Journal of the Scientific Agricultural Society of Finland
Vol. 48:427-521, 1976
Maataloustieteellinen Aikakauskirja

AN ECONOMETRIC MODEL FOR THE FINNISH
EGG INDUSTRY
Selostus: Suomen kananmunasektoria kuvaava ekonometrinen, malli

MARKKU NEVALA

Agricultural Economics Research Institute,
Rukkila, SF-00001 Helsinki 100, Finland

To BE PRESENTED, WITH THE PERMISSION OF THE FACULTY
of Agriculture and Forestry of the University of

Helsinki, for public criticism in Auditorium xii on
February 25, 1977 at 12 o’clock

SUOMEN MAATALOUSTIETEELLINEN SEURA HELSINKI

https://www.c-info.fi/en/info/?token=RlFkjkWz8tCAtJ81.oZdtlsZMFqlb0YaWG2ej4A.mGuh3xAYiMylKcxtqWyTkFnxocpDJdVpDhEqiGlCWHCVGmQ8Q3DeatvgkcuuCwIbFysgpl0nyXXXRF3MtzPIzu1yhuf10lkE41tLsTeRIvnD88ocT8b--mAquk2KxUZWcNXghbZ2v-6THAlFmjBTA_lobMLL5y0eK1gl




Preface

The present study has been conducted mainly at the Agricultural Economics Research
Institute. As to be expected, my acknowledgements of indeltness on completing this study
are substantial, and I wish to acknowledge my debt of gratitude to all those who have contri-
buted to its completion.

First of all, I wish to express my gratitude to Professor Matias Torvela, Head of the Agri-
cultural Economics Research Institute, for the support and encouragement he has given me
firstly in this research project, and secondly in my entire study program during my employment
at the Institute. It is also my pleasure to express my sincerest thanks to Professor Lauri Ket-
tunen, Head of the Marketing Research Department of the Institute, for his continual guidance
and undivided interest throughout the development of this study, and for the comments he
has made on the first drafts of the manuscript.

I also owe a particular debt of gratitude to my teacher, Professor Risto Ihamuotila, for
all the constructive suggestions and support he has given me not only in this research but also
in my studies at the University of Helsinki. He also read the manuscript and made many useful
comments. My sincerest thanks are also due to Professor K. U. Pihkala for discussions I have
had with him Concerning this study.

A special debt of gratitude is due to Professor K. L. Robinson, of Cornell University, who
gave me valuable guidance especially on the methods of analyzing agricultural price policy in
the final stage of this study. In addition, he was kind enough to read the entire manuscript
to provide me with detailed suggestions that have led to substantial improvements.

I wish to express my thanks to Dr. Lulu Siltanen, Mr. Erik Haggren, Mr. Juhani Rou-
hiainen, and Mr. Jouko sir£n, who have helped in different ways with the present manus-
cript, and to Mr. Clive A. Volk, who checked the English text. Also, I would acknowledge
the patience and skill of Mrs. Patricia Wells and Mrs. Ritva Lindroth, who typed the dif-
ferent drafts of the manuscript.

My thanks go also to the Academy of Finland and to the W. K. Kellogg Foundation for
their financial support of this study. I am especially grateful for the grant from the W. K.
Kellogg Foundation which provided me with an opportunity to complete this study during
ray visit to Cornell University in the academic year 1975 76. lam also grateful to the Scien-
tific Agricultural Society of Finland for including this study in its series of publications.

As always, my greatest debt is to my children, Samuli and Hannaleena, who in the final
phase of the study kept asking me whether I had already finished the manuscript or not. To
them I wish to dedicate this study.

Helsinki. October 1976
Markku Nevala





TABLE OF CONTENTS

1. INTRODUCTION 433
1.1. Background to the Problem 433
1. 2. Objectives of the Study 439

2. ECONOMETRIC MODEL BUILDING 442
2, 1. On the Methods for Assessing the Impact of Alternative Policies on Agricultural

Production 442
2. 2. General Framework for Econometric Model Building 444
2.3. Basic Model 447
2. 4. Characteristics of the Basic Model 454
2. 5. Recursive Production Response Model 457

3. DATA AND ESTIMATION METHODS 458
3. 1, Estimation Period 458
3. 2. Elimination of the Seasonal Variation 459
3.3. Estimation Methods 460

3.3.1. General 460
3. 3. 2. Polynomial Lag Formulations 462

4. ESTIMATION OF THE MODELS 463
4. 1. Recursive Production Response Model 463

4. 1. 1. Estimates from the Period 1956—74 463
4. 1.2. Variations in the Coefficients of the Recursive Production Response

Model in Time 467
4. 2. Basic Model; Estimates from the Period 1956—70 469

5. EVALUATION OF THE BASIC MODEL 472
5. 1. On the Methods for Evaluating Econometric Models 472
5. 2. Historical Simulation 474
5. 3. Ex-Post Forecast Simulation 478

6. PRICE POLICY ANALYSIS FOR THE PERIOD 1956-70 BY MEANS OF THE
BASIC MODEL 480
6. 1. General Framework 480
6. 2. Derivation of the Equilibrium Policy Alternative 483
6. 3. Development of the Egg Industry Under the Alternative Price Policy Directions 485
6. 4. Effectiveness of the Policy Alternatives 487

7. EX-POST SIMULATION OF ALTERNATIVE PRICE POLICY DIRECTIONS AND
THEIR EVALUATION 491
7. 1. Policy Alternatives and the Simulated Outcomes of These Policies 491
7. 2. Evaluation of the Effectiveness of the Policy Alternatives in Attaining

Policy Goals 494
S. SUMMARY AND CONCLUSIONS 499

REFERENCES 503

SELOSTUS 506
APPENDICES 508





433

JOURNAL OF THE SCIENTIFIC AGRICULTURAL SOCIETY OF FINLAND
Maataloustieteellinen A ikakauskirja

Vol. 48: 427-521 . 1976

Nevala, M. 1976. An Econometric Model for the Finnish Egg Industry.
J. Scient. Agric. Soc. Finl. 48:427 521.

Abstract. An econometric model for the Finnish egg industry was derived in this
study to provide information concerning the impact of alternative price policies on the
development of the egg industry for price policy analyses. The basic model specified
in this study is an eight-equation model including distributed lag formulations and consist-
ing of three different blocks. The basic assumption behind the model specification is
that the price levels of eggs are mainly determined by the target price level (set up by
the government) and the domestic market situation. Producers are assumed to respond
to changes in the producer price of eggs and in the other factors influencing the pro-
fitability of egg production. The basic model was estimated on the basis of semiannual
observations from period 1956 70. In addition, the stability of the coefficients of pro-
duction equations in the model over the years was tested by using the stepwise reg-
ression method. The structure of the basic model was evaluated by means of deterministic
simulations to gain some idea of the model’s ability to simulate the actual development
of the egg industry. Both historical simulations and ex-post forecast simulations were
conducted and the »goodness of fit» was tested by the use of Theil’s inequality coeffi-
cients and graphical examinations.

The basic model was also used for analyzing the price policy pursued by the govern-
ment in period 1956 70 in order to illustrate the type of analyses that can be conducted
by the basic model. An alternative policy mix consisting of the target price system for
maintaining the domestic market balance and the direct payment scheme for attaining
the income target of producers was defined. Given the assumptions made in this study
this policy mix would have been a more effective policy tool from the standpoint of
society for attaining the policy goals than the target price system as implemented in
that period. Similar conclusions can be also drawn from the results of the ex-post price
policy simulations, in which the target price level was assumed to be used for adjusting
the domestic egg production gradually to the domestic consumption. Also in this alterna-
tive the government was assumed to bear the burden of attaining the income target of
producers in the system by the use of direct payments. However, we must point out
that results depend on the assumptions necessary for computations of this type.

1. Introduction

1. 1. Background to the Problem

According to the total accounts of agriculture the value of annual egg pro-
duction in Finland at the producer price level has somewhat more than quad-
rupled during the last two decades: from some 70 mil. Fmk in 1956 to 328 mil.



434

Fmk in 1974.1 The growth in the total value is due to the fact that both the pro-
duction volume and producer price of eggs has approximately doubled during
this period. This kind of development has also increased the relative importance
of eggs within the agricultural production. Eggs accounted for some six per-
cent of the total value of agricultural production in 1974, while the correspond-
ing proportion in 1956 was some four percent.

The average growth rate of egg production during the above mentioned
period was 4.2 percent per year computed by using a least squares exponential
curve fit. 2 Accordingly, egg production has increased from 35.5 mil. kg to
77 mil. kg (See Appendix 2). The annual growth rate of production has not,
however, been stable but varied from year to year (Figure 1.1). Yearly increases
in production were relatively high in the late 1960’s and in the early 1970’5:
for example, in the period 1969—1974 the production grew at a yearly rate of
6.7 percent. On the other hand, the domestic consumption of eggs has increased
at a slower rate than the egg production. The per capita consumption has
during the last two decades risen by some 3 kg. Accordingly, the total consump-
tion increased by some 17 mil. kg within this period: from 33 mil. kg to 50
mil. kg. The average growth rate of egg consumption was 2.5 percent per year
computed from the period 1956—74 by using the exponential curve fit.

This kind of development has also led to increases in the egg surpluses on
the domestic market. In the early 1950’s the domestic egg production and
consumption were almost in balance: the level of self-sufficiency for eggs varied
in the range of 99 and 106 percent in these years. Since then, the surpluses on
the domestic markets have increased at varying annual rates such that in the
early 1970’s the domestic production exceeded the domestic consumption by
some 50—57 percent. 3 As can be seen in Figure 1.1, the egg exports amounted
to some 27 mil. kg in the years 1973 and 1974.

There are, of course, many reasons for the growth of the total egg production
described above. First of all, it seems reasonable to argue that recent changes
in the structure of the egg production sector and price development are the main
factors for the continuity of production increases. Especially the structure of
egg production has changed drastically during the last two decades. Impro-
vements in production technology have been remarkable and they have contri-
buted to changes in the producing sector towards bigger and more efficient
producing units. Flygare (1972, p. 252) states that this phenomenon is univer-
sal and hardly in any other sector in agriculture has the structural change
been as rapid as in the egg industry.

The graphs in Figure 1.2 illustrate the structural change which has taken
place in the Finnish egg industry in period 1950 74. According to these estima-
tes, for example, some 55 percent of the total number of laying hens belonged

1 These figures are based on the total accounts of agriculture prepared by the Agricultural
Economics Research Institute. Finland.

2 The formula used is: y = be mx, where y = yearly production of eggs, x = trend
(1, 2,3, . . .), b and m are the coefficients to be estimated.

3 In this context it is also worth mentioning that in the 1920’s and 1930’s the level of self-
sufficiency for eggs was, on the average, even higher than in the recent years. In some years
of the 1930’s over half of the total egg production was exported.



435

to producing units larger than 500 hens in the year 1974, while the corresponding
percentage in the year 1959 was some 2—3 percent. Thus, egg production on
a small scale on family farms has decreased significantly: while about 82 per-

Figure 1.1. Production, Net Exports and Producer Price Index of Eggs as Well as the
Producer Price Index of Agricultural Products (1956/57 = 100) in 1956 74.

Figure 1.2. The Distribution of Laying Hens Between Different Size Categories
of Producing Units. Estimates are Based on the Censuses of Agriculture. Sources:
The Official Statistics of Finland III: 45, 53, 66 and Unpublished Data Prepared
by the Board of Agriculture.



436

cent of the layers in the year 1950 belonged to producing units smaller than a
hundred laying hens, in the year 1971 the share of this size gategory accounted
only for some 18 percent of the total number of laying hens. These graphs
also clearly indicate that the structural change has been especially rapid in the
late 1960’s and early 1970’5. The recent concentration of egg production can
also be illustrated by the fact that 23 percent of laying hens in the year 1971
belonged to some 600 producing units, which accounted only for some 1.2
percent of the total number of the producing units in the country.

As to the development of egg prices, the recent trend has been based mainly
on the government’s intervention in the egg market in order to regulate the
price level. To attain its agricultural policy goals the Finnish government
adopted the so-called target price system for the main agricultural products
in the 1950’5. Since crop year 1956/57 this program has been implemented
under several Agricultural Price Acts. The contents of the Price Acts have
been changed to some extent in the course of time.1 However, the main policy
tool the target price system has remained the same as have also the two
main objectives of the price policy (given explicitly or implicitly in the Acts):
1) to promote development of incomes for agricultural producers to a level

comparable with the other sectors of the economy, and 2) to stabilize prices in
order to maintain a level of self-sufficiency for the products which can be pro-
duced within the country at reasonable costs.

In this system the target prices for the main agricultural products are set
annually such that the income objective can be attained. However, the general
principle common to all the past Acts is the fact that only the average
income level of producers is, in one way or other, tied to the development of
the incomes of other population groups. Accordingly, the policy makers are
relatively free to change the target prices for individual products: they can be
raised proportionately or disproportionately to the change in the average target
price level required. This kind of flexibility is mainly due to the government’s
desire to guide the allocation of resources in agriculture according to market
considerations. On the other hand, this freedom enables policy makers to take
into account the possible differences in the development of the production costs
of different products when fixing target price levels for a given period. Within
this framework, the government has achieved and maintained the target price
levels by paying export subsidies to exporters so as to enable them to dispose
of surpluses on world markets if production exceeds the domestic needs of the
target price, and allowing imports if the producer price is over the target price
level. Every Price Act has included specific provisions on the conditions which
must be met before the government intervenes in the market. For most pro-
ducts (for example, eggs) the producer prices have been allowed to fluctuate
within ± 5 percent range around the target price level without state interven-
tion. In this system domestic production is protected against relatively low
world market prices for agricultural products by means of variable import
levies.

1 Good descriptions of the contents of different Price Acts are given by Sauli {1971, p.
48 71), Kettunen (1972, p. 135 150) and Ihamuotila (1972 c, p. 10 11).



437

There is no question about the basic applicability of this program in support-
ing agricultural prices in Finland. Under certain conditions the system can serve
as a very effective policy measure (See Gulbrandsen and Lindbeck 1973,
p. 245). However, measuring the necessary changes in the target prices of
different products such that the basic policy goals could be attained, seems to
have been difficult in the past for many different reasons. For some products,
setting the target price levels has been successful in this respect, while for other
products it has been difficult to find proper target price levels.

Eggs have, without doubt, been one of the products for which measuring the
necessary changes in the target price level has been problematic as long as the
target price system in its current form has been implemented under the Agri-
cultural Price Acts. Adopting that policy program has certainly stabilized the
fluctuations of egg prices which were typical and frequent before the interven-
tion of the government in the egg market (See Kaarlehto 1964, p. 382—390).
Because of the lack of appropriate data, it is impossible to explore how the
incomes of the egg producers have developed under this price policy. 1 However,
increases in the total egg production imply that production of eggs at least
in certain producing units must have been profitable relative to the other pro-
duction branches of agriculture. Similarly, some book-keeping results from the
year 1974 indicate that the profitability of the poultry farms was better than
the average profitability in this farm group (See Anon. 1974, p. 32).

As to the efforts to maintain a reasonable balance between domestic pro-
duction and consumption, the past price policy seems as we shall argue
to have been rather unsuccessful in this respect. The domestic production of
eggs has, as stated, increased under the price policy pursued in 1956—74 at
varying yearly rates as shown in Figure 1.1. This has happened in spite of the
fact that yearly increases in egg prices have, on the average, been smaller than
increases in the average price level of agricultural produce (Figure 1.1). It seems
obvious that one of the most difficult problems in fixing target price levels for
eggs has been assessing the impact of the remarkable structural change of egg
production on a necessary increase in the target price level of eggs in a given
period. Increases in productivity have resulted in improvements in the compe-
titive position of egg production in spite of smaller increases in the target price
of eggs than on the average in other agricultural product prices. Due to relati-
vely free entry into the industry, big firms using the best technology available

have been set up; thereby, increasing the production potential (Figure 1.2).
On the other hand, egg production on a small scale on family farms has not
decreased sufficiently to outweigh this kind of growth in production potential.
That is why, an increasing proportion of the total egg production has been
exported annually at relatively low world market prices.

Since egg prices on the world market have remained at a relatively low level
during the last two decades (See Appendix 2), the growth in export volume of
eggs has led to increases in government expenditure on export subsidies (See

1 In this connection we have to point out that there are studies on the profitability of egg
production in the last two decades conducted by Westermarck. However, these results may
not be generalized since they are not based on a random sample of the Finnish egg industry
(See Westermarck 1975 and Torvela 1959).



438

Appendix 2). The problem can be illustrated by the graphs in Figure 1.3, which
indicate the value of the total egg production at the retail price level and how
it is composed of different items in period 1956—74.1 As can be seen from the
figure, the domestic consumption accounted for some 95 percent of the total
value of egg sales in 1956, whereas in 1973—74 the same proportion was only
some 63 —65 percent. Egg exports in the early 1970’s amounted to some 34
36 percent of the total production volume. However, in the same time period
total export revenues from eggs accounted only for some 14 percent of the total
sales of eggs. Accordingly, export subsidies by the government have increased
substantially: in the years 1973 and 1974 they amounted to 21—23 percent of
the total value of egg production.

Because of failure in maintaining a reasonable balance between production
and consumption by the use of the target price system, some additional measu-
res have been adopted recently to curb the growth of egg production (Anon.
1971 b, 1975, 1976). Since they are assumed to have exerted only a minor in-
fluence on the egg industry during the period this study covers we will not
discuss them in more detail here. It may, however, be mentioned that the pur-
pose of these measures is to protect the continuity of egg production on family-
farms by restricting the freedom to set up new large producing units. Also a

1 The total value of egg production has been computed by adding the export subsidies paid
by the government and export revenues to the value of domestic consumption (including the
home consumption on farms) at the retail price level.

Figure 1.3. The Estimated Total Value of Egg Production at the Retail Price Level and the
Components of the Total Value in 1956 74, mil. Fmk.



439

progressive marketing fee is levied on the production of large producing units
(over 5 000 laying hens) to curb the egg production of these firms. As a tempo-
rary measure for reducing the egg production the government adopted a
slaughter levy scheme. On the other hand, the government has also supported
the advertisement campaigns for eggs to increase their consumption.

1.2. Objectives of the Study
The primary objective of this study is to explore the possibilities of generating

information by means of an econometric model to help the policy makers in
adopting price policy directions for eggs which would be consistent with our
major policy goals. If we regard attaining the producers’ income target and
the reasonable level of self-sufficiency in eggs (Section 1.1) as our main po-
licy goals, we find that past experiences indicate that it has been extremely
difficult to reach both of these goals simultaneously by employing only with
the target price system. Mainly due to the low mobility of agricultural resources,
the policy makers have been most commonly faced by conflicting decision al-
ternatives concerning the future target price levels: the necessary change for
attaining a better balance between domestic production and consumption
(or maintaining domestic market equilibrium) conflicts at least in the short
run with the change required for achieving the income objective of egg
producers. Since the income objective established by law has been the
most important, the target price levels have been fixed mainly according to
this kind of consideration in the past; thereby, leading to serious surplus prob-
lems as stated previously.

At the same time the role of the target price system in the case of eggs
as we shall argue has changed: it no longer serves as a mere stabilization
vehicle by which the government eliminates temporary downward and upward
fluctuations of egg prices to stabilize the domestic egg market as orginally
implicitly planned when the system was introduced for supporting egg producers.
Nowadays it can rather be regarded as an income support program the govern-
ment applies to the egg industry. 1 However, it is not without question clear
that this kind of program is the most effective and best policy instrument
available for supporting incomes of producers from the standpoint of the whole
of society in the case of the egg industry. There are many other policy tools,
which are, in theory, available to the government and which may be more
effective in attaining our major policy goals than the target price system as
implemented for the egg industry during the last two decades.

Selecting an optimal policy mix for a certain agricultural product is a com-
prehensive and complex issue influenced by many factors and considerations.

1 We know that this kind of classification of policy programs is a simplified view of reality,
because the prices and producers’ incomes are positively related to each other. However, by
making this distinction we want to emphasize the fact that given a certain flexibility the
target prices could, at least in theory, have been fixed at such a level that the target price system
would have functioned as a mere price stabilization vehicle associated only with minor exports
and imports of eggs.



440

which are not discussed in more detail in this context. 1 In this study we con-
centrate on only a few concrete policy alternatives which we believe
can be considered as alternatives to the actual price policy of eggs which have
been pursued by the government. We recognize that the policy makers cer-
tainly need more information on the probable outcomes and consequences of
a certain policy action in order to implement effectively the alternative policies
than they necessarily need to employ with the target price system in its current
form. It is one of the main purposes of this study to provide the policy makers
with an aid to generate this kind of information for the basis of their policy
decisions in order that they may avoid major errors in target price fixing, which
may in the long run lead to structural bias in the producing sector: an econo-
metric model based on the data currently available and including the necessary
policy instruments as its variables.

To specify our policy alternatives we assume that there are two major policy
goals, which have to be attained under alternative price policies: 1) the income
objective as defined in the previous chapter and 2) maintaining a reasonable
balance between the domestic egg production and consumption. We assume that
the target price system should be used only as a stabilization tool; i.e. for
maintaining (and attaining) the balance between domestic production and
consumption so that the target price of eggs is always attempted to set at a
level which would lead to a reasonable market equilibrium in future periods.
If the income target of egg producers is not reached at this price level, we as-
sume that the government makes up the difference between the actual and
target income levels in the form of direct cash payments to the producers em-
ployed in the egg industry. In order that this scheme be effective, payments
to individual producers must not be related to the production volumes of their
firms or there must be a certain upper limit in the payment to an individual
producer.

This kind of policy mix can be considered as a plausible policy alternative
especially in the case of the egg industry due to the current development of the
structure of the egg industry. On one hand, the number of big »industrialized»
producing units, where production costs are obviously relatively low compared
with the average level in the sector as a whole (and also with the level of the
target price), have been increasing during the last few years. On the other hand,
however, a great proportion of the total egg production is still produced on small
family farms. It seems reasonable to argue that, because of the relatively few
big firms, the target price system in its current form is no more a plausible
and fair vehicle for supporting the incomes of egg producers, because a great
deal of the support actually goes to these big firms which may not need that
much support.

In this study we will evaluate empirically the comparative effectiveness
of the policy mix mentioned above and the actual price policy implemented
by the government through the target price system in attaining the two policy
goals specified above. The main question in the following analyses is whether

1 We only suggest consulting the studies of Nerlove (1958, p. 222 225), Josling (1969,
p. 175 191) or Gulbrandsen and Lindbeck (1973, p. 243 247).



441

adopting this policy mix instead of the actual price policy would have resulted
in any net savings to society as a whole proceeding from different situations in
the egg industry in the past.

Accordingly, to complete these research objectives our intentions in this
study are to:

1) construct econometric models which can be used for analyzing and forecast-
ing the impact of changes in the values of the policy instruments of the
government (the target price) and other socio-economic forces faced by
the egg producers in a given period. Due to its crucial importance in the
following price policy analyses, relatively much attention is paid to the
specification and estimation of the model in this study. For the same reason,
our intention is also to evaluate the model’s validity by the use of different
tests in order to get some ideas about the model’s ability to simulate the
behavior of the real system (Sections 2 —5).

2) conduct historical price policy simulations with this model for analyzing
the effectiveness of the policy mix defined above compared with the actual
price policy in period 1956—1970. In this attempt, we will use the model
to find the target price path which would have kept the domestic
production and consumption of eggs in reasonable balance during that
period. The evaluation of these policy alternatives from the standpoint of
society is then based on the values generated by the model (Section 6).

3) perform ex-post price policy simulation with the model in order to show
the nature of the policy analyses that can be made by means of this model
in a »real» decision situation and, also, to evaluate the effectiveness of
different policy alternatives in period 1971—74 compared with the actual
price policy (Section 7).

There are some earlier econometric studies on the Finnish egg industry.
The main attention in most studies has been focused on estimating the pro-
duction response of eggs or the price and income elasticities of demand for
eggs only (Pihkala 1941,Kaarlehto 1958 and 1961) or on explaining variations
in the marketing margins of eggs (Waananen and Kaarlehto 1965). The
only multi-equation model for the egg industry in Finland taking into account
the most important phenomena of the sector has been developed by Koponen

(1964). However, since these studies are based on relatively remote time periods,
their results cannot directly be used for the purposes of this study. Nevertheless,
these studies provided valuable a priori information on the relationships be-
tween variables in the system for building models in this study.

As for the problem formulation and methodology used, this study has
some similaritieswith those made by Agarvala and Ball (1970) and Chayat et.
al. (1974). The former study deals with econometric model building for the egg
market in the U.K. and with policy analyses using this model (the policy instru-
ment of the Egg Marketing Board is the proportion of the total supply that is
sold on the shell market). The latter study is an interesting econometric deter-
mination of the welfare impact of giving bargaining power to egg producers
(so that they could stabilize the market prices).



442

In addition, many other interesting ideas and procedures in the art of model
building for the egg market can be found in the recent literature. It may not
be necessary to describe these models in more detail here. Building an econo-
metric model for whatever purpose is usually a form of tailoring: the individual
solutions usually depend on the use of the model (See Niitamo and Soivio
1964, p. 137), institutional circumstances in the country, data availabilities,
etc. However, in this context it is worth mentioning Gerra’s (1959 a and
1959 b) and Fisher’s (1958) works, which are among the earliest efforts to
quantify the reactions of egg producers and consumers by means of multi-
equation models. More recently, Hartman (1974) identified an egg cycle in the
U.S. by use of spectral analysis and built different kinds of recursive models
for explaining this cycle. The study by Bange and Bender (1969) can be
mentioned among the recent demand analyses for eggs. Interesting applications
of econometric models and other methods for forecasting egg production, con-
sumption and prices can also be found in the studies of Roy and Johnson
(1973), Miller and Masters (1973), Stojkovic (1971), Winfridson (1974) 1

,

Bröcker (1973) and Kersten (1973).

2. Econometric Model Building

2. 1. On the Methods for Assessing the Impact of Alternative Policies on
Agricultural Production

One of the primary objectives of this study is to build models for simulating
and forecasting the reactions of egg producers and consumers to changes in
various economic forces, among which the price level of eggs as a policy instru-
ment of the government is the most important. For predicting and simulat-
ing agricultural production responses, several kinds of techniques have been
developed and tested in empirical analyses in the past. In this study, an econ-
ometric model approach was chosen for assessing impact of alternative policy
directions on the egg industry. The alternative methods fcr simulating the
reactions of egg producers are different kinds of programming models, such
as representative farm programming and recursive programming.

Much attention was paid to representative farm programming 2 (i.e. deriv-
ing aggregate supply response from the supply functions of the representative
farms) in the estimation of aggregate supply response in the 1960’5. In spite
of an appealing idea to predict aggregate response via the micro route the
empirical applications of this method were not very successful (See Sharples

1 This is a good example from among the econometric models built for short-term forecast-
ing of egg production.

2 Applications of this method and discussion about its applicability can be found for example
in the articles by Sharples (1969, p. 353 361), Sheehy and McAlexander (1965, p. 681
695), Frick and Andrews (1965, p. 696 700). The study of Wipe and Bawden (1969, p.
170 178), in which the aggregate supply curve is derived from the production functions of
single firms, also, belongs to this type of approach to aggregate supply analyses.



443

1969, p. 360). According to Barker and Stanton (1965, p. 711) and Sharples

(1969, p. 355), the main difficulties of this approach can be classified as follows:
1) interdependence or externalities, 2) changes in farm size, 3) unrealistic firm-
level assumptions, 4) the selection of representative farms, and 5) mechanical
problems. They considered many of these problems very difficult ones to solve
and, accordingly, Sharples (1969, p. 360) was pessimistic wich respect to the
applicability of this method stating that »what we do need (instead of this app-
roach) is to further explore less cumbersome models that identify and better
utilize stable aggregate relationships».

Recursive programming (the introduction of flexibility constraints into
regular programming models) was initially used as an alternative when efforts
to build all the behavioral constraints explicitly into the model led to unsatis-
factory results. 1 This technique has, as stated by many economists,2 certain
advantages over econometric models: 1) it is more effective in dealing with
dramatic changes in production technology or in agricultural policies, 2) data
problems at least in regional supply analyses are easier to solve when
using programming models, 3 3) econometric analysis may lead to forecasts
that are inconsistent with resource availabilities, while a constrained optimi-
zation framework such as resursive programming might solve this dilemma,
4) the use of recursive programming gives the analyst a better understanding
of the reasons for changes in production than econometric analysis, etc. How-
ever, building plausible flexibility constraints into the model has proved to be
difficult. This dilemma has led to more complex methods (such as that of
Sahi and Craddock 1974) in estimating these coefficients, thereby, making the
use of the model more difficult.

In the past studies, econometric production response models have performed
somewhat better than recursive programming models (See, for example, Schal-
ler and Dean 1965, p. viii) and, that is why, the econometric model approach
was chosen in this study. However, as some economists have pointed out,
neither method has provided completely satisfactory results. The most dif-
ficult problems in production response analysis via econometric models are
according to Nerlove (1961, p. 31): 1) the complex structure of production,
2) technological change, 3) aggregation, 4) investment in fixed and quasifixed
factors and 5) uncertainty and expectations. No general solutions to these
problems have been found thus far and, therefore, this study also involves these
shortcomings. Individual solutions to these problems made in this study are
discussed in the following sections.

1 The pioneering work was made by Day (1961 and 1962) in the field of recursive programm-
ing. Among the most interesting applications of this method is that of Schaller and Dean
(1965) for prediction of regional crop production. More recently, Sahi and Craddock (1974,
p. 344 350) have made their contribution in developing recursive programming by introducing
new techniques of estimating flexibility coefficients.

2 For example, Heady (1961, p. 16 19), Day (1962, p, 139 148), Schaller and Dean
(1965).

3 However, recent applications (for example, that by Sahi and Craddock) use time series
data for estimating the flexibility coefficients of the recursive programming model.Thus, this
advantage becomes less important when discussing more complex recursive programming models.



444

2. 2, General Framework for Econometric Model Building
Before we start building an econometric model for evaluating the effects

of alternative price policies on the behavior of the egg industry, one of the most
important questions to be solved from the methodological standpoint is the
definition of our policy problem. This relates closely to the question of the
kind of information we are able to generate for the policy maker by means of
an econometric model, and how reliable and useful this kind of information is.

As stated by Naylor (1970, p. 263—265), three alternative approaches
have been suggested by economists for using macro-econometric models for
policy analyses; each producing different kinds of information for policy decis-
ions: 1) the Theil approach, 2) the Tinbergen approach, and 3) the policy si-
mulation approach. Each of these approaches assumes that we start with a
given econometric model. In the Theil approach the policy problem is defined
as one of maximizing a welfare function which contains a set of policy goals
ranked by the decision maker and which is subject to the constraints imposed
by the model. The Tinbergen approach assumes that the policy maker has
specified a fixed target value for each of the endogenous variables and, then,
for given values of exogenous and lagged endogenous variables, the equations
of the model are solved for the set of values of the policy variables that are
consistent with the targets (the policy instruments are endogenous variables
in the model). On the other hand, the policy simulation method determines
values for the endogenous variables from the reduced form or structural form
of the model based on the given values of the policy instruments and other
predetermined variables (thus, the policy instruments are exogenous variables
in this approach).

Theoretically, all of these approaches are also applicable in evaluating the
impact of policy alternatives on a commodity economy (See Labys 1973,
p. 113) like the Finnish egg industry. As to their usefulness for real world
policy analyses, empirical evidence, however, suggests that the policy simulation
approach may represent the only methodology currently available for policy
analyses (Naylor 1970, p. 265). Although Theil-Tinbergen approaches may
be appealing to economists from a purely theoretical standpoint, they have
not, in general, provided any operational solutions to the policy problems
because of many difficulties in applying these methods. 1

Accordingly, the policy simulation method was chosen in this study for the
reasons mentioned above. Policy instruments the target price of eggs in
this study are exogenous variables in the model, and we can generate the
time paths of the endogenous variables (those variables that the policy maker
is assumed to be interested in) for any given values of these policy instruments
proposed by the policy maker. On the other hand, we may ourselves analyze
some policy alternatives, and based on these analyses we may propose a few

1 The Theil approach suffers from one major shortcoming: in the real world we do not know
the functional form and parameters of the social welfare function of the policy maker. The main
problem in applying the Tinbergen approach is that we have to specify the targets of the policy
maker explicitly in the model and, thus, before we can start model building we have to try to
find out what the goals of the policy maker are. In general, this task has proved to be difficult.



445

policy directions of our own which to our mind seem to be appropriate
for consideration by the policy maker. Thus, since we can with simulation
show the consequences of different policy directions, we can help the policy
maker to select the policy that is most compatible with his preferences, mini-
mize conflicts, etc.

Econometric model building for the production response of eggs is based
on the assumption that producers react to changes in the profitability of pro-
duction when planning the output volumes of their firms at different points
in time. One of the main determinants of the profitability of egg production
is the producer price level of eggs, which can be controlled by the government
through the target price system. In addition, consumers are assumed to react
to changes in the retail price of eggs and change the consumption level of eggs
according to changes in the price level of eggs and other socio-economic foices.
This is the general framework for building an econometric model for policy
analyses defined above.

As to the production response of eggs, the total change of egg production
in a given period depends on how producers as a single group react to
changes in the economic circumstances where they operate. On the basis of
the biological characteristics of egg production it appears reasonable to argue
that there are three major production decisions through which the producers
are able to influence the production volumes of their firms. These decisions,
which producers make at different stages of the production process, deal with

1) purchasing chickens
2) culling layers
3) changing efficiency to influence the output per layer.

It seems reasonable to assume that identifying and assessing producers’
reactions in these three decision situations separately will serve as a plausible
point of departure for building an econometric model for egg production respon-
se. Accordingly, a behavioral equation, describing producers’ reactions to
changes in the values of the likely decision variables, will be specified for each
decision situation in this study. The total response of egg production is then
derived from these equations through an identity. For these reasons, the single
equation approach (such as that of Chayat et. al. 1974, p. 5—6) for describing
production response of eggs is not applicable in this study. We believe that
this approach has some advantages over the typical single-equation production
response model. First of all, it may give us a better understanding of the reasons
for changes in the production levels and, in addition, estimating the production
response may be easier in this approach due to greater variations in the values
of the above mentioned variables than in the total egg production.

In this context it will be pointed out that the potential chances of egg produc-
ers to change the output in these decision situations are not similar due to the
differences between economic alternatives faced by the producers in each of
these decision situations. It is clear that producers have the greatest chances
to adjust production volumes of their firms according to their targets when
they plan to purchase baby chicks: new building capacity can be built or —in



446

the opposite case the existing capacity may be utilized more inefficiently
than earlier. On the other hand, this decision is also the most difficult to pro-,
ducers. Since raising baby chicks to the laying maturity takes some 56 months,
in the decision situation producers do not know what the prices of eggs and
production inputs will be in the period when these chickens reach the laying
stage. That is why, they have to base their decisions on expected profitability.

On the other hand, culling of layers is always directed to the existing laying
flock. A priori, we can assume that by changing the culling pattern between
periods producers can to a certain extent influence the total output of eggs.
However, taking into account the economic and biological aspects, we can
argue that the potential changes of total egg output through changing culling
rates are smaller than those through the decisions dealing with purchasing
baby chicks. Similarly, we can a priori suppose that the importance of the
third type of decision (changing output per layer) in changing the total output
of eggs in the short run is, in practice, negligible (See Section 2.3).

Within this framework, a typical iterative procedure of econometric model
building for a certain type of policy analyses has been followed in this study.
This procedure, which has been described, for example, by Naylor (1971,
p. 11—36) in an illustrative way, can from the technical point of view be divid-
ed into subsequent stages as follows:

Y I r
Specification >- Estimation >■ Evaluation >. Final use

| A -f-
Proceeding from the problem formulation and model specification, the model

goes through the stages of estimation and evaluation towards the »final» model.
Each successive stage provides feedback information to the researcher for
improving the decisions made at the previous stages. There is no exact scienti-
fic criterion for choosing the best model structure among many alternatives.
Therefore, the iterative rounds described above can be thought to continue
almost endlessly; new data will be published and, on the other hand, the econo-
metric theory tends to make progress in the course of times. The rule: the best
possible explanation or results with the minimum number of explanatory
variables which Box and Jenkins (1970, p. 17) call the principle of parsimo-
ny in model building may provide some guidance in the different stages of
the procedure. On the other hand, the ease of computing and forecasting values
of exogenous variables may be another criterion in choosing explanatory va-
riables. This usually leads to the introduction of lagged endogenous variables
into the model and recursive ordering of the equations of the model.

To start the procedure described above, the model specifications developed
in the present study are introduced in the following section with short comments
on each equation. We have to point out that the model as presented in this
chapter has already undergone several iterative rounds of model building.
However, we have to bear in mind when evaluating the model that these



447

models like econometric models in general include only the most important
variables and relationships of the actual economic system whose behavior it
attempts to explain, namely, the Finnish egg industry.

2. 3. Basic Model

The basic model which emerged from this study was developed for semi-
annual data.1 The main reason for this was the availability of data; forseveial
key variables in the model the observational values are available only on a
semiannual basis. On the other hand, taking into account the characteristics
of the egg production process, it seems reasonable to argue that a semiannual
model is rather plausible in analyzing producers’ responses for policy purposes.

The specification of the model to be tested in the following stages of
estimation and evaluation can be presented as follows (the sign of each
coefficient reflects a priori expectations of the logical signs):

Endogenous Variables of the Basic Model

SPt
= Number of chickens at the end of period t.

KK t = Number of layers culled in period t.
SK t = Number of layers at the end of period t.
TSt = Total production of eggs in period t.
PKTt = Producer price of eggs in period t.
PKV[ = Retail price of eggs in period t.
COt = Consumption of eggs per capita in period t.
UK( = Net exports of eggs in period t.

Exogenous Variables of the Basic Model
PR t = Price of commercial feed in period t.
RVS t = Total availability of feedgrain in period t.
TAVt = Target price of eggs set by the government in period t.
Yt = Consumers’ disposable incomes in period t.
YM t = Wage index in commerce in period t.
F.I t = Consumer price index in period t.
T = Time trend (1952/1 = 9 . . . 1974/1 = 54).
H t = Total population in period t.
D 2 = Seasonal dummy. D 2 = 1 in the second half of the year, otherwise, D 2 = 0.
D 2T = Product of variables D 2 and T
a and b = Coefficients to be estimated.
Uj Random variable.

Equations of the Basic Model

Number of chickens at the end of period t;
n—l n—l

(2.1) SP t =a, + 2 b ni PKTt-i “ J bl2i PR‘-i + b^T - b i4D 2
- bl5D 2T +u,

i=o i =0

Number of layers culled in period t
n— l

(2.2) KK t =a2 + b 2li SP t_k _; b 22PKT t + b23PR t - b 24RVS t + b25 +u 2
i =0

1 Selecting the time unit of observations is a relevant aspect in the process of model building,
since the length of observation period usually influences the specification and nature of the
model (See Theil 1971, p. 462).



448

Number of layers at the end of period t (identity):

(2.3) SKt = SKt_j 4- SP t_!
- KKt

Egg production in period t;

(2.4) TS t a4 + b 4l (SK t_j + 0.5SP t 0.5KKt ) -f- b42 l -f- b 43 -f- u 4
Producer price of eggs in period t:

(2.5) PKTt =a s +b M TAV - b 52 (UK/H), + b53 +u5

Retail price of eggs in period t;

(2.6) PK\ t a 6 + b6iPKl t H“ b 62YM t + b63 +ue

Consumption of eggs in period t:

(2.7) COt =a, - b 7l (PKV/El) t + b72 (Y/ET) t + b7iD 2 +u 7

Net exports in period t:

(2.8) UK t = TS t - (CO • H) t

Number of Chickens at the End of Period t: Equation (2.1) is a behavioral
equation including the main factors on which egg producers are assumed to
base their decisions when purchasing chickens. The main assumption behind
this formulation is that chick placings depend on the profitability of egg pro-
duction. Accordingly, the number of chickens purchased in a given period
is hypothesized as a function of the producer price of eggs, the price of feed
and the existing state of production technology.

However, since the decisions of purchasing chickens at a given moment
always refer to the future egg production and producers, therefore, have to base
their decisions merely on their expectations of the future price development, 1

we assume that there is a lagged response to change in the prices affecting the
profitability of production due to this imperfect knowledge. This assumption
of a lagged production response can also be justified on the basis of the nature
of the production process underlying the decisions. Some adjustments can be
made in a short period, while others require considerably more time due to
capital restrictions of firms, amount of fixed production resources, etc. (See
Heady and Tweeten 1964, p. 68—79).

Hence, distributed lag formulations including several lags of both the pro-
ducer price and feed price are used for describing producers’ responses to price
changes in purchasing chickens. As equation (2.1) indicates, the length of the
lag structure for neither variable has been specified in this stage. Practically,
the only way to find the most appropriate length of lags and estimation method
for this equation is estimation experiments using different lengths of lags and

1 Due to its crucial importance, much attention has been paid to this question in the past
supply analyses of agricultural products. Besides the use of typical cobweb models, different
kinds of modifications for example, incorporating equations for producers’ price expectations
into the model have been developed. Among them, Nerlove’s partial adjustment model
(1958, p. 45 65) and Larson’s harmonic motion model are the most well known (1964, p.
375 386). Because of its special interest with regard to this study, it should also be mentiond
the model formulated by Miller and Masters for the price expectation of egg producers:
X* = B 2 + a (Xt X t_4) + X t_!, where Xj = the expected price of eggs and Xt = the
current price in period t respectively and B 2 and a are coefficients to be estimated (1973, p. 485).



449

estimation methods.1 On the basis of a priori information, however, we expect
that the shapes for coefficient vector bui (coefficients of different lags of the
producer price) as shown in Figure 2.1 would be the most probable.

The shape of the coefficient vector in Figure 2.1 a indicates that the response
to a price change is strongest in the period when the price change occurs decreas-
ing gradually with time. Instead of that, Figure 2.1 b reflects a relatively small
response in the first period and, accordingly, the effect of a certain price change
is at its highest some periods later. Without empirical testing it is impossible
to decide, which kinds of coefficients vectors are the most probable for the
producer price of eggs and the feed price in equation (2.1). Therefore, we will
explore different kinds of estimation methods and lengths of lags for selecting
the »best» alternative. In addition, in this context we will point out that res-
ponse to rising prices may be different from response to price decreases.2

In this formulation prices of production inputs other than feed (such as
capital, wages, etc.) are ignored. This is mainly due to 1) their relatively small
importance in the total production costs of eggs, and also 2) difficulties in com-
puting operational variables for these cost items. It will be pointed out that
feed costs alone account for about two-thirds of the total production costs of
eggs3 and these in relation to egg prices should exert a major influence on chick
placings.

Changes in production technology may have had substantial impacts on the
profitability of production through the sample period. As stated before, the
problem of how to take these changes into account is a difficult problem in the

1 See Almon (1965, p. 178 180). He states, for example, that one can start by computing
simple correlation coefficients between the dependent variable and different lags of the expla-
natory variables when seeking the most appropriate length of lags.

2 This phenomenon, which is called the irreversibility of the supply function (Gerra 1959 a,
p. 12 17), has been an interesting problem in supply analyses in the past. This kind of pro-
ducers’ behavior has been verified by some empirical analyses. See, for example. Tweeter
and Quance (1969, p. 351 352) and Hartman (1974, p. 259).

3 See Westermarck (1972, p. 264) and Torvela (1959, p. 42—43).

Figure 2.1. Alternative Shapes for Coefficient Vector bn; in Equation (2.1) Reflecting
Different Kinds of Lagged Response to Producer Price Change.



450

time series supply analysis and there is no general solution to this problem. 1

In this study, perhaps the most typical method introducing a time trend
variable (T) into the equation as a proxy for technological changes has been
applied. The chick placings have significant seasonal variation and this may
have been changing over the years. It is this type of seasonal variation that is
taken into account by dummy variables D 2 and D 2T in the basic model.2

Number of Culled Layers (Equation 2.2): On the basis of the biological
aspects of egg production, we assumed that the potential number of layers to
be culled in a certain period is mainly determined by the size of the oldest age
groups of the laying flock. 3 Unfortunately, no homogenous data exist on age
distribution of laying hens for the entire research period. However, chick pla-
cings lagged in an appropriate way may provide good estimates for the sizes
of those age groups that are supposed to form the culling potential in a given
period. Therefore, variables SPt_k _j (i = 0 . . . n —1) have been introduced
into the model.

The length of the lag structure (values for k and n) was not yet specified in
this stage, but the decision on the length of lags and also on the estimation
method will be made on the basis of estimation experiments using different
lengths of lags and estimation methods for reasons discussed above (See the
equation for chick placings). In practice, it seems plausible to assume that,
on the aggregate level, the culling operations by producers may be directed
to many age groups, because the objectives and production technology
both most probably affecting the pattern of removing layers from production
activity are likely to vary among the firms in the egg industry. Some pro-
ducers cull layers regularly after one year of laying, while others consider it
profitable to keep their layers in production for two or even more years (some
producers may also use more flexible rules when deciding culling operations).
This kind of consideration suggests relatively long lag structures for variable
SP t_k _; in the estimation experiments. Thus, a priori information suggests
that the relevant shape of coefficient vector of this variable (b 2li, k= 3,
i •0 . . . 3) would be something like presented in Figure 2.2.

It has been hypothesized, however, that egg producers do not adhere
strictly to the age criterion in removing layers from the laying flock, but they
tend to change their »normal» pattern of culling layers according to the price
situations. If egg prices are low or feed prices high in a certain period, they
tend to remove more layers than in periods of favorable prices. Also, available
supplies of domestic feed grain may be a determinant in culling layers. That
is why, the current producer price (PKTt) and feed price (PRt) as well as supp-

1 Different kinds of proxy variables such as accumulated investments, variables indicating
levels of education, trend variables, etc. has been included to the supply and production functions
to solve the problem of technological change in recent time series analyses. As to the applica-
tion of these methods in production function and supply analyses in Finland, see Ihamuotila
(1972 a, p. 56-58) and (1972 b, p. 17-18).

2 Elimination of seasonal variation is further discussed in Section 3.2.
3 For example, when building their model, Agarvala and Ball (1970, p. 537 555) impli-

citly assumed that the age of layers is the only criterion for removing them from the laying
flock.



451

lies of domestic feed grains (RVS t) has been introduced into the model. In this
context it is to point out that the variables mentioned above cannot exert very
strong influence on the number of culled layers. Prices have not fluctuated
widely enough in the past and therefore, for example, high price level may
not have outweighed the deteriorating productivity of old age groups. On the
other hand, the continuation of production in periods of low prices would in
many cases be a better alternative than removing very young age groups of
the existing flock from the economic standpoint of producers.

In order to take into account the impact of changes in production technology
and the structure of the egg industry, a trend variable (T) was originally in-
cluded into the model. However, after the preliminary estimation experiments
it was deleted from the equation due to its statistically insignificant coefficients.
A seasonal pattern of culling exists and, that is why, seasonal dummy variable
D 2 has been introduced into this model.

Number of Layers: Equation (2.3) is an identity. The number of layers at
the end of a given period can be derived from the number of layers at the end
of the previous period (SK t _t) by adding the number of chickens at the end
of the previous period (time unit = six months) to that figure and subtracting
the number of layers culled in the current period (KK t) from that figure.

Output per Layer: Although originally planned, no equation for output per
layer was included into the basic model. This is mostly due to the fact that
preliminary analyses of the data of this study (See Nevala 1974,p. 11) could not
bear out hypotheses of producers’ attempts to change output per layer according
to the price situation (for example, by changing feeding rates). On the other
hand, all the models developed for output per layer using existing data could
not explain variations in the output per layer in the sample period any better
than a simple trend model. Obviously, there are so many factors influencing
the rate of lay which can be grouped into trend in genetics and management
practices that we could not estimate the separate effects of these factors on
the output per layer. On this basis, the equation for output per layer was
removed from the original model and total egg production equation was reformu-
lated according to this change: a trend variable as a proxy for improvements

Figure 2.2. The Expected
Shape of Coefficient Vector
b.ji Reflecting the Age Dis-
tribution of Culled Layers in
a Given Period.



452

in genetics and management was included into this equation to take into account
the influence of increase in output per layer on the total production.

Total Egg Production: Equation (2.4) expresses technical relationships
between total egg production and the laying flock in a given period. The total
egg production is hypothesized as a function of the average number of layers
and the output per layer in that period.

The average number of layers is computed according to the following formula:

SK t = SK t _j + - 0.5KKt

where SK t_j and SPt-1 are the number of layers and chickens at the end of
the previous period respectively, and KK t is the number of culled layers in
the current period. As to estimation, this variable is interesting. The main
question in this respect is, whether 1) to use the original specification as such
and estimate coefficient b4l for variable + 0.5SP t_x —0.5 KK t )

or 2) to estimate separate coefficients for each of the variables included to this
combination.

In practice, when using the first procedure, the coefficients are restricted to
be the same for each variable included. However, there are certain reasons
for assuming that the coefficients most probably differ from each other. For
example, the output per layer in the youngest age group (indicated by SPt-1)

is likely to be different from the average rate of lay and, on the other hand,
the total output of this age group in period t also depends on the average number
of days after reaching laying maturity. In the first procedure, we assume that
purchases of chickens are equally distributed over a given period and, thereby,
the chickens purchased in period t—l is assumed to lay eggs, on the average,
for half of the period t and their rate to lay is the same as the average rate for
the whole laying flock (therefore variable 0.5SP t_x ). Similar arguments can
also be raised against restricting the coefficient of variable 0.5K (number
of culled layers) to be the same as that of variable SK t_1 .

1 The major advan-
tage of the first alternative is the fact that, when using this approach, there
is only one coefficient to be estimated instead of three coefficients in the second
approach. Because of this kind of consideration both estimation procedures
were used to find out the most appropriate method.

Producer Price Equation (2.5): As stated before, the target price fixed by
the government (TAV) is the main determinant of the producer price of eggs.
According to the Price Acts, however, the producer prices have been permitted
to vary within a certain range around the target price level (generally within
i 5 percent range). Every Price Act has included specific provisions on the
conditions, which have to be fulfilled, before the government can intervene in
the egg market (paying export subsidies or allowing imports).

Without describing these provisions in more detail and to make things
operational we simply assumed here that the discrepancy between the do-
mestic egg production and consumption also exert an influence on the producer

1 Usually, layers to be culled in a certain period have passed the most productive stage,
and therefore, the output per layer in this group is likely to be lower than the average rate of
lay.



453

price in a given period.1 For example, if egg production increases more than
domestic consumption in a certain period, the producer price falls under the
target price level. The impact of this discrepancy between production and
consumption will last until the government is allowed (according to the Act)
to take actions to restore the current producer price to the target price level.
F'or these reasons, variable UK, [= TS, (CO-H),; net exports, and when
negative; net imports] indicating the magnitude of the discrepancy between
domestic production and consumption was introduced into the model. As can
be seen from the model, a negative sign is, a priori, expected for this variable.

Retail Price Equation (2.6) ; 2 Above we specified an equation for the de-
termination of the producer price of eggs and operationally the retail price
of eggs in a given period can be computed by adding a marketing margin to the
producer price in that period. Hence, the problem in building econometric models
for the retail price of eggs is to formulate operational hypotheses on the deter-
mination of the marketing margins for eggs. 3

An earlier study on marketing margins for eggs in Finland, made by Waa-
xanen and Kaarlehto (1965), provided an excellent starting point for hypothe-
sis formulation. They concluded, for example, that 1) the marketing margin
was primarily a linear function of the price level in absolute terms. 2) However,
the market situation influences the magnitude of the marketing margins; when
prices were moving up the marketing margin was smaller than when prices
were falling (p. 12). On the basis of these findings besides the variables
included in the final model a dummy variable (=1 for rising prices and
= 0 for falling prices) was also initially incorporated into the model to take
into account the above suggested behavior of the marketing sector in fixing
retail prices. However, the coefficient of this dummy variable in the prelimi-
nary estimation experiments was not statistically significant in any model
formulation. In addition, fixing values for this type of variable when using
the model for forecasting or policy analysis might have given problems and;
therefore, it was deleted from the final model.

Thus, the remaining explanatory variables in the retail price equations are:
the producer price, the wage level in commerce (to take into account the in-
creases in the marketing costs) and the seasonal dummy. This is a rather simple
model for retail price formation. Many factors which might have been relevant

1 In this context it will be pointed out that numerous hypotheses were tested and different
models were used in order to find an appropriate model for the producer price of eggs. Initially,
some efforts to incorporate certain regulations of the Price Acts explicitly into the model were
made. In addition, a somewhat different model formulation: PKTt = TAV t -f- Ap t , Ap t =

f (UK t , . . .), where Apt = the difference between the producer price and target price for eggs
in period t, was tested. However this model did not perform any better in the validity testing
than the model described above.

2 The wholesale price is ignored in the model building for the price determination of eggs,
because there exists no homogenous time series on the wholesale prices of eggs for the entire
research period.

3 Relatively much attention has recently been paid to the marketing margins of agricultural
producers in Finland. Kettunen (1968) and Pölkki (1971) have built different kinds of eco-
nometric models for exploring variations in the marketing margins for pork and beef. Also,
two special committees set by the government have investigated marketing margins; for eggs,
see: Anon. 1966, 1971 a.



454

in the determination of the marketing margins have been ignored mairly be-
cause of lack of data or difficulties in making any operational variable for these
factors.

Consumption of Eggs: Equation (2.7) refers to the total per capita consump-
tion of eggs in a given period. Building separate models for different compo-
nents of the total egg consumption, whose determinants may be different, was
not possible in this study, because no data on these components were available. 1

It is, however, reasonable to argue that this is not a very serious shortcoming
in the structure of the model with respect to the purposes of this model building:
an appropriate model for the total egg consumption is likely to be sufficient
for policy analyses in this study.

Equation (2.7) is a typical demand function used in many empirical studies.
The consumption of eggs is assumed to depend on the retail price of eggs (PKVt),

disposable income of consumers (Y t ) and the prices of all other consumer goods
(El). The effect of the prices of competing and complementary products was
assumed to be insignificant, this assumption being based on the statistical
findings of preliminary analyses of this study and these of earlier studies (Ko-
ponen 1964, p. 21). As it can be seen from the model, the price index of con-
sumer goods has been included into the model as a deflator of price and income
variables. This was done after using these three variables as separate variables
in the model which led to unsatisfactory results due to high intercorrelations
between these explanatory variables. However, at the same time a very restric-
tive assumption of consumers’ behavior was made: the reaction to changes
in the price ratio (PKV/El) t is the same irrespective of whether the change
is due to a change in egg prices or in consumer price index (the same holds also
to variable (Y/EI) t).

Net Exports: Equation (2.8) is an identity: equations for the determination
of both the domestic production and consumption have been specified earlier
and, thus, the net exports of eggs in a given period can be calculated by subtract-
ing total consumption from total production in that period. Net storage mov-
ements were not allowed in this model mainly because of lack of data and,
on the other hand, since it was assumed that this item comprises only a minor
portion of the total volume of eggs in a given period (semiannual data).

2. 4. Characteristics of the Basic Model

The eight-equation model including eight endogenous variables was present-
ed above without considering mutual relationships between the endogenous
variables. Since current endogenous variables appear also on the right-hand
side of the equations, there might be jointly dependent variables2 in the model
which will require the use of simultaneous estimation methods. Because of the

1 In many empirical demand studies, separate models have been specified for the con-
sumption of shell eggs and for other consumption (such as the use for breaking, hatching).
See, for example, Bange and Bender (1969, p. 29) and Roy and Johnson (1973, p. 209|.

2 Theil (1971, p. 429 436) calls the current endogenous variables, whose values are de-
termined jointly and simultaneously by predetermined variables and disturbances in the way
prescribed by the equations of the system, jointly dependent variables.



455

complexity of those methods in large models, recursive ordering of the equations
of the basic model to identify the simultaneous equation system(s) of minimum
size in this model may be useful (i.e. checking which equations are to be solved
simultaneously and which equations or subsystems are to be solved before a
particular subsystem can be solved). 1

An arrow scheme is a convenient way to illustrate relationships between
endogenous variables for purposes of recursive ordering of the equations (Fi-
gure 2.3).2 On the basis of arrows indicating the directions of the influence of

1 See Holt et.al. (1967, Chapter 5).
2 Also, special computer programs have been developed for this purpose. For example,

program SIMULATE II also used in this study analyzes the recursive ordering of the
equations in a given econometric model. This program is especially useful in the case of large
econometric models {See Holt et.al. 1967, Chapter 5).

Figure 2.3. Relationships Between the Endogenous Variables of the
Basic Model Indicating the Dynamics of the Model. (PKT = producer
price of eggs, SP = number of chickens, KK = number of culled layers,
SK = number of layers, TS = egg production, PKV = retail price of
eggs, CO = per capita consumption of eggs, UK = net exports).



456

each variable, we can conclude that the following endogenous variables are
jointly dependent in the basic model: PKT t , KK t , TSt, PKV t , CO t and UK t .

The model assumes that the current producer price influences the number of
layers to be culled in period t and the latter, in its turn, is the determinant of
the total egg production in the same period. On the other hand, the producer
price is, according to the model, the main determinant for the retail price, which

in its turn influences the egg consumption in period t. Thirdly, the net
exports (UK t), which are determined by the production and consumption of
eggs, are assumed to influence the producer price in this model. Thus, these
equations form a simultaneous block of six equations in the basic model.

As to equation (2.1), two endogenous variables (SPt and PKT t) are included
into this equation: producer price of eggs is one of the determinants of chicks
placings, but on the other hand the number of chickens purchased in period t
does not exert any influence on the other endogenous variables in the model
(only lagged SP’s have been included in the other equations). Accordingly
given the predetermined values, the number of chickens is determined by PKT t

(producer price) and ut (disturbance). Assuming that the disturbances in the
model are stochastically independent, we can expect that the values of PKT
are predetermined with respect to SP t . Therefore, equation (2.1) including only
predetermined variables on the right-hand side can be estimated separately.
Similarly equation (2.3), which is an identity, can be solved for SK t after the
value of variable KK t is determined from the simultaneous block mentioned
above.

Hence, this eight-equation model can be considered as a block-recursive
equation system containing three blocks, 1 which link each other as follows:

BLOCK 1 BLOCK 2

1 equation (2.I)SP tequation (2.2)KK t
» (2.4)TS t
» (2.S)PKT t
* (2.6)PKVt
* (2.7)C0 t
* (2.B)UK t

BLOCK 3

KK t equation (2.3)SKt

The model is dynamic; the typical feature of this model is numerous lagged
endogenous variables. On the other hand, an attempt has been made to keep
the number of exogenous variables to a minimum to facilitate forecasting and
simulation experiments with the model. The other characteristics which are
to be considered before estimation of the model are the linearity of the relation-
ships between variables and theidentificationof the simultaneousequation system.

Both the one-equation subsystems of the basic model (equations 2.1 and
2.3) are linear in variables and parameters (in absolute terms). On the other

1 We have a block-recursive equation system, when the coefficient matrix of the endo-
genous variables is block-tritangular (with square diagonal blocks) and when the variance-
covariance matrix of the disturbances is block-diagonal with conforming blocks (See, for example,
Theil 1971, p. 460—461).



457

hand, the simultaneous subsystem includes some nonlinear elements. Since
the equation of egg production has been specified for the total output and the
consumption equation on per capita basis respectively, the identity for net
exports of eggs is to be nonlinear in variables (including the nonlinear com-
bination of one endogenous and one exogenous variable (CO • H) t). Similarly,
one nonlinear combination of an endogenous and an exogenous variable
(PKV/EI), was included in the consumption equation instead of using these
as separate variables in order to reduce intercorrelation of the variables on the
right-hand side of this equation. Moreover, equation (2.6) includes a nonlinear
combination of variables UK t and Ht .

This is a typical situation in practical model building (See, for example,
Naylor 1971, p. 139), and problems may arise from these kinds of nonlinear-
ities in the estimation and simulation experiments with the model. However,
since all the nonlinear elements in the model are combinations of one endo-
genous and one exogenous variable, relatively easy solutions can be found
to these problems (such as polynomial approximations of these variables; see
Naylor 1971, p. 181). These will further be discussed in the stages of estimation
and model evaluation through simulation experiments.

Generally speaking, the identification of an econometric model can be
considered as a problem of deriving the coefficients of the structural form from
the reduced form coefficients of the system. 1 In the case of linear models simple
rules have been developed for checking the identification of a certain equation
system. However, as stated above, the simultaneous subsystem of this study
is nonlinear in variables and these rules do not apply to this model. The iden-
tification theory for nonlinear models is difficult and has only been partially
developed (Johnston 1972, p. 372). Therefore, we will ignore the identifi-
cation problem of the present model. Before estimation of the simultaneous
submodel, we need only check whether this model satisfies the specific con-
ditions necessary for using a typical simultaneous equation estimation method
in the case of this model (See Section 3.3.1).

2. 5. Recursive Production Response Model

Analyzing different specifications and the stability of the structural coef-
ficients of the model over periods of time is an important part of model evalua-
tion. As stated by many economists, the coefficients are not likely to remain
constant over long periods of time. As to this study, especially producers’
responses to changes in socio-economic forces may have changed over the years
of the sample period due to the technological changes in production. Therefore,
to get some kind of idea as to the trends of these changes over periods of time,
the model should be estimated from different time periods.2

1 For more detail, see Johnston (1972, p. 341—356).
2 Teräsvirta (1971, p. 4) states that in this way we obtain many estimates of the structu-

ral coefficients, and by examining the coefficients estimated from different time periods we
can obtain information about the changes in the process which have generated the values of
the endogenous variables.



458

However, the eight-equation model presented above is rather inconvenient
for this kind of testing because of the complexity of estimating the coefficients
of the simultaneous subsystem. That is why to facilitate the computational
procedure a simple model for the egg production was derived from the basic
model. If we ignore our initial assumption on the relationship between the
producer price of eggs and the current market situation (i.e. variable UK,
is deleted from equation (2.5)), the producer price becomes an exogenous va-
riable in the model. Therefore equation (2.2), explaining the number of culled
layers, includes only one endogenous variable. In this case we can write the
equation system for the egg production determination in matrix notation as
follows: _

1 O O O-1 - SP t 1 r-1 iigp
0 10 0 KK, + C Z = u KK
0 b 1 0 TS, u-pg
0 10 1- SK, L 0

Where B and C are the coefficient matrices for the endogenous and pre-
determined variables respectively, and Z is the vector of the predetermined
variables. As can be seen from the matrix presentation of the model, which
is henceforth called the recursive production response model, the coefficient
matrix of the endogenous variables is subdiagonal (i.e. it has only zero elements
above the diagonal) and, therefore, this set of equations can be regarded as a
recursive system. 1 The two first equations, determining the values of SP t and
KK, respectively, include only one endogenous variable each. The third equation
implies that, given the predetermined values, TS t is determined by KK t and
uTS . If we assume that the disturbances are stochastically independent, we
can expect that KK t is predetermined with respect to TSt . The fourth equation
is an identity including two endogenous variables (KK, and SK t). However,
the value of KK t is determined by the second equation and, thus, the fourth
equation explains the values of SKt in terms of predetermined variables only.

Accordingly, assuming that the variance-covariance matrix of disturbances
is diagonal this equation system can be considered as four one-equation sys-
tems, and the ordinary least squares method can be used for estimation of the
parameters.

3. Data and Estimation Methods
3. 1. Estimation Period

The estimation of the basic model was based on semiannual time series
from the year 1952 to 1970. Thus, taking into account the lagged variables
incorporated into equations (2.1) and (2.2), this period provided some thirty

1 According to Wold (1964, p. 17), the recursive system consists of a set of equations which
form a causal chain so that the first equation contains only one endogenous variable YL and
the second explains Y.2 in terms of Y, and other predetermined variables and so on. The key
point is that each successive equation may include only those endogenous variables that have
appeared in the previous equations.



459

observations for regressions depending on the lengths of lags applied to variables
PKTt_j, PRt_i, and SPt _k_; in each estimation experiment. The reason
why the latest observation periods (from the year 1970 on) were not included
was to provide a comparatively long time period for the evaluation of the
model via ex-post forecasting and, also, for comparing the outcomes of ex-
post simulations of different price policies generated by the model to those
under the actual price policy. Accordingly, time period 1971—74 was used
for testing the model by means of predictive tests.

However, besides the estimation from the base period, stepwise regressions
with regard to observations 1 were applied to the recursive production response
model (See Section 2.3) in order to explore the stability of the coefficients over
periods of time.

Sources of the data used in this study as well as the main principles for
computing observational values for the variables of the model are described
in Appendix 1. In order to obtain homogeneous time series for the entire research
period, some adjustments to the official figures published for certain periods
had to be made due to changes in the system for compiling these statistices
over time. Generally speaking, data availabilities for this study were not, in
part, quite satisfactory with regard to the coverage of certain phenomena in
the egg industry nor, on the other hand, with regard to the accuracy. Due
to the lack of data, some probably relevant factors had to be excluded
from the model. This is, however, a rather typical situation in the practical
model building for policy purposes: one can hardly find a empirical econometric
study, in which no complaints about the data shortages have been presented.

3. 3. Elimination of the Seasonal Variation

As the estimation of the model will be based on the semiannual data, an
additional problem arises from the seasonal patterns of those phenomena of the
egg industry that the model describes. Some of the time series, such as the
number of chickens (SP t ) and the number of layers (SK t ) seem to have very
clear seasonal components. As can be seen from the equations of the basic
model, seasonal dummies (D 2) were included into the model to take into account
this kind of seasonality in the observed values. Another approach to the sea-
sonality problem that was tested before selecting the seasonal dummy method,
was eliminating seasonal variations from the time series and using these sea-
sonally adjusted series as a statistical basis for the model.2

1 The stepwise regressions with regard to observations is in this study defined as several
successive estimations of the parameters of the model such that the estimation period is syste-
matically changed by dropping one or more of the oldest periods from the beginning and adding
a similar number of new periods to the end of the estimation period (See Teräsvirta 1971).

2 There are many alternative methods for computing seasonally adjusted series (See Kuk-
konen 1968 and Saito 1972). In this study the iterative method developed by Kukkonen
was used to eliminate seasonal variation from the semi-annual time series. The applied model
was multiplicative: the original time series (Ot ) were divided into three components; log O t =

log P t + log S t + log I t , were P t is trend-cycle, S t seasonal and I t residual component res-



460

However, since the estimation results of these two methods did not in
essence differ from each other, the seasonal dummy approach was chosen
mainly for two reasons. The main advantage of this approach over the use of
seasonally adjusted series is that the model directly generates the final values
of the endogenous variables in the seasonal dummy approach. On the other
hand, when seasonally adjusted data is used the model generates seasonally
adjusted forecasts and, therefore, an additional block of equations has to be
included into the model for forecasting values of the seasonal components of
the time series in which we are interested. In addition, Klein (1965, p. 35)
for example suggests the use of explicit dummy variables, because in this way
one avoids using seasonally adjusted time series that have already had syste-
matic computations built into them. On the other hand, the main advantage
of the separate elimination of seasonal variation is that, unlike in using the
seasonal dummy approach, we can take into account possible changes in the
seasonal components of time series by applying some specific methods deve-
loped for this purpose (such as that by Kukkonen 1968. See also Saito 1972).

This dummy variable (D 2 = 1 in the second half of the year, otherwise
D 2 = 0) is a kind of shift variable. Its coefficient indicates how much the level
of the function has to be adjusted in the second half’s seasonal influence. In
addition, to take into account changes in the seasonal pattern of purchasing
chickens over periods of time, variable D 2T (the product of the seasonal dummy
variable and the trend variable) was incorporated into equation (2.1).

3. 3. Estimation Methods

3. 3. 1. General
The following step in the model building selecting the estimation methods
refers to the problems of obtaining parameter estimates, which would have

desirable statistical properties (unbiasedness, consistency, efficiency, etc.).
In this respect the ordinary least squares method (OLS) can be applied to the
one-equation subsystem (Block 2) in the block-recursive basic model of this
study as well as to the estimation of the four-equation recursive production res-
ponse model described in Section 2.5 under certain conditions.

As for the recursive production response model, the application of the OLS-
method to the structural equations leads to consistent and asymptotically
efficient estimates, if this model is diagonally recursive: that is, if besides
the tritangularity of the coefficient matrix of the endogenous variables also
the variance-covariance matrix of the structural disturbances is diagonal

pectively. Seasonally adjusted series were computed according to the following formula: OS t
=

Ot /S t. Also, the regression method has been used to estimate seasonal components (as for eggs;
see, for example, Kersten 1975, p. 89—97). In this context it will be shown out that Lovell
has proved that, when the time series adjusted by means of the least squares regression on
seasonal dummy variables are used, the same estimates of the coefficients of the model are
obtained as when using the seasonal dummy variables in the original equation (Johnston
1972, p. 189-190).



461

(See Kmenta 1971 p. 586). 1 Similarly, the variance-covariance matrix of the
disturbances of the basic model has to be block-diagonal to obtain OLS-esti-
mates with desirable statistical properties for equation (2.1) of the basic model.
A priori, there were no reasons why these basic assumptions on the distur-
bances of these models could not be made in the case of these equations.

Instead of that, the OLS-method applied to simultaneous equation systems
generally leads to inconsistent estimates, because one of the basic conditions
is violated in this case: the endogenous variables on the right-hand side of the
equations are, in general, correlated with the disturbance of the equation in
which they appear (Kmenta 1971, p. 533). Therefore, the two-stage least
squares estimation method (2 SLS, see Theil 1971, p. 451—460) was applied
to estimation of the six-equation simultaneous subsystem of the basic model
(Block 1). Being a single-equation method it provided a relatively flexible and
easy method for estimating different model specifications when searching for
the »best» model structure. In the case of linear models the 2 SLS-method
can be described as a two-step procedure, in which 1) the first stage consists
of obtaining proxies of the right-hand side endogenous variables (Y jt) from the
i-th reduced form equation as estimated by the OLS-method and 2) the second
stage uses these proxies (^ jt) in the place of the right-hand endogenous variab-
les in regressions using structural equations.

However, the simultaneous subsystem of the basic model is nonlinear in
the variables (including nonlinear combinations of endogenous and exogenous
variables such as (PKV/EI),, (UK • H) t). Problems arise from these nonlinear-
ities in the estimation, since in nonlinear models the reduced form is, in general,
nonlinear in the parameters (See Edgerton 1972, p. 27). To avoid difficulties
of this kind in applying the 2 SLS-method to models nonlinear in the variables,
the econometric theory proposes using polynomial approximations to the re-
duced form in the first stage of the procedure (See Kelejian 1971, p. 373
374 and Edgerton 1972, p. 26—32). Edgerton explored two such approx-
imations: 1) a method which used a linear function of thepredetermined variables
in the first stage and 2) a method which used a second order polynomial in the
first stage. The consistency of these methods is established in his paper by
showing that they are equivalent to the instrumental variables method and,
also, conditions for their application are given.

The first variant of the 2 SLS-method2 is used for the six-equation simulta-
neous subsystem. According to Edgerton (p. 30), the condition that must
be met for this method to be applied to the i-th equation of a nonlinear model
is that the total number of variables and functions on the right-hand side of
the i-th equation is less or equal to the number of predetermined variables in

1 Of course, also the other assumptions on the nature of the disturbances necessary for
applying the OLS-method have to hold in these structural equations: 1) the disturbances are
normally distributed with zero mean and constant variance, 2) not serially correlated and 3)
independent of the explanatory variables (See Johnston 1972, p. 12, and 122 123).

2 In fact, this variant is the 2 SLS-method for the nonlinear situation, if we define 2 SLS
for the linear case as a two-step regression method {See Edgekton 1972, p. 27).



462

the system. 1 Fulfilling this condition in the case of the present simultaneous
submodel does not seem to be any problem; even though we do not take into
account the lags of SP t_k_i in equation (2.2), there are ten predetermined
variables in the system. Hence, for example, in the case of equation (2.2) which
is the most critical equation in this respect, the above condition will be easily
fulfilled when using alternative estimation methods for lagged variables SPt _k _j.

3. 3. 2. Polynomial Lag Formulations
The number of chickens at the end of a given period (equation 2. 1) was

assumed to be determined by several past values of predetermined variables
PKT (the producer price of eggs) and PR (the feed price). Similarly, equation
(2.2) is a distributed lag model consisting of several lags of variable SP. The
straightforward estimation of these equations by the use of ordinary estimation
methods (such as OLS, etc.) is possible, but on the other hand, intercorrelations
between explanatory variables may be a problem in this case: it may disturb
the »natural» pattern of the coefficient vectors of these variables.

An alternative estimation procedure, suggested first by Almon (1965, p.
178—196) to the above is a finite distributed lag formulation in which the
coefficients of different lags are restricted to lie on a polynomial of low order.
By making this restriction the number of parameters to be estimated can be
reduced, because a polynomial of low order can be defined by a few parameters.
The individual lag coefficients can be then computed from these estimated
parameters.

Initially both methods were used in this study. The latter approach was,
however, chosen for the final model; this decision being based on the estimation
experiments by using both the alternatives. Because there are computationally
different procedures for estimating finite polynomial lag models2

, the method
used in this study is described in detail in Appendix 3. Only polynomials of
first and second order were used in this study. The main idea of the procedure
is to include transformations from the original explanatory variable into the
equation instead of the separate lags of that variable as shown in Appendix
3. The OLS-method (or 2 SLS-method in the case of simultaneous equation
systems) can be applied to the estimation of these equations. After estimation,
the individual lag coefficients can be computed from the estimated coefficients
of the transformed variables according to the formula given in Appendix 3.
If the disturbances (u t) satisfy all the assumptions of classical regression models,
theresulting estimates will have all the desirableproperties (Kmenta 1971,p.493).

This method provides a flexible procedure for estimating distributed lag
models: the length of the lag and the degree of a polynomial can be easily altered
to find the »best» lag structure. On the other hand, one of the disadvantages

1 Functions are, in this case, defined as g(Yt,Z t), where Yt and Z t is an endogenous and
exogenous variable respectively and that it contains no unknown parameters. If this condition
does not hold then a singular matrix will be required to be inverted in the second stage.

2 For example, estimation by Lagrangian interpolation polynomials (Almon 1965) or by
the use of simple polynomials (Chen et. al. 1972, p. 78 79).



463

of this method is that alternative choices to be tested increase rapidly: we have
to specify a priori both the degree of a polynomial and the length of the lag
for each variable.

4. Estimation of the Models
As stated, the recursive production response model was developed in order

to evaluate different specifications of the production response equations and to
explore the stability of their coefficients over periods of time. Accordingly,
the final decisions on the structure of the basic model of eight equations were,
in part, based on the estimation results of the above mentioned model. Thus,
to be consistent with this order in this section, the estimation of the recursive
production response model is first presented; the estimates of the parameters
of the model selected after many estimation experiments with different
specifications are first discussed with some comments on alternative spe-
cification possibilities (Section 4.1.1). Section 4.1.2 deals with variations in
the parameter estimates of the same model over periods of time. Finally, the
estimates of the basic model are given in Section 4.2.

In this context, it should be also mentioned that due to a great number
of alternative choices (such as different variable combinations, distributed lag
formulations, estimation methods and periods, etc.), it is almost impossible
to test all the choice alternatives that a priori seem plausible. Therefore, in
interpreting the estimation results one has to bear in mind that if some choices
during the model building of this study had been done differently, perhaps,
we would have ended up with different model structures for the Finnish egg
industry.

4. 1. Recursive Production Response Model

4. 7. 7. Estimatesfrom the Period 1956—74
The different specifications of the recursive production response model were

tested on the entire research period. Accordingly, the OLS-estimates for the
selected model from period 1956/1—74/1 are as follows1:

Number oj Chickens:

(4.1) SP t = 0.432 + 0.551PKT t + 0.367PKTt_1 + 0.184PKT t_ 2 -2.662PR t
2.2** 2.2** 2.2** —2.4**

- 1.775PR t !
- 0.887PR t_ 2 + 0.063T - 0.036D 2T - 1.366D2 + u SP

-2.4** -2.4** 3.6*** -s.o*** —s.l***

R 2 = 0.98 su = 0.228 d = 1.67

1 Numbers below regression coefficients are Student t-values. The significance levels are
indicated as follows:

* statistically significant at the 90 percent level
** statistically significant at the 95 percent level

*** statistically significant at the 99 percent level.

R 2 = the coefficient of determination, s„ = the standard deviation of the residual and d = the
Durbin-Watson test statistic.



464

Number of Culled Layers:
(4.2) KKt = 0.748 -0.608PKT t + 3.046PR t -0.148SPt_2 + 0.137SPt_3

I.7** 2.l** -1.1 I.B**

-f* 0.256SP t -f- 0.210SP t g -f- 0.259D2 “f* Urk
2.B*** 2.B*** 0.8

R 2 = 0.81 su = 0.258 d = 2.03

Egg Production :

(4.3) TSt = -3.594 + 4.795SK t_1 + 1.756SPt_l -0.943K + 0.167T
12.5*** 3.4*** —l.s* 4.6***

+ 0.814D 2 + u-j-s
0.7

R 2 = 0.99 su = 0.825 d = 1.40

Number of Layers (Identity):
(4.4) SK t = SKt_, + SP t !

- KK t

The coefficients of PKT t-i (producer price for eggs, i = 0 . . .2) and
PR t _i (feed price, i=o.. . 2) in equation (4.1) were estimated by using
a polynomial of first order, because this formulation proved to yield the most
plausible results in the experiments using different lengths of lags and degrees
of the polynomial. Hence, there were only six coefficients to be estimated by
OLS in this equation. The estimated coefficients 1 were, however, unscrambled
(See Section 3.3.2) in order to present the model in the original form. Similarly,
the coefficients of SPt_2 .. . SPt _5 in equation (4.2) were estimated by using
a polynomial of second order.2

Assessing on the basis of the Revalues and the standard deviations of the
residuals, the equation system seems to explain producers’ reactions rather
well. In order to test the assumption on the non-correlation of the disturbances
of this recursive model, a linear correlation matrix of the computed residuals
was calculated:

U SP UKK U TS
1.0U SP

"KK
UT S

0.30 1.0
0.29 0.01 1.0

As can be seen from the figures, the highest correlation coefficient (r) was
obtained for residuals uSP and uKK (equations 4.1 and 4.2). However, it is not
statistically different from zero even at the 5 percent level of significance 3

1 The estimated coefficient for (PKT) Dgl was 1.102 (t-value = 2.2) and for (PR)DgI 5.324
(t-value = —2.4) respectively.

2 The coefficient for (SP)DgI was 2.929 and for (SP)Dg2 2.474 with t-values 2.8 and
2.6 respectively.

3 The number of observations (n) is in this case 37. For n = 30, r has to be S: 0.36 and Ss

0.31 for n = 40 at the five percent level of significance to be statistically different from zero
(See D:xon and Massey 1957, p. 468).



465

and, accordingly, we can conclude that the residuals belonging to this recursive
model are not at least linearly correlated with each other. 1

All the coefficients in equation (4.1) are statistically significant at least at
the 95 percent level; thus, bearing out our basic hypotheses on the criteria of
producers in deciding the quantities of chickens to be purchased in agiven period.
These results also indicate comparatively sensitive and rapid producers' reactions
to changes in the prices affecting the profitability of egg production; at the mean
level of the variables, the elasticity of chick placings with regard to the pro-
ducer price of eggs (PKT t_i( i= 0 ... 2) is 1.89 and with regard to the feed
price (PR t_i, i=o. . . 2) —1.73 respectively.2 Thus, in absolute values they
are almost the same.

The elasticities above are cumulated price elasticities referring to the entire
response period (3 periods in this case). As to the distribution of the response
between periods, the results indicated that the producers’ response to a price
change is at its highest in the same period during which the price change occurs
declining gradually with time. In testing different lengths of lags and degrees
of polynomials, a polynomial of first order proved to give the most plausible
results for every length of lag tested in these experiments (2—6 periods). The tests
also proved that the extension of the lag structure over three periods did not
essentially change the estimated total response to price changes as indicated
by the lag coefficients below, which were estimated using 3 and 4 periods as
a length of lag for the price variables in equation (4.1):

Lag Coefficient
Period Variable t t —1 t —2 t —3 £

3 PKT 0.551 0.367 0.184 - 1.102
4 PKT 0.496 0,372 0.248 0.124 1.240
3 PR -2.662 -1.775 -0.887 - -5.324
4 PR -2.247 -1.685 -1.123 -0.562 -5.617

In this context, it must be pointed out that a number of efforts were made
to incorporate the producer price index of agricultural products into equation
(4.1) to represent the prices of alternative agricultural products, but in every
case the estimation results were not consistent with a priori expectations
partly due to the high correlation between the producer price of eggs (PKTDgI)
and the producer price index of agricultural products. In addition, it was
attempted to incorporate a variable indicating the feed grain availabilities
(RSV t) into the model. However, its coefficients were not statistically signifcant.

In the second equation (explaining the number of culled layers in a given
period), the R 2-value remained relatively low compared to those of the other

1 However, as Kettunen (1968, p. 61 62) points out, we cannot conclude that the theo-
retical assumption of non-correlation of disturbances is valid since the test above concerns only
linear correlations.

2
2 These elasticities were computed as: e = £bj • X/Y, where b; = the regression

i=o
coefficients of different lags of the explanatory variable X, x = the mean of the explanatory
variable X and Y = the mean of the dependent variable computed from the entire estimation
period.



466

equations. This does not necessarily refer to the failure of the specification.1

The time series for KK t was derived from the other time series (SK t and SP t)

computationally and, thus, it may contain some errors. On the other hand,
it seems reasonable to argue that the number of layers culled in a given period
is subject to unpredictable variations due to the total effect of several minor
factors such as health conditions, feed grain situations, etc.

In order to choose an appropriate lag structure for SPt_k_j (lagged num-
ber of chickens), the equation (4.2) was estimated many times using different
lengths of lags (from t 2 to t —7) and degrees of a polynomial. Lags SPt _2
.. . SPt_5 and a polynomial of second order proved to be the most appropriate
for this variable, although logic suggests that SPt_2 not be included in the
equation on the basis of the age criterion.2 However, the estimation of the equa-
tion without SP t_ 2 (i.e. using lags from t—3to t —5) did not yield statisti-
cally significant coefficients for transformed variables (SP)DgI and (SP)Dg2.

The signs of the coefficients of the current producer price of eggs (PKTt)

and the feed price (PRt) are also logical, but on the other hand, the coefficient
of PKT t is statistically significant only at the 90 percent level. High inter-
correlations between the explanatory variables obviously disturbed the estima-
tion of this equation leading to the relatively large standard errors of the
coefficients. Therefore, some modifications of this model (polynomial lag for-
mulations for PKT and PR) were also estimated. However, the results did not
give us any reasons to change the model specification. It should also be men-
tioned that variable RVS t (feed grain availability) was deleted from the equation
because of the statistical insignificance of its coefficients in every model for-
mulation tested in this stage.

According to these results, the elasticity of culling layers with respect to the
producer price of eggs is —1.2 at the mean levels of variables and with respect
to the feed price 1.1 respectively. Thus, in absolute values they are almost
the same; an increase of one percent in the producer price of eggs or a decrease
of the same size in the feed price reduces the number of culled layers by just
over one percent.

In estimating the egg production equation no restriction were here imposed
for the parameters of variables SK t_ 1( SP t-1 and KK t 3 to find out the se-
parate impacts of variations in these different components of the total laying
flock on the egg production. All those variables had logical signs and the mag-
nitudes of the coefficients reflect comparatively well a priori assumptions. How-
ever, the coefficient of KK t was not statistically significant in this structure.
One reason for this may have been the high correlation between SKt-1 and

1 In this context it shouldbe mentioned that the Durbin-Watson test value computed from
the residuals of this equation was 2.03; thus, indicating no autocorrelation of the residuals.

2 Incorporating SPt _2 into the equation is not consistent with the age criterion. However,
the negative sign of this lag is logical, indicating a negative relationship between the number
of layers removed in a given period and the number of chickens purchased two periods earlier.

3 When the coefficients are restricted so that the coefficient both for SP t-1 and for KK t
is assumed to be half of the coefficient of SK t_1; estimation yields the following equation:
TS t = 3.318 + 4.971 (SKt_l + 0.5SPt _! -0.5K + 0.176 T + 0.322D 2 + u TS

13.2*** s.2*** 1.3
R 2 = 0.99 su = 0.781 d = 1.38



467

KK t in the estimation period. In addition, we have to point out that the Durbin-
Watson test applied to this equation is inconclusive (d = 1.40, in this case
the lower and upper bounds of the 1 percent points of the D-W test statistic
are 1,00 and 1.59 respectively). Therefore, one needs further observations to
be able to conclude whether the residuals are positively autocorrelated or not.

The coefficient of the trend variable (T) in equation (4.3) is statistically
significant at the 99 percent level indicating the importance of the growth in
output per layer for the growth of the total production. Because the model
seemed to underestimate the total egg production in the years 1964—1968 and,
respectively, overestimate in the years 1969—1970, some additional variables
were included into the model in order to explain these variations (variables
indicating variations in the age structure of the laying flock, feed grain avail-
ability). Unfortunately, these additional variables could not reduce the residual
variations of the model in those years.

4. 1. 2. Vaviations in the Coefficients of the Recursive Production Response
Model in Time

The method for exploring variations in the production response over periods
of time was stepwise regression with respect to observations. In the beginning,
the model was estimated from the first 25 observations of the research period
and, after that, three successive estimations of the parameters were performed
such that in each stage the four oldest observations were dropped from the
regression and four new observations were included in the regression analysis.
Thus, four different model structures were obtained from the entire research
period. In addition, for price policy simulation purposes two different estimation
periods ending in the year 1970 were also used.

The results are summarized in Table 4.1 (the more detailed results arc given
in Appendix 4). As can be seen from the Table 4.1, the coefficient of determina-
tion of any equation does not vary widely over the years, the greatest difference
between Revalues is in the equation explaining the number of culled layers:
the difference between R 2 max and R 2 min is 0.052. The standard deviation of
the residuals seems to grow in absolute terms. However, since the values of the
dependent variables also grow over the years, the standard deviations of each
equation measured by percentages remain almost at the same level for every
estimation period.

Table 4.1. Variations in the Coefficients of the Recursive Production Response Model with
Respect to the Estimation Period. The Coefficients Estimated from Period 1956/1 70/11 = 100.

Equation (4.1): SP {= number of chickens) as the Dependent Variable 1

Coefficient of
Period

(FKT)DgI (PRiDgl T DJ P 2 R 2 su

56/1 -69/1 76 133 119 70 128 0.980 0.195
58/1 -70/1 108 111 96 91 112 0.971 0.261
60/1 -72/1 102 102 80 73 134 0.973 0.265
62/1 —74/1 77 93 98 77 126 0.975 0.266
56/1 -70/11 100 100 100 100 100 0.974 0.237
57/11 —7O/11 104 110 100 51 110 0.973 0.250
56/1 -74/1 97 119 112 82 117 0.977 0.288



468

Equation (4.2): KK t (= numbei of culled layers) as the Dependent Variable.

Coefficient of
Period

SP t 2 SPt_,
SP t_4 SPt_ 5 PKT t PR t D ä R 2 su

56/1 -68/1 75 136 123 120 127 97 114 0.858 0.201
58/1 -70/1 70 84 81 81 90 93 161 0,838 0.220
60/1 -72/1 -85 49 20 13 63 65 362 0.862 0.202
62/1 -74/1 47 58 56 56 37 62 236 0.813 0.274
56/1 -70/11 100 100 100 100 100 100 100 0.825 0.222
57/11 -70/11 126 92 99 101 96 93 79 0.804 0.234
56/1 -74/1 130 89 98 100 100 114 96 0.811 0.259

Equation (4.3): TS t (= egg production) as the Dependent Variable

Coefficient of
Period

SKt _! SPt_,
KK t T D R 2 su

56/1 -68/1 96 91 106 128 161 0.979 1.28
58/1 -70/1 124 168 197 51 -117 0.977 1.12
60/1 -72/1 104 92 118 76 259 0.982 1.21
62/1 -74/1 105 96 68 66 735 0.976 1.39
56/1 -70/11 100 100 100 100 100 0.978 0.98
57/11 — 7O/11 108 104 110 78 141 0.980 1.06
56/1 -74/1 98 82 56 99 144 0.985 1.39
1 The index numbers are computed for the coefficients of transformed variables (PKT)DgI

and (PR)DgI, because the coefficients of different lags are proportional to them in the case of
a polynomial of first order.

In order to illustrate better the variations in the regression coefficients, these
were transformed into index numbers (coefficients from period 1956/1—1970/11
= 100). Although there are rather wide fluctuations in the coefficients of some
variables, no clear systematical changes with respect to time can, however, be
perceived in the magnitudes of the coefficients. For example, the coefficients
of PR t_; and PKTt_i in the first equation seem to have some kind of down-
ward direction with respect to time; possibly referring to a weakening in the
sensitivity of producers’ reactions to changes in prices in purchasing chickens
over the observed periods. On the other hand, the standard errors of the coeffi-
cients estimated from the latest periods are relatively large. 1

In the second equation, in which the number of culled layers served as the
dependent variable, especially the coefficients of SP t_2 .. . SPt _5 estimatedfrom
observations 1960/1—1972/1 and 1962/1—1972/1 seem to differ significantly from
those obtained from the other estimation periods. In addition, the t-values
of the coefficients estimated from those periods remained relatively low (See
Appendix 4). No clear explanation for this phenomenon could be found (in-

1 One reason for the growth of standard errors of these coefficients is that intercorrelations
of the explanatory variables are higher in the regressions from later periods than from earlier
periods.



469

tercorrelations between explanatory variables were in those regressions higher
than in the rest, which is certainly one reason). However, because coefficients
estimated from the base period (1956/1—1970/11) and from the entire period
(1956/1 —1974/1) were almost the same, the specification of the equation (4.2)
was not changed.

The »fit» of the egg production equation (equation 4.3) is rather good for
every estimation period and the coefficients of the key variables (SK t-1 and
SPt-1) do not vary widely over time. On the other hand the coefficient of
KK t (the number of culled layers) is not stable: the most exceptional coeffi-
cients for KK t were obtained from periods 1958/1 —1970/1 and 1962/1—1974/1.

To summarize the results of the stepwise regression we may conclude that
the estimates of some parameters were not stable but varied depending upon
the estimation period. It is obvious that the sensitivity of producers’ reactions,
for example, to changes in prices have changed over periods of time due to
changes in the production technology and the other economic circumstances
of production. However, these results did not give us any clear indication of
systematical changes in the parameters of the model over the research period.
Reasons for this may be numerous; such as a relatively short research period
which limited the number of observations available for the anatysis. On the
other hand, the irregular variations of the coefficients of the model may be
due to multicollinearity problems, which disturbed the estimation of the coeffi-
cients from certain periods. In any case, the results obtained in this section
reflect the problems of econometric model building: the specification of the basic
model might have been different, if the model had been developed on the basis
of other time periods that used in the case of this model.

4. 2. Basic Model: Estimates from the Period 1956—70
Although many different specifications of the basic model were tested in

the estimation experiments and many estimation periods were used, only one
structure of the basic model is given below: the structure whose coefficients
were estimated from semiannual observations for period 1956—1970 (called
the base period). This is the structure of the model which will be used for policy
simulation experiments in this study.

One of the three blocks (Block 2) can be estimated by the OLS-method as
stated in Section 3.3. 1 In fact, OLS-estimates for this equation from period
1956—7O are given in Appendix 4, where the results of the stepwise regressions
of production response model are presented. However, the whole structure
of the basic model to be used for the price policy simulation is presented in this
section to avoid confusion.

In the estimation of the simultaneous subsystem (Block 1) by the 2 SLS-
method, a polynomial of second order was again used for the estimation of the
coefficients of SPt_2 . . . SPt_5 in equation (4.5). Accordingly, the trans-
formations of variable SP as shown in Appendix 3 [(SP)DgI and (SP)Dg2]
were included in the first stage of the 2 SLS among the other predetermined

1 Block 3 included only an identity and needed no estimation.



470

variables of the simultaneous equation system. However, in order to present
the model in the original form, the coefficients of (SP)DgI and (SP)Dg2 es-
timated in the second stage of the procedure were unscrambled according to
the formula presented in Appendix 3.

The estimated structure of the basic model is:

Block 1: Simultaneous Six-Equation Subsystem (2SLS-estimates) 1 :

Number of Culled Layers:
(4.5) KK t = 1.927 + 0.190SPt_ 6 + 0.229SPt_4 + 0.115SP t_.,

- 0.152SPt_ 2

2.1»* 2.l** 1.7* -1.1

0.217(PKT/PR) t + 0.238D2 + u 2
-I.9** 0 6

Egg Production:
(4.6) TS t = -2.236 + 4.595(5K t_! + 0.5SPt _j - 0.5KK t) + 0.192T

7.l*** s.2***
-f- 0.375D,,

1.5*

Producer Price of Egg:
(4,7) PKT t = -0.405 + 1.190TAVt - 0.237(UK/H) t + 0.069D 2 4- u 5

B.7*** -I.9** I.B**

Retail Price of Eggs:
(4.8) PKVt = 0.408 + 0.854PKT t + 0.639YM t 4- 0.050D 2 +uc

s.9*** 2.s*** 1,0

Consumption of Eggs:
(4.9) CO t - 5.032 - 1.261(PKV/El) t + 2.686(Y/El) t + 0.221D2 +u,

—2.l** 3.B*** 1.5*

Net Exports (Identity):
(4.10) UK t =TS t - (CO • H) t

Block 2: One-Equation Subsystem (OI.S-Estimates);

Number of Chickens:
(4.11) SP t = 0.164 + 0.568PKT t + 0.379PKT t_! -f 0.189PKT t_2 -2.223PR t

I.B** I.B** I.B** -I.B**

- 1.482PR t_1
- 0.741PR t_., + 0.056T - 0.044D 2T - 1.164Da +u 2

I.B** —l.B** 2.4** -4.2*** -3.4***
R 2 = 0.97 su = 0.237 d = 1.69

Block 3: One-Equation Subsystem (Identity):
dumber of Layers:
(4.12) SK t = SKt—l + SPt_l

- KK,

1 Due to the ambiquity of the coefficient of determination (R 2) in the case of 2SLS-method,
no R2-values for the estimated equation are given above (See Tomek 1973, p. 670). For com-
parison purposes it may, however, be worth mentioning the Revalues computed from the se-
cond stage of the 2SLS procedure: eq. (4.5) 0.82, eq. (4.6) 0.97, eq. (4.7) 0.93, eq. (4.8) 0.98 and
eq. (4.9) 0.78. The alternative values for R 2 based on the residuals of the equations when
actual values for the explanatory endogenous variables are used were not calculated in this
study. Numbers below the coefficients are Student t-values. As to significance levels, see
Section 4.1.1.



471

As for the production response equations in Block 1, their specification was
changed a little on the basis of the estimation results of the recursive production
response model. In equation (4.6), parameters of SK t_ 1; SP t-1 and KK t were
restricted so that the coefficients of SP t-1 and KKt (number of chickens
and culled layers, respectively) were assumed to be one-half of the coefficient
of SKt-1 (number of layers at the end of period t —1) because of multi-
collinearity problems in estimating this equation. Tor the same reason, instead
of using the producer price of eggs and feed price as separate variables, price
ratio (PKT/PR) t was incorporated into equation (4.5). Thus, another nonlinear
combination of an endogenous and an exogenous variable was incorporated
into the model. However, equation (4.5) still satisfies the condition given in
Section 3.3 for applying the 2 SLS method to this nonlinear model.

The production response equations of the basic model need no further com-
ment: all the coefficients of the variables except that of SPt_ 2 are statis-
tically significant at least at the 90 percent confidence level according to the
t-ratios computed1 and, in essence, they are similar to those of the recui'sive
production response model (to the extent they are comparable).

The producer price relation seems to be a plausible model for determination
of the producer price level of eggs in a given period. As expected, the target
price is the main determinant of the producer price level. Negative sign for
variable UK t (net exports) is also logical. It expresses the influence of the market
situation on the producer price level compared to the target price level in a

given period: surpluses in the domestic market tend to reduce the producer
price level below the target price level, and on the other hand, domestic market
prices are over the target price level in the periods of shortages. Also, in the
equation for the retail price determination (4.8) the coefficients of the producer
price and the wage index in commerce are statistically significant and accept-
able from the theoretical standpoint. 2

In the equation for the per capita consumption of eggs, the retail price of
eggs and the disposable income of consumers were deflated by the consumer
price index to reduce the intercorrelation between the explanatory variables
(the linear correlation coefficient between Yt and EI t was 0.99). In this case,
statistically significant coefficients were obtained for both the variables. Accord-
ing to these results, the demand for eggs is rather inelastic: at the mean level
of variables, the price elasticity of demand is —0.60 and the income elasticity
0.37 respectively. Thus, a change of one percent in the deflated retail price
of eggs changes, ceteris paribus, egg consumption by some 0.6 percent. Due to
different model specifications, these price and income elasticities are hardly
comparable to those given in some other sources (Koponen 1964 and Kaarlehto
1958).3

In this context, it should also be mentioned that an attempt was made to
include the linear trend (T) into the consumption equation to take into account

1 As to the use of the t-ratio in the case of simultaneous estimation methods; see Kmenta
(1971, p. 584).

2 The results, however, differ from those obtained by Kaarlehto and Waananen (1965).
The main reasons for this are obviously different model specificiations and estimation periods.

3 For example, Koponen (1964) did not include the consumer price index into his model.



472

possible changes in consumers’ preferences. However, the estimation did not
yield statistically significant coefficients for this variable.

The linear correlation matrix of the equation residuals of the basic model
is given in Appendix 5. As can be seen from the correlation coefficients, this
model structure does not satisfy the assumption of the block-diagonal variance-
covariance matrix of the disturbances, which should hold for estimating equation
(2.1) separately by the OLS-method. The correlation coefficient between the
residuals of equation (4.6) and (4.11) differs from zero even at the 1 percent level
of significance. However, since the correlation does not seem to be very high,
we did not change the estimation of the basic model.

5. Evaluation of the Basic Model

5. 1. On the Methods for Evaluating Econometric Models 1

Even though the structure of the basic model seems assessed on the
basis of the test values computed in estimation stage to be rather plausible
and even though many alternative specifications were tested, we cannot without
further tests be certain, whether this model structure is a good instrument for
assessing alternative price policy directions. There is some evidence in the
field of econometric model building that the model even though estimated
properly and based on sound economic theory may yield simulation results
that are nonsensical. Therefore, further testing is needed for evaluating the
validity of the present model for price policy analyses and this testing has close
linkages to the previous stages of the model building: specification and estima-
tion.2

The problem of evaluating the validity of an econometric model is a difficult
one including many practical, statistical, theoretical and even philosophical
complexities (Naylor 1971, p. 21—22). Without dealing with these problems
in detail in an applicative study of this kind3

, we will only discuss the
methods available for evaluating the validity of the basic model of this
study below.

In general, the nature of the stability properties of an econometric model
constitutes an important measure of its validity, and therefore the analysis of
dynamic stability of the model is useful in the evaluation process (See Labys

1 Some authors use the term validation in this context, while others speak about the eva-
luation of a model; thereby, emphasizing the fact that the validation of an econometric model
(= to prove that the model is true) is, in practice, impossible. In this study, we use the term
evaluation of a model, since we are in this stage interested only in the model’s ability to lore-
cast the values of the endogenousvariables when values for the exogenous variables are changed.

2 For example, in this study the producer price relation was entirely respecified after si-
mulation experiments with the original model. Although the estimation yielded plausible and
statistically significant coefficients for the variables included in the original specification, the
simulation of this model »exploded*: after some observation periods, the simulated values of
the endogenous variables were in complete conflict with the actual values.

3 We refer to Naylor’s presentation (1971, p. 153 163).



473

1973, p. 169). Two different approaches depending upon the characteristics
of the model have been used for exploring the stability conditions: 1) anal-
ytical methods, and 2) simulation methods.1 Because of difficulties in using the
first approach in the case of the present model, which includes nonlinearities
and lags of five periods, no efforts were made to derive the analytical solution
of the model in this study. As stated, for example, by Naylor (1971, p. 299),
the analytical solutions for models of this kind are difficult if such a solution
exists in the first place. Therefore, the simulation remained the only method
for evaluating the basic model.

For the model evaluation purposes, two tests of the basic model were per-
formed in this study using the simulation approach; 1) How well do the simulated
values of the endogenous variables compare with known historical values in the
estimation period (Section 5.2). 2) How accurate are the model’s predictions
of the behavior of the system in the future observation periods (Section 5.3).

Because the basic model is nonlinear in the variables, solving the values
of the endogenous variables from this model is not as straightforward as in the
case of linear models for which the reduced form can be easily derived from the
structural form. However, since the only nonlinear elements the basic model
includes are the products or ratios of one endogenous and one exogenous va-
riable, the reduced form solution approach is still applicable to solving this
nonlinear model for the endogenous variables of a given period. 2

Assuming that the values of the exogenous variables are known, the basic
model can for a given period t be expressed in terms of a linear equation system
(in matrix notations):

. 5
B t Yt -f- L t _j + CZ t Ut

i=1
(5.1)

where B* is the coefficient matrix of the endogenous variables, in which the
coefficients of those endogenous variables belonging to the nonlinear variable
combinations have been adjusted by the values of the corresponding exogenous
variables (for example, coefficient b* for PKT t in equation (4.5) is 0.217/PRJ.
Yt is the vector of the endogenous variables, Bj’s (i = 1 . . . 5) are the coeffi-
cient matrices of the lagged endogenous variables (Y t-i are the vectors of the
lagged endogenous variables) and C is the matrix of the coefficients of the exoge-
nous variables (including the intercept terms). Z is the vector of the exogenous
variables (including 1 for intercept terms of the equations) and U t is the vector
of the structural disturbances. Therefore, with simulation at time period t, the
equation system (5.1) can be solved for Y t and these endogenous variables

1 In analytical method, the stability conditions of an equation system are determined by
solving a set of difference equation system derived from the deterministic part of the model.
Simulation methods permit examination of the response of the endogenousvariables to discover
whether the latter converge to their equilibrium values after the exogenous variables undergo
impact changes.

2 Alternative approaches, which could have been used, are so-called structural form solution
methods, among w hich the Gauss-Seidel method is the most used in the cases of nonlinear
equation systems (See Holt 1967).



474

can be expressed in terms of the coefficient matrices, predetermined variables
and disturbances, whose values we already know:

6

Yt=-B*-1 Z B; Y t_i - B*-1 CZ t + B* -1Uti =1
(5.2)

In this way, when solving values of the endogenous variables for subsequent
periods, the B* matrix changes according to the changes in the exogenous
variables of the nonlinear combinations. Because the inverse of B* has to be
computed separately for every period, this is arelatively costly method. However,
B* being a small matrix (6 by 6), the computation of the inverses of B*’s was
not any big problem in this study'.

Program SIMULATE II a simulation language developed by Holt et. al.
(1967) was used in this study to solve values for the endogenous variables
for different time periods. Only the deterministic simulation procedure was
applied to the basic model (disturbance vector U t was set to its expected value
= 0). SIMULATE II is also capable of performing stochastic simulations of
an econometric model, but because of a great number of computer runs
necessary in stochastic simulation procedures this approach was not applied
to the present model due to the limited resources available for this study. 1

For the same reason, simulating the model for several time periods (for example,
50 100) with the same values of the exogenous variables in order to test
whether the endogenous variables converge to their »equilibrium» values, was
not performed in this study. However, we believe that the two tests taken in
this study will reveal the most essential properties of the model for evaluating
the validity of the model.

5. 2. Historical Simulation
For the purpose of the historical price policy analysis that will be conducted

later in this study, it is important to test, how well does the basic model
simulate historical time paths of the actual system. Therefore, the basic model
was simulated to generate the values of the endogenous variables for the estima-
tion period (1956/1 —1970/11). In this simulation, we treated the model as a closed-
loop simulation model. That is, for given starting values of the lagged endo-
genous variables and given values of the exogenous variables, we solved the
model each period for the current values of the endogenous variables.2 Thus,
in this method also values for the lagged endogenous variables are generated
by the model (except the starting values for them).

1 This may be a shortcoming of the model evaluation process, for Howrey and Kelejian
(1971, p. 299 319) suggested that the properties of dynamic nonlinear models should be studied
in terms of stochastic simulation procedures. They also stated that nonstochastic simulation of
nonlinear models yields results that are not consistent with the properties of the reduced form
of the model.

2 In dealing with different kinds of methods of generating simulated values Labys (1973,
p. 201) calls this approach the final method.



Figures 5.1 and 5.2 contain graphical comparisons of the simulated time paths
and the actual time paths of the key endogenous variables: egg production
(TS t), per capita consumption (CO t) and prices (PKTt and PKV t). A casual
observation of these results would lead us to the tentative conclusion that the

Figure 5.1. Egg Production and Per Capita Consumption in 1956 70: Actual and Simulated
Values.

Figure 5.2. Egg Prices in 1956 70: Actual and Simulated Values, Fmk/kg.

475



476

basic model does a reasonably good job of simulating the behavior of the egg
industry especially in the latter part of the sample period. On the other
hand, the model seems to underestimate both the production and consumption
of eggs in the last periods of the 1950’5. One reason for this underestimation may
be the fact that the starting values of the lagged endogenous variables are not
in accordance with this model structure leading to the underestimation of
production and consumption in the first periods of simulation.

To provide a better criteria for the evaluation of the model’s performance
than the visual examination, Theil’s inequality coefficients were computed
for the endogenous variables.1 Since there are two different inequality coeffi-
cients (U x and U 2) suggested by Theil (1958, p. 32 and 1966, p. 28), the for-
mulas of both the coefficients are given below;

tj =

[/^(Pt - A t )«
TI =

V£( Pt - A t ) 2

VY?f+ /SAt 2 2 V(A t - At_J»
(5.3)

Where P t is the model prediction and A t the corresponding actual value. The
value of Ux is bound to the interval 0 and 1. A value of 0 for Ux indicates perfect
prediction, while a value of 1 corresponds to perfect inequality. U 2 is also bound
by 0 with perfect forecasts, but U 2 has no upper bound. U 2 = 1 when the
prediction model is the naive no-change extrapolation (P t = A t_j).2

There is only one model structure to be evaluated and, therefore, the inter-
pretation of Ux and U 2 test values is a little difficult because of the lack of an
explicit reference point. 3 As for U 2, one possibility is to compare U 2 values from
the basic model with those from the hypothesized naive no-change extrapolation
method referred to above (where U 2 is always = 1). Taking into account that
with this simulation we actually predict values of the endogenous variables
fifteen years onward by the use of the basic model, we can argue that this
alternative U 2 value {= 1) provides a reasonably rigorous basis for the model
evaluation. Since actual changes in the endogenous variables are relatively
small, one may need a rather complete model to predict endogenous values for
the next fifteen years as accurately as does the hypothesized naive no-change
extrapolation model defined above.

The Uj and U 2 statistics computed from the simulated values for periods
1956/1—1970/11 are given in Table 5.1. Since the graphical comparisons suggest-
ed that the discrepancy between the actual and simulated values for the key

1 Many different techniques, such as the analysis of variance, regression analysis, nonpara-
metric tests, spectral analysis, etc., have been suggested for testing of simulation models (See
Naylor 1971, p 159-153).

2 As to the advantages and disadvantages oi these two coefficients, see Lkuthold (1975,
p. 344 346). Theil suggests two reasons why one might prefer U 2 over Up 1) U 2 is related more
directly to the concept of the failure of the forecast and 2) the denominator of Ux depends on
the forecasts and it is therefore not true that the coefficient is uniquely determined by the
mean square predictive error (ref. I.eutfold 1975, p. 345).

3 Of course, comparisons between the inequality coefficients of the final version of the
basic model and the initial specifications of that model would have been possible but probably
not very interesting.



477

endogenous variables is greater in the early part than in the latter part of the
estimation period, Uj and U 2 coefficients were also computed for the simulated
values of periods 1963/1—1970/11.

Table 5.1. Theil’s Inequality Coefficients (I \ and U 2) for Evaluating the »Goodness of Fit» of
the Basic Model for Period 1956 —70. 1

Endogenous Variable
Period

SP t KK, TS t PKTt PKV t COt UK,

] 956 -70
U, 0.0540.076 0.0340.022 0.0170.023 0.129

UM 0.0020.007 0.0020.000 0.0000.001 0.002
Us 0.0190.030 0.131 0,050 0.0470.015 0.085
Uc 0.9790.963 0.8670.950 0.9520.984 0.913

U 2 0.0820.268 1.3450.620 0.5900.773 1.255

1963-70
Uj 0.0560.060 0.0190.015 0.0140.023 0.075

U M 0.0190.031 0.1750.191 0.1940.000 0.218
Us 0.0070.055 0.0370.002 0.1170.111 0.241
Uc 0.9740.914 0.7880.807 0.6890.889 0,541

U* 0.0860.205 0.7330.868 0.9240.736 0.741
1 The numerator of U, has been decomposed into its bias component U M , variance com-

ponent Ug, and covariance component Uq (See Theil 1958, p. 34 35) for evaluation of the
source of the forecast error. The optimal values are: Uk = 0. Ug = 0 and Uq = 1 and the fol-
lowing equation holds for these components: Uj[ + Ug + Uq = 1.

All the U 2 statistics for the latter part of the sample period are smaller than
one; thereby, indicating that the basic model predicts the values of the endogenous
variables more accurately than the naive no-change extrapolation model for
this period. This also holds for the U 2 statistics computed from the entire period
except for those of production (TS t) and net exports (UK t). As stated, this is
mostly due to the discrepancy between the actual and simulated values in the
first years of the sample period. Decomposition of Uj into UM, Us and Uc
leads us to the same kind of conclusions on the model’s ability to forecast the
values of different endogenous variables as the U 2 coefficients. The covariance
component, Uc, got its lowest values from production and net export forecasts
and, in the latter period, also from the retail price forecasts. It should also be
mentioned that regressing the actual values on the generated values the
test of the »goodness of fit» suggested by Cohen and Cyert (1961, p. 120) also
led to the similar results from the standpoint of the model’s validity to forecast
the endogenous values. 1

1 For example, when regressing the actual values of KK t (the number of culled layers)
on the generated values (K f\t ), the resulting regression equation is

KK t = 0.0286 + 0.966Kft t + u t .

(0.1358) (0.095)
where the intercept does not differ statistically from zero (ta 0.21) and the regression coeffi-
cient from unity {tp = 0.36), which are the optimal values of the parameters.



478

5. 3. Ex-Post Forecast Simulation

To evaluate the performance of the basic model outside the estimation period,
ex-post forecasts for the endogenous variables for periods 1971/1—1974/1
(seven periods = 3.5 years) were generated. Similar to the historical simulation,
only the actual values of the exogenous variables and the starting values for
the lagged endogenous variables were given in the model and all the other
variable values were generated by the model. Since the main field which the
basic model will be applied to, is to forecast the future outcomes of different
price policy alternatives for the basis of policy decisions, this test is one of the
most important parts of the model evaluation. In this way we are able to
explore how the model behaves when the exogenous variables (including the
control variable of the policy maker) take value combinations which diverge
from those typical to the estimation period. As stated by Naylor (1971, p.
158), the final decision concerning the validity of the model must be based on
its predictions.

The ex-post forecasts and forecasting errors are given in Appendix 6 and,
in addition, Figure 5.3 contains the graphs of four of the variables whose time
paths were predicted as well as the actual time paths of these variables. As can
be seen from Figure 5.3 a, the production forecasts compare reasonably well
with the actual production except in the first half of the year 1973, for which
the forecast error is some —8 percent (the errors of the other periods are all
smaller than five percent). 1 As to the egg consumption (the total consumption

1 On the other hand, there is always a possibility that errors occur also in the actual pro-
duction figures.

Figure 5.3. Ex-Post Forecasts for the Most Important Endogenous Variables Generated
by the Basic Model and the Actual Values of These Variables in 1971 74/1.



479

(CO • H) t in Figure 5.3 a to ease the scaling problems of the figure), the forecast
errors of the model are under five percent of the actual values in all the periods.
Similarly, the model simulates rather well the actual time path of the retail
price of eggs: the largest forecast errors account only for some 2.5 percent of
the corresponding actual values.

On the other hand, the model consistently underestimates the producer
price over the entire prediction period: the predictive values are, on the
average, some 6 percent lower than the actual values. One reason for this kind
of underestimation is certainly the remodification of the target price rules for
eggs in the early 1970’s due to serious surplus problems. 1 Without taking into
account this rule change, the model continues to generate the producer price
values which are consistent with the earlier provisions of the Price Acts (in the
surplus situation; producer price PKTt = TAV t —0.05TAV t ). This problem
will be further discussed in Section 7.1.

Theil’s inequality coefficients, Uj and U 2, were also computed for these
forecasts (Table 5,2).

Table 5.2. Theil’s Inequality Coefficients from the Ex-Post Forecasts Generated by the Basic
Model for 1971/1-1974/1.

Endogenous Variable

SP t KK t TS t PKT t PKVt CO t UKt

U, 0.0580.093 0.0230.035 0.0080.014 0.061
U M 0.3050.047 0.1680.936 0.421 0 000 0.208
Us 0.2140.247 0.2600.001 0.0000.089 0.192
Uc 0.480 0,706 0.5720.063 0.5790.911 0.600

U, 0.0990.268 1.0001.453 0.3030.505 0.911

The U 2 values suggest that the basic model could predict the values of
most endogenous variables, on the average, more accurately than the hypothesiz-
ed naive no-change extrapolation model. In the case of forecasting egg produc-
tion the models are equally efficient and, as expected on the basis of graphs,
U 2 computed from the producer price forecasts is greater than one referring to
the poor fit of the forecasts made by the basic model. The high value of U M

(the bias component of the numerator of U t ) for the producer price forecasts
indicates the source of the inequality; the difference between the means of the
actual and predicted values.

Although evaluation of the model via predictive tests remained imperfect,
we may summarize the above test results by stating that the basic model seems
to have some good properties for price policy analyses and its forecasting ac-
curacy is at least satisfactory. These results cannot, however, be interpreted

1 The provisions concerning the computing export subsidies were changed such that the
price paid to producers can reach the average target price level even in the surplus situations.
According to the earlier provisions export subsidies were computed on the basis of the lower
limit of the target price range (= 0.95TAV t ).



480

in such a way that the same model structure once estimated is valid for
price policy analyses for a long time in the future. The efficient use of this
econometric model requires continuous development of the model: testing
alternative specifications, and re-estimating coefficients, when the model is used
for analyzing policy alternatives in the course of time.

6. Price Policy Analysis for the Period 1956—70 by Means
of the Basic Model

6. I. General Framework

The econometric model developed and tested above allows us to conduct
historical price policy simulations of certain kinds and, thus, enables us to assess
the outcome of alternative price policies. In this section we will use the basic
model for evaluating the price policy the government had pursued in period
1956—7O when implementing the target price system under the Agricultural
Price Acts. In this task we will concentrate on the question: would some other
price policy or policy mix have been more effective from the standpoint of
society than the actual price policy for achieving the main policy goals
(Section 1). This kind of analysis might be of interest, because the price policy
applied to the egg industry in the above mentioned period has resulted in sur-
plus problems and, accordingly, in increases in government expenditures on
export subsidies.

There are, in theory, a great number of different policies or policy mixes
available to the government to be applied to the egg industry (different kinds
of price support programs, production quota schemes, marketing fees, etc.).
That is why, we wish to emphasize in this connection that the following price
policy analysis does not purport to have discovered an optimal policy for the
egg industry in the above mentioned period. Selecting an optimal policy or
policy mix for a given agricultural product is, in itself, a comprehensive and
complex issue involving many kinds of aspects and considerations. On the other
hand, even the available econometric model restricts our possibilities in this
respect; we can only use the instruments included in the model (and instru-
ments whose impact on the endogenous variables can be assumed to be neg-
ligible). Accordingly, the following analysis can better be viewed as an illustra-
tion of the type of simple price policy analyses that can easily be conducted
by means of the basic model in a real decision situation.

Hence, in this section we will concentrate only on one alternative to the
actual policy and compare these alternatives with each other as to their effec-
tiveness in reaching the two price policy goals. The alternative policy is a policy
mix consisting of: 1) the target price system for maintaining a reasonable ba-
lance between domestic production and consumption. In our policy analysis, we
assume that the domestic egg production should exceed the domestic consump-
tion by five percent as suggested by the Agricultural Committee in 1969.
2) making direct cash payments (which are not related to the production volume



481

of firms) to producers to attain the income objective if the target income level
is not reached at the prevailing target price level in a given period.

This policy alternative henceforth called the equilibrium price policy
alternative consists of two policy instruments for attaining two quantitative
policy objectives. Thus, this policy mix is consistent with the theory of
economic policy: a necessary (but not sufficient) condition for reaching a number
of quantitative objectives is that the policy maker employs with a similarnumber
of policy instruments (Josling 1969, p. 188—189).

Before we can draw conclusions from the desirability of these two policies,
we have to provide some kind of criterion for evaluating alternatives from the
standpoint of society. Some authors (for example, Wallace 1962, p. 590
594, Dardis 1967, p. 597-609, Josling 1969, p. 175-191, CHAYATet. al. 1974,
p. 1—44) have used certain concepts of welfare economics such as the net
social cost1 associated with a given policy as a criterion for ranking the policy
alternatives. Since the objective of our agricultural price policy is mainly to
guarantee the stability of domestic food production and to improve unequal
income distribution in society, 2 this criterion taking into account only the
economic efficiency does not seem to be appropriate for the purposes of this
study.3 In addition, the net social cost of a policy program is difficult to measure
empirically (See, for example, Wallace 1962, p. 580 594).

That is why, the evaluation of alternative price policies in this study will
be made on the basis of more concrete criteria: changes in producers’ incomes,
government and consumers’ expenditure associated with the policy alternative.
An additional reason for using these criteria is that usually the policy makers
are interested in just these issues when making decisions concerning future
policy.

The problem formulation of the price policy analysis of this section car. be
illustrated by means of supply and demand schedules in Figure 6.1.4 The actual
price policy alternative can be characterized by price level px : the target prices
have been set in excess of the level that would have sustained the domestic
market balance in each period. Therefore, to maintain the price level px the
government has to pay subsidies such that egg exporters are able to export
surpluses (= EC in Figure 6.1, the subsidy per unit =Pi pw , where pw

1 The social cost concept can be viewed as the amount of resources or utility forgone when
an unstable system either produces too much or too little (for more detail; see Tweeten and
Tyner 1966, p. 33-38).

2 In this context we have to point out that even though our target price system certainly
improves the income distribution between the agricultural and non-agricultural sectors, it is
not obviously a good measure to reduce income inequality within the agricultural sector.

3 Ray and Tweeten (1976, p. 28) in dealing with different criteria for ranking policy alter-
natives state, for example, that »the chief weakness of net social costs is failure to account for
equity and stability in the system. In general, society seems willing to trade some efficiency
for more equity and stability.*

4 Since the purpose of Figure 6.1 is only to illuminate the problem, the supply and demand
curves have been drawn arbitrarily and, therefore, they may not correspond exactly to the
estimated relationships in this study. The demand curve in Figure 6.1 refers to demand at
the producer price level, which can be derived Irom the original demand relation (See Tomek
and Robinson 1975, p. 44—49).



482

is the world marked price of eggs). Accordingly, the total revenues of egg pro-
ducers are represented by area OABC, consumer expenditures (at the producer
price level) by area OADE, export revenues by area ENKC and export subsi-
dies by area NDBK, respectively.

In the case of the equilibrium price policy direction we assume that the
target price level could have been fixed such that domestic production would
have exceeded domestic consumption by only some 5 percent in each period.
Assuming that price level p 2 corresponds to this objective, area OFGH represents
the total revenues of egg producers in this alternative. Consumer expenditures,
export revenues and export subsidies are represented b}' areas OFIJ, JMLH
and MIGL respectively. The excess resources corresponding to egg produc-
tion HC could have been used for producing other goods in the case of the
equilibrium price policy. Due to the low producer price level of eggs, the net
incomes of producers associated with the equilibrium price policy is most likely
to be lower than in the case of actual price policy and; therefore, we assume that
the government will make up the difference in the form of direct payments
to producers to attain the income target. In this analysis we assume that
direct payments do not influence the quantity of egg produced, since the direct
payment scheme is considered to function so that payments to a single pro-
ducer are not related to the production volume of his firm.

There are many difficulties in this kind of analysis and, therefore, many
simplifying assumptions have to be made. In addition to problems in deriving
the equilibrium target price path for period 1956—7O (discussed in the follow-
ing section), difficulties also arise from estimating the net incomes generated
by those production resources that would have been donated to other pro-

Figure 6.1. Supply, Demand and Net Exports of Eggs at Different
Price Levels.



483

duction activities in the case of the equilibrium price policy. The easiest
and perhaps the most frequently used way to solve this problem would
have been to make an assumption that they would have generated the same
net income for producers in their alternative use without government subsidies
as they have actually generated in the egg industry.

However, since there have been excess capacity in the agricultural sector
as a whole, and since many production factors have had no alternative use
outside agriculture, it may be more realistic to assume that these excess resour-
ces would have been devoted to the production of other agricultural products,
whose domesticproduction may, in reality, have exceeded the domestic consump-
tion in period 1956—70. To simplify the problem, we assume here that they
would have been devoted to pork production only. In practice, another
econometric model would have been necessary to forecast and assess the im-
pacts of this kind of shift in resource use on prices and consumption of pork
and other agricultural products. However, instead of building an econometric
model for these purposes, we made some simple assumptions concerning these
impacts which are discussed later in this study.

6. 2. Derivation of the Equilibrium Price Policy Alternative

The equilibrium price policy alternative was above defined as a policy
under which the domestic egg production would have exceeded the domestic
consumption by five percent in each observation period. Since the target price
is assumed to be the only policy instrument available to the policy maker
for guiding domestic production and consumption, problems in determining
the equilibrium values of the endogenous variables for period 1956—7O relate
to those of finding the target price path that would have led to the above de-
fined policy goal. Since the target price is an exogenous variable in the model,
a straightforward solution of the equilibrium price levels is not possible: we
can only simulate the impact of different target price levels on the development
of the egg industry.

Therefore a kind of iteration approach successive simulation runs by
the model was used to find out the equilibrium target price path for period
1956 70. After selecting the starting path of the target price development 1

and solving the values of the endogenous variables associated with this price
path, we made several successive simulation runs devising the target price
path in each run on the basis of the balance between domestic consumption
and production generated by the previous run. The time paths for the other

1 Actually, the initial target price path to be used in the first simulation run was determined
by solving it from the model which was derived from the basic model through the following
transformations; net exports (UKt ) were defined to account for 5 percent of the total consump-
tion (the equation for net exports; UK t = 0.05 (CO • H) t) and the target price was defined
to be an endogenousvariable in this model. These transformations allowed us to use the model
for solving the equilibrium target price path for the sample period. However, due to the dis-
tortion of the model through these transformations, the model solution resulted in a target
price path with widening fluctuations. The linear trend values based on the generated target
price path were, therefore, used as starting values for the target price in the iteration procedure.



484

exogenous variables were not changed in these simulation procedures. Finally, we
ended up with solutions which yielded the production and consumption paths as
shown in Figure 6.2 and producer price development as shown in Figure 6.3
(the values of the most important endogenous variables are given in Appendix 7).

Figure 6.2. Production and Total Consumption of Eggs Under the Hypothesized
Equilibrium Price Policy in 1956 70, mil. kg. The Time Paths are Generated by the
Basic Model.

Figure 6.3. Producer Price of Eggs Under the Actual Price Policy and Hypothesized
Equilibrium Price Policy as Generated by the Basic Model in 1956 70.



485

As can be seen from the Figure 6.2, this solution does not exactly satisfy
our original assumption of the equilibrium policy alternative. The proportion
of net exports varies around its target value five percent of the total consump-
tion averaging 6.8 percent in the entire period. However, since the resulting
balance between production and consumption is reasonably close to our original
definition of this policy alternative, and since the solution in any case remains
a tentative effort to describe the development of the egg industry under a price
policy different from the actual one in period 1956—70, this solution was consid-
ered as sufficient to represent the development of the egg industry under the
hypothesized equilibrium price policy. 1

In this context we also want to emphasize the tentative nature of policy
simulations of this type. Due to many simplifying assumptions, which have
had to be made, one should be very careful in drawing conclusions from the
simulation referred to above. First of all, we have assumed that the egg in-
dustry resulting from the equilibrium price policy would have reacted and
behaved in the same way as the actual egg industry has done in the sample
period. However, since the equilibrium price policy would have resulted in
different structure and scale of the egg industry, the production responses to
price changes might have been quite different from the actual one in this case.
Another question to be considered at this stage is the values of the exogenous
variables; for example, the feed price developments might have been different
from each other in these two policy alternatives.

In interpreting the results we should also bear in mind that the equilibrium
solution is based on the assumption that the government would have consis-
tently pursued the equilibrium policy alternative over the entire period. Thus,
the solutions associated with the policy alternatives gradually turn separate
over periods of time as the values of the lagged endogenous variables of a given
period associated with the policy alternatives diverge from each other. Accord-
ingly, for example, the generated price paths must be interpreted separately.
Therefore, on the basis of these results we cannot conclude that, for example,
adjusting the prices to the level of the equilibrium policy only in the latter part
of the sample period would have resulted in a reasonable equilibrium between
production and consumption, because the states of the egg industry associated
with the solutions of the endogenous values of the policy alternatives for that
period differ from each other.

6. 3. Development of the Egg Industry Under the Alternative Price Policy
Directions

To get a more detailed view of the differences between the actual and the
hypothesized equilibrium price policy, we will in this section discuss briefly
the development of prices, production and consumption of eggs under these

1 On the other hand, the use of this solution to represent the equilibrium price policy alter-
native may be justified on the basis of the desirability of consistent price policy in the long run.
Without permitting any flexibility in the production-consumption ratio, we might have ended
up in fluctuating price paths with large, casual, downward and upward swings as indicated
by the simulation of the model referred to in the first footnote of this section.



486

alternative price policy directions. Due to its crucial importance in the evaluation
of the effectiveness of the policy alternatives, the main attention is focused upon
the differences between the producer price paths generated by the model in
the simulation of the price policy alternatives (Figure 6.3) 1 to identify those
decision situations, in which the changes in the taiget price level have most
likely been overmeasured from the standpoint of the domestic market equilib-
rium for eggs; thereby, leading to the surplus accumulation in futuie periods.

Our general conclusion from these results is that they reflect clearly the
sensitiveness of the egg producers reactions to changes in the producer price
level and, in this way, also the difficulties of the policy makers in setting the
target price of eggs such that the domestic market equilibrium can be maintain-
ed. According to these results, only slightly smaller price changes compared
to the actual ones during the period 1956—7O would have been required for
maintaining a reasonable market balance in that period assuming that the
policy makers could have adhered strictly to this policy alternative in every
decision situation. On the other hand we have, however, to bear in mind that
the state of the egg industry in the middle of the 1950’s would have provided
the policy makers with a good starting point for this kind of price policy; in this
period net exports namely accounted for some 5 7 percent of the total consump-
tion; thus, being almost in accordance of our definition of the equilibrium policy.

As to differences in the producer price paths in the individual periods, the
results suggest that the price level was in periods 1956/1—1957/1 too low from
the standpoint of sustaining the market equilibrium in the long run. On the other
hand, beginning from the latter part of the year 1957, the target price changes
have been overmeasured leading to the surplus problems and, finally, to drastic
price declines in the early 1960’5. One reason for this kind of relatively high
increase in the target price of eggs in the late 1950’s was obviously the imper-
fect knowledge of egg producers’ reactions in that situation. On the other hand,
one can argue that these increases were also due to the government’s efforts
to adjust the structure of agricultural production according to the existing
export possibilities through the use of price policy measures. The possibility
of exporting the dairy products surpluses significantly deteriorated during
this period (See Sauli 1971, p. 58). That is why to pull excess resources out
of the dairy industry target price levels for the rest of the products included
in the target price group were raised relative to that of milk.

Accordingly, the target price of eggs was raised by 0.35 Fmk through several
separate decisions from September 1957 to September 1960. The simulation
analysis of this study suggests that the balance in the domestic egg market
could have been maintained through a producer price level some 0.10 0.15
Fmk lower than the actual price level.2

1 We have to point out that the producer price path under the actual price policy in Figure
6.3 refers to that generated by the model and not to the historical producer price. The diffe-
rences between these two last mentioned paths can be evaluated on the basis of Figure 5.2.

2 It should be mentioned that the increase of 0.35 Fmk in the target price level also resulted
in the increase in the price ratio between the producer price of eggs and the feed price. A 0.10
0.15 Fmk lower increase in the producer price would have sustained the price ratio approxima-
tely unchanged.



487

Due to the deterioration of exports possibilities and to increasing surpluses,
the producer price of eggs could no longer be supported at its 1960/11 level
(2.65 Fmk) in the following periods. Therefore, egg prices began to fall reaching
their minimum in the first half of the year 1962 (2.14 Fmk). Because of these
difficulties in maintaining the producer price of eggs within the stipulated
target price range, the target price of eggs was decreased by 0.10 Fmk both
in the year 1961 and 1962. According to the policy simulations of this study,
the actual producer prices in these years were lower than the equilibrium pro-
ducer prices generated by the model (Figure 6.3).1

The comparability of the two producer price paths degenerates over periods
of time, because they are in the latter periods associated with quite different
states of the egg industry (i.e. the values of the lagged endogenous variables differ
from each other). However, it seems obvious that the decisions to raise the target
price of eggs by 0.20 Fmk in the year 1966 and 0.15 Fmk in 1968 have especially
contributed to the continuity of the surplus situation in the egg market. It
is also interesting to note that, according to these results, the consistent equilib-
rium price policy for eggs in period 1956—70 would have resulted in only about
a 0.10 Fmk lower producer price level in 1970 than the actual producer price
level for this year.

The production of eggs under the equilibrium price policy (Figure 5.3)
increases steadily from 37.58 mil. kg in 1956 to 43.39 mil. kg in 1970. The
average growth rate per period is 1.2 percent compared to the corresponding
actual growth rate of some 2 percent. 2 The results also suggest that only a minor
upward shift would have taken place in the total consumption of eggs in the
case of the equilibrium price policy: the estimated growth rate of both the time
series is 1.3 percent per observation period.

6. 4. Effectiveness of the Policy Alternatives

Under both the price policy alternatives the self-sufficiency objective would
have been attained in period 1956 70. As for reaching the other policy goal

the income objective we have first to define the target development of
the net incomes of producers employed in our system for the entire period to
be able to assess the effectiveness of these policy alternatives in attaining our
policy goals. Since we have no possibility of defining objectively the target
income of these producers, for comparison purposes we simply assumed that
the actual price policy had been successful in this respect and, accordingly,
the income objective of producers has been reached under this price policy.
Hence, the amount of direct payments which would have been necessary in the
case of the equilibrium price policy can be computed on this basis: such that
this policy alternative would have guaranteed the same income development
as the actual policy. In this respect we have to point out that the evaluation
of the policy alternatives is based on the total target income of the producer

1 As for Figure 6.3, we have to bear in mind that the model overestimates the producer
price for those periods in the case of the actual price policy.

2 Computed from a least squares exponential curve fit.



488

group in our system and, thus, the net income of an individual producer need
not be the same under these two policy alternatives. Accordingly, as a criterion
for evaluating the comparative effectiveness of these policy programs we use
total net savings in export subsidies and consumer expenditures compared with
the total decrease of the incomes of the producers in the system associated with
adopting the equilibrium policy alternative instead of the actual price policy.
Thus, in this analysis we compare whether the net savings mentioned above
are sufficient to compensate the producers for the decreases in their incomes
by making direct cash payments. The figures given in Table 6.1 help us to
assess this question. For conformity, they are based on the values of the en-
dogenous variables generated by the model in both the alternatives and refer
to the entire period.

Table 6.1. Producers’ Gross Revenue, Government and Consumers’ Expenditure as well as
Export Revenue Under the Actual Policy and Their Changes Associated with Adopting the
Hypothesized Equilibrium Price Policy Alternative in Period 1956 70, mil. Fmk. 1

Actual Equilibrium Price Policy
Policy Eggs Pork Total

Change Change Change
mil. Fmk mil. Fmk mil. Fmk mil. Fmk mil. Fmk

1) Producers’Gross Revenue ....
1848.50 1555.01 -293.49 260.84 -32.65

2) Export Revenue 220.83 60.70 -160.13 138.79 -21.34
3) Consumers’ Expenditures 1863.73 1855,99 —7.74 2 —7.74
4) Government Expenditures:

Export Subsidies 261.83 65.49 -196.34 146,23 -50.1I 3

5) Net Change in Consumers’
and Government Expenditures
(items 3 + 4) - - -204.08 146.23 -57.85

6) Net Savings 4 25.20

1 Data, on which the figures are based, is given in Appendices 1,2, and 7.
2 A part of the increase in pork production could have been consumed at home: in periods

when pork has been actually imported. But the total consumption expenditures on pork would
not have changed since the increase in domestic production would have only been substituted
for imports.

3 Actually this is a net change. Since pork imports would have decreased during some
periods, government revenues from import levies would have decreased by 4 mil. Fmk. This
decrease has been subtracted from the total net savings in export subsidies.

4 Assuming that the production costs of the system would not have changed. Computed
by subtracting item 1 from item 5.

The main assumptions which have been made during these computations are:
1) The excess resources actually used in egg production would have been

donated to pork production in the case of the equilibrium price policy
alternative. 1 The transformation coefficient between pork and eggs was

1 In the first few observation periods the egg production of the equilibrium price policy
is a little higher than that of the actual price policy alternative. In these cases we assumed
that resources necessary to produce the additional egg output in these periods would have
been taken from resources actually assigned for pork production.



489

estimated to be 0.8 (i.e. in terms of resource requirements 8 units of pork
are equivalent to 10 units of eggs).

2) Changes in pork production do not change theprices and domestic consump-
tion of pork. Therefore, changes in pork production influence exports and
imports of pork only.

3) Although the consumption level of eggs under the equilibrium policy is
higher than under the actual policy, we assumed that this would have
exerted no influence on the consumption levels of other agricultural pro-
ducts for two reasons: 1) the differences in egg consumption levels are
small (on the average, the egg consumption under the equilibrium policy
alternative would have been 0.7 percent higher than under the actual price
policy), and 2) it was difficult to identify the products for which eggs would
have been substituted.

4) In computing the gross revenue of producers in the system we took into
account only the final products (eggs and pork). The gross revenue figures
are based on the total production including the direct home consumption
on farms. Similarly, consumer expenditures include also the home consump-
tion on farms.

5) The prices of the exports and imports of eggs and pork were assumed to be
independent of the agricultural policy pursued in Finland. Thus, the dif-
ferences between per unit subsidies for eggs associated with the policy
alternatives are determined by the differences between domestic price levels
associated with the policy alternatives.

Under these assumptions, the gross revenue of producers in this system
would have decreased in all by some 32.7 mil. Fmk in period 1956 70, if the
government had adopted the equilibrium price policy alternative instead of
the actual policy (i.e. 1.6 percent of the actual gross revenue in that period).
The decrease is mainly attributable to the lower price level of eggs; the excess
resources assumed to have been devoted to pork production would have generated
approximately the same gross revenue as they have actually done in the egg
industry under the actual price policy. Changes in the intermediate production
costs of the system due to the policy shift are difficult to estimate. However, if
we simply assume that they are approximately the same under both the policy
alternatives, we end up with the conclusion that the net incomes of the producers
under the equilibrium price policy would have been some 32.7 mil. Fmk lower
than under the actual policy. This is the amount which the government ought
to have paid to producers in the form of direct payments in order to attain the
income objective of the producers in the entire period.

According to these results, the total export subsidies for eggs which the
government ought to have paid in the case of the equilibrium price policy would
have totalled some 65 mil. Fmk in the entire period. Since almost all the pork
produced by this system would have been exported, the government subsidies
for pork exports would have increased by 142.5 mil. Fmk. Moreover, the total
amount of the import levies from pork imports would have reduced by some
4 mil. Fmk in the case of the equilibrium policy alternative. Thus, the policy
shift would have resulted in a net savingos to the government of 50.1 mil. Fmk.



490

Moreover, domestic consumers could have obtained eggs at lower prices
in the case of the equilibrium policy than under the actual policy. However,
consumers’ total expenditures on eggs would have decreased only moderately
(by 7.7 mil. Fmk in the entire period), because the total egg consumption would
have increased in the equilibrium policy alternative. As stated before, we
have not taken into account the substitution effect between eggs and other
agricultural products in computing this change in consumer expenditure due
to a policy shift. 1

Thus, the net savings to the government and consumers, resulting from the
adoption of the equilibrium price policy would have been sufficient to compen-
sate the losses of the egg producers incurred by the lower producer price level
of eggs resulting from that alternative. If the government had made up the
difference of 32.7 mil. Fmk between the net incomes of producers under thepolicy
alternatives in the form of direct payments, the net savings to the government
would have totalled some 17 mil. Fmk in the entire period. This together with
the decrease in consumers’ expenditure suggests that the net savings to society
through adopting the equilibrium price policy alternative instead of the actual
policy would have been some 25.2 mil. Fmk in period 1956—70under the assump-
tions mentioned above.2 The net savings discussed above refer to period 1956—

70 only. In practice, the actual price policy resulted in an excess capacity of some
40 percent in the egg industry by the year 1970. Since the rapid downward
adjustment of the price level is not possible without causing unfair losses to
some producer groups, the government will most likely continue to pay export
subsidies for eggs also in the future. Therefore, in evaluating the policy alterna-
tives we should also take these future periods into account. However, there
are many difficulties in estimating the future savings of the equilibrium price
policy and, so, we will not deal with them in this context. To get some idea of
their likely magnitude, we will only refer to the net savings of the equilibrium
price policy in 1970, which is the last year of our simulation period. The reduction
in the producers’ gross revenues would have been some 3 mil. Fmk, while decrea-
ses in government and consumer expenditure would have totalled to some
15 mil. Fmk. Thus, the net savings resultings from the adoption of the equi-

librium price policy would have been some 12 mil. Fmk in the year 1970 only.
However, we must point out that future net savings depend on the world mar-
ket prices of eggs and pork (in the year 1970 they were comparatively favourable
to pork).

To summarize the results discussed above we may conclude that the equi-
librium price policy alternative the target price system together with the
direct payment scheme would have been a more effective vehicle for attain-

1 Consumers’ expenditure under the equilibrium price policy would have been some 21 mil.
Fmk lower than the same consumption level at the actual retail prices. Assuming that the price
ratios between foodstuffs would have been »correct» from the standpoint of consumers’ welfare,
this difference can be regarded as a net saving to consumers from the equilibrium price policy
alternative, if we also take into account the substitution effects.

2 The net savings are mainly due to the fact that supporting consumers abroad in the form
of export subsidies for eggs and pork would have decreased if the government had adopted
the equilibrium price policy for eggs.



491

ing our price policy goals in the case of the egg industry during the period 1956
70. This conclusion is based on the net savings throughout the entire period.
As to individual observation periods, only in a few periods net savings resulting
from the equilibrium price policy are not positive which is mainly due to the
low world market price of pork.

7. Ex-Post Simulation of Alternative Price Policy Directions
and Their Evaluation

7. 1. Policy Alternatives and the Simulated Outcomes of These Policies

In this section we will discuss in brief the results of the ex-post simulations
of different policy alternatives for period 1971—74 generated by the basic
model of this study. These ex-post simulations were done above all to get some
idea of the nature of policy analyses that can be conducted by the model and,
also, to evaluate the actual price policy the government has been pursuing in
that period. We concentrated on only a few alternatives in which the policy
makers would we suppose have been interested in fixing the target price
level of eggs for the year 1971.

For defining the policy alternatives to be analyzed in this section, we assume
that the policy makers in that situation anticipated because the number
of layers in the end of the year 1970 was relatively high that the domestic
egg production would grow substantially also in the year 1971 and exceed the
domestic consumption by some 50 percent at almost any reasonable price level.
Because such production increases are most likely to lead to a substantial
growth in government expenditures on export subsidies in the future, we assum-
ed that the policy makers would have been interested in the possibilities of
changing the direction of the price policy for eggs: in such a way that a better
balance between the domestic production and consumption could be attained
in the future. In this respect we assumed that they would have been consider-
ing adjusting the domestic production gradually to the domestic consumption
by using target price levels as their policy instrument. Since the other objective
of our agricultural policy to promote producers’ net incomes to a level com-
parable with those of the other sectors in the economy most probably cannot
be achieved at the price level required for the adjustment of egg production and
consumption, we assumed that the policy makers would have been ready to
adopt some kind of direct payment scheme for attaining this policy objective.1

Let us assume that the policy makers would have specified their alternative
policy goals in terms of annual adjustment rates: to attain a reasonable equi-
librium between the domestic production and consumption

1) in 5—6 years (alternative 1),
2) in 7—B years (alternative 2),
3) in 10—11 years (alternative 3),

1 Again we have to point out that direct payments to an individual producer must not be
related positively to the production volume of his firm in order that this policy alternative should
be effective.



492

and simultaneously taking care to reach the income objective of the producers
in the system (= to raise the total net income of the producers at the same rate
as the average incomes of employees in the other sectors of the economy increases).
Let us further assume that the policy makers could have used the basic model
of this study in searching for the target price paths which would have led the
egg industry towards the above mentioned goals and for simulating the future
production and consumption paths as well as the price levels associated with
each alternative.

The solutions of the endogenous variables which correspond to the above
mentioned adjustment target of each policy alternative, were generated by the
model using the same kind of iterative procedure as was used in the previous
chapter to find out the equilibrium target price path. 1 Again, we have to empha-
size that these solutions only approximate the policy directions specified above:
the adjustment rate varies between periods in each policy alternative, however,
the average rate in each solution is such that the equilibrium objective could
be attained within the defined time period.

Figures 7.1 a and 7.1 b indicate the simulated development of egg pro-
duction and net exports under each policy alternative (values for the most
important endogenous variables are given in Appendix 8). Since we already

1 To get starting values for the target price paths for the iteration procedures, the model
was reformulated by fixing the net exports and consumption to each other with the following
equation: UK t =K t • (CO • H)t , where K t is an exogenous variable, whose values were specified
separately for each adjustment alternative. For example, in the first alternative K = 0.55
for 1971/1, K = 0.50 for 1971/11, etc. In this model the target price (TAV) was an endogenous
variable, whose values could be generated by the model.

Figure 7.1. Production and Net Exports of Eggs Under Different Policy Directions in
1971 74. Production and Net Exports Figures are Generated by the Basic Model, mil. kg.



493

know the actual price policy the government has pursued in period 1971 1974,
we use the values generated by the simulation of the actual price policy as a
reference point in evaluating the effectiveness of these policy alternatives in
the next section. One reason 1 for using generated values rather than actual
ones is to provide a better consistency between the outcomes of different policy
alternatives, because the model seemed to underestimate the producer prices
in the simulation of the actual price policy for periods 1971/1—1974/11, as
stated in Section 5.3.

The differences between producer price time paths associated with the
actual policy and the above specified policy alternatives are as follows:

Period Alternative 1 Alternative 2 Alternative 3
Fmk/kg

1971 I -0.21 -0.12 -0.05
II -0.13 -0.04 0.04

1972 I -0.12 -0.04 0.02
II -0.13 -0.08 -0.06

1973 I -0.18 -0.17 -0.17
II -0.36 -0.34 -0.31

1974 I -0.42 -0.35 -0.19
II -0.49 -0.34 -0.19

As a general conclusion from these figures we can state that a compara-
tively big downward adjustment of the actual price level would have been
necessary to adjust domestic egg production to domestic consumption within
the 5 —6 years proceeding from the situation at the end of the year 1970. Accord-
ingly, the producer price level for the latter part of the year 1974 is some 0.50
Fmk lower under the first policy alternative than under the actual policy.
However, the results suggest that policy alternative 3 (adjusting production to
domestic consumption within 10—11 years) would have required only a minor
downward adjustment of the actual price level during these first years assuming
that the policy makers would have adhered consistently to this alternative from
the year 1971: at the end of the test period (1974/11) the producer price of this
alternative is only 0.19 Fmk lower than that of the actual policy.

The average changes in the values of the key variables per year in different
alternatives are as follows:2

Production Net Exports Total Consumption
Alternative TS t UKt (CO • H)t

Average change per year, percent

Alternative 1 —4.3 —24.2 +2.7
Alternative 2 —2.2 —14.5 +2.7
Alternative 3 —O.B 8.2 +2.6

1 Using the generated values of the endogenous variables to represent the outcomes of the
actual policy is also consistent from the standpoint of analyzing future policy directions in a
real decision situation which we are, in fact, attempting to do in this section.

2 Average changes per year have been computed by using a least squares exponential curve
fit (estimated from yearly observations).



494

For comparison purposes it may be worth mentioning that the average
growth rate of egg production under the actual policy is +1.6 percent per year
computed from the generated values for egg production. Accordingly, also the
average growth in net exports is positive but small; some 0.2 —0.3 percent. 1

The average growth in egg consumption is almost the same in each alternative:
2.6—2.7 percent per year.

In practice, there is no way to analyze empirically the reliability of these
solutions (i.e. whether the adjustment target of each alternative would have
been attained under the corresponding producer price paths generated by the
simulations or not). The question remains open in this study as it will remain
open also in analyzing and evaluating alternative policy directions for future
periods in a real decision situation. Changes in egg production in a given period
can be assumed to depend on changes in the relationship between the marginal
cost of producing one unit of eggs and the producer price of eggs. Estimating
the marginal costs of the entire egg industry associated with different scales
of the industry is, however, difficult (if not impossible) 2

. Therefore, in this respect
we must rely merely on the model’s validity to forecast producers’ and consu-
mers’ reactions to changes in price levels. Evaluating the model indicated that
the model was able to forecast rather well the development of the egg industry
for period 1971—74 under the actual price policy. Assuming that the reactions
of producers and consumers under the alternative price policy directions would
not have been quite different from their actual reactions, the solutions asso-
ciated with alternative price policies can be considered to represent the devel-
opment of the egg industry under these policy alternatives.

7. 2. Evaluation of the Effectiveness of the Policy Alternatives in Attaining
Policy Goals

In order to evaluate the effectiveness of the policy alternatives defined
above in attaining our policy goals (producers’ income target and the self-
sufficiency) we made calculations similar to those in the previous chapter in
comparing the equilibrium policy alternative to the actual price policy. Thus,
for comparison purposes we assumed that the income target of producers in
the system had been reached under the actual price policy in period 1971—74.
As to the alternative use of the excess resources which would have been released
from the egg industry if the government had adopted one of the price policy
alternatives I—3, we made two different assumptions: 1) they would have

1 Actually, the exponential growth curve did not fit very well to the observational values
of net exports.

2 Some book-keeping results on the profitability of the egg production (See Westermarck
1973, p. 230 232) suggest that the average production costs of producing eggs were 3.12 Fmk
per kg in 1971 and 3,21 Fmk per kg in 1972. However, since these results are based on a com-
paratively small group of firms which is not a random sample from the Finnish egg industry,
these results cannot be generalized. Moreover, the variations of the production costs were wide
even in this group: in 1972 the production costs of ten most profitable firms were some 14 per-
cent below and those of ten less profitable firms was 15 percent above the average level.



495

been devoted to pork production (See Section 6.4), and 2) the excess feed grain
would have been exported directly without processing it into eggs. Also, the
entire increase in pork production that would have been produced by these
recources, would have been exported since, in reality, the domestic pork pro-
duction already exceeded the domestic consumption in that period.

Given these policy goals and assuming that all the other effects associated
with each policy alternative1 are approximately equal from the standpoint of
society, the desirability of these policy alternatives depends on government
and consumers’ expenditures necessary to attain the specified policy goals
under these policy alternatives. In this respect, we can evaluate the compara-
tive effectiveness of policy alternatives I—3 and the actual policy in period
1971—74 by exploring the net savings to society associated with adopting
hypothesized policy alternatives I—3 instead of the actual price policy, and
which are presented in Tables 7.1 and 7.2. Again we have to point out that
these figures are based on many assumptions whose validity in the real world
is difficult to verify. However, we believe that these figures indicate at least
the magnitudes of the net savings associated with each price policy alternative
for assessing the desirability of these policy tools in guiding the egg industry
according to the goals of our agricultural policy.

As can be seen from Table 7.1, producers’ gross revenue would have decreas-
ed altogether by some 55 mil. Fmk in the entire period, if the government had
fixed the target prices of eggs according to alternative 1 and the excess re-
sources had been devoted to the production of pork. On the other hand, the
government would have saved some 46 mil. Fmk in export subsidies by adopting
policy alternative 1 instead of the actual policy, since egg exports would have
decreased and exporting pork would not have required the same amount of
export subsidies as the equivalent amount of eggs (in terms of resource require-
ments). Moreover, consumer expenditures on eggs would have decreased by
some 26 mil. Fmk. Assuming that the intermediate production costs of the
system would not have changed due to the policy change, the income target
of producers in the system would have been possible to attain by compensating
producers by some 55 mil. Fmk in the form of direct cash payments.

Hence, the net savings to society from adopting policy alternative 1
instead of the actual policy would have totalled some 17 mil. Fmk in period
1971—74. As to their distribution between the individual years within this

four year period, our results suggest that the net savings would have grown
with time. However, there would have been some casual variations around
this general trend depending mainly on the variations in the export prices of eggs
and pork: the biggest net savings would have been obtained in 1973 (11.2 mil.
Fmk) and they would have been negative in 1971 (—2.5 mil. Fmk). In this
context no attempts were made to measure net savings associated with the
adoption of policy alternative 1 beyond the year 1974 because of difficulties

1 Adopting totally different policy directions may in the short run result in some side-
effects negative or positive such as unfair losses to some individual producers, capacity
problems in the marketing sector, changes in feed imports etc.



496

Table 7.1. Estimated Producers’ Gross Revenue, Export Revenue as well as Government and
Consumer Expenditures under the Actual Price Policy in 1971—74 and their Changes Associated
with Adopting Price Policy Alternatives 1—3. Excess Resources Were Assumed to Be Used
in Pork Production.

Actual Policy Price Policy Alternative
mil. Fmk 12 3

Change, mil. Fmk

1) Producers’ Gross Revenue
Eggs 1012.49 -163.65 -109.37 -60.82
Pork - 108.16 65.53 31.79
Total & Net Change 1012.49 - 55.49 - 43.84 -29.03

2) Export Revenue
Eggs 185.82 - 60.25 - 37.45 -19.01
Pork - 79.30 48.12 23.59
Total & Net Change 185.82 19.05 10.67 4.58

3) Domestic Sales
Eggs 919.76 - 25.85 - 18.67 -11.83

4) Government Expenditures1

Egg Exports 285.40 -104.53 - 67.94 -36.58
Pork Exports - 58.27 36.13 18.31
Total & Net Change 285.40 - 46.26 - 31.81 -18.27

5) Consumers Expenditures 519.76 25.85 18.67 —11.83

6) Net Change in Government
and Consumers’ Expenditures
(items 4 + 5) - - 72.11 - 50.48 -30.10

7) Net Savings2
- 16.62 6.64 1.07

1 Only export subsidies have been taken into account in computinggovernment expenditures.
2 Assuming that the production costs do not differ between the alternatives.

in forecasting the prices and exports of pork for which we had no econometric
model.

Similarly, adopting policy alternative 2 instead of the actual policy would
have resulted in net savings to society. The estimated savings are quite
logically a little smaller than those in the first policy alternative: some 6.6
mil. Fmk in the entire period. In terms of government and consumers’ expen-
ditures in supporting producers’ incomes and maintaining self-sufficiency in
eggs, policy alternative 3 and the actual policy do not differ substantially
from each other, because the development of the egg industry under these al-
ternatives would not have been very different due to the long adjustment period
allowed in alternative 3. Net savings associated with adopting policy alternative
3 would probably have grown in the future periods. However, even evaluating
on the basis of period 1971—74 policy alternative 3 would have been a little
more effective than the actual policy in attaining the policy goals.



497

When we, on the other hand, assumed that producers would have sold the
excess feed grain at prevailing producer prices and all this feed grain had been
exported directly without processing it into livestock products we ended
up with the results given in Table 7.2. In these calculations the feed requirements
per kg of eggs in the withdrawing egg production was assumed to be 4.0 feed
units (prices, export subsidies for feed grain used in the analysis are given in
Appendix 9).

Table 7.2. Estimated Changes in Producers’ Gross Revenue, Export Revenue, as well as in
Government and Consumers’ Expenditures Associated With Adopting Policy Alternatives
1—3 Instead of the Actual Policy in 1971 74. Feed Grain Released from Egg Production is
Assumed to be Exported.

Price Policy Alternative
12 3

Change, mil. Fmk

1) Producers’ Gross Revenue
Eggs -163.65 -109.37 -60.82
Feed Grain 43.78 27.67 3.36
Net Change -119.87 - 81.70 -57.46

2) Export Revenue
Eggs - 60.25 - 37.45 -19.01
Feed Grain 42.13 27.14 3.50
Net Change - 18.12 - 10.31 -15.51

3) Domestic Sales
Eggs - 25.85 - 18.67 -11.83

4) Government Expenditures
Egg Exports -104.53 - 67.94 -36.58
Feed Exports 7.53 4.44 0.26
Net Change - 97.00 - 63.50 -36.32

5) Consumer Expenditures 25.85 18.67 —11.83

6) Net Change in Government and Consu-
mers’ Expenditures (items 4 + 5) —122.85 82.17 —48.15

As to Price Alternative 1, producers’ gross revenue would have decreased
by some 120 mil. Fmk in 1971—74, which accounts for some 12 percent of the
actual gross revenue in this period (Table 7.1). Likewise, the export revenue
would have decreased by some 10 percent. However, on the other hand decreases
in government expenditures on export subsidies and in consumers’ expenditures
would have totalled to some 123 mil. Fmk in the corresponding period. Thus,
our calculations suggest that net decreases in export subsidies and consumers’
expenditures would have been sufficient for compensating producers’ losses



498

due to adopting price policy alternative 1 instead of the actual price policy in
period 1971—74. Even the gross revenue of the producers could have been
raised to the same amount as under the actual price policy by paying direct
cash payments from the net savings of the government and consumers. In
addition, we have to point out that some of the processing costs of egg produc-
tion could have been eliminated (such as labor, maintenance costs, etc.). In
this study we did not, however, estimate these savings. Their magnitude is
mainly determined by the possibilities to allocate these inputs to alternative
uses. For example, these savings can be estimated differently depending upon
the assumption of the impact of price policy alternative 1 on the structural
development of egg production (alternative assumptions in this context are,
for example, that price policy alternative 1 had reduced investments in new
producing units or that this alternative had eliminated existing producing
units). We have no knowledge of these impacts and, therefore, we have to
bear in mind that the above mentioned 120 mil. Fmk can be regarded as an
approximate maximum for a necessary compensation in order to maintain the
net incomes of producers in the same way as the actual price policy has done.

Similar conclusions can be also drawn from the results concerning price
policy alternative 2 in this analysis. On the other hand, our results suggest
that price policy alternative 3 associated with exports of excess feed grain would
have resulted in a net reduction of 48 mil. Fmk in the government and consu-
mers expenditures in 1971—74, while the decrease in producers’ gross revenue
is some 57 mil. Fmk. Thus, when ranking the actual price policy and price policy
alternative 3 as defined in Table 7.2 in terms of their effectiveness, decrea-
ses in the production costs of the sector are an essential question. In addition,
differences in the future development of the egg industry between policy alter-
natives are an important aspect in this regard.

Hence, on the basis of these analyses we would recommend a new policy
mix for supporting egg production in Finland instead of employing only the
target price system: 1) using the target price system for adjusting the domestic
egg production gradually to the domestic consumption and for maintaining a
reasonable market balance after the adjustment period, and 2) introducing a
direct payment scheme for attaining producers’ income target if the target in-
come level cannot be attained at the existing target price level in a given period
(in this scheme the yearly payments per firm must not be related to its pro-
duction volume). The results imply that the faster the adjustment rate, the
greater the net savings to society under the relatively strict assumptions made
in this study. However, there are many factors other than discussed above
which influence the desirability of a certain adjustment policy (for example,
equity considerations among producers and consumers). Moreover, we have
analyzed only one type of policy mix above, and we do not know the effectiveness
of other policy instruments available to the government. On the other hand,
many kinds of assumptions can be made on the alternative uses of excess re-
sources and the results of our analysis might change depending on these assump-
tions. Therefore, we may conclude from these results that adjusting the egg
production to domestic consumption would have provided net savings to so-
ciety in period 1971—74, but further analyses may be necessary to decide what



499

instruments are the most effective in this situation. But such analyses are
beyond the objectives of this study.

These empirical results conform with earlier theoretical findings that in
terms of government expenditures the guarantee price system like our target
price system, in which the price level is controlled by exports (and imports),
is not as effective as the direct cash payment scheme in supporting agricultural
prices if the supply and demand for those products is elastic. 1 Moreover, one of
the goals of our agricultural policy should be improving the income distribution
among the agricultural producers, and in this respect the target price system
is also inefficient (it may even worsen the income distribution within agri-
culture). Therefore, the additional advantage of adopting a direct payment
scheme for supporting producers’ incomes is that it provides the government
with a vehicle to equalize the income distribution among producers. There
are, of course, difficulties in implementing the direct payment scheme (See
Sisler 1966, p. 49) but we do not think they are unsolvable.

8. Summary and Conclusions
Providing appropriate information for price policy decisions concerning the

Finnish egg industry was the primary purpose of this study. As described in
Section 1, the egg industry has been one of those sectors in Finnish agriculture,
where in order to attain our major policy goals an price policy employing only
a target price system as established in the Agricultural Price Acts at
reasonable government and consumer costs has been problematic during the
last two decades. The difficulties can be attributed to a number of reasons,
among which the sensitivity of egg producers’ responses to price changes and
the recent rapid technological development in egg production are probably
the most relevant.2 This is why, in order to ease the problems of the policy
makers in measuring the necessary changes in the target price of eggs we deri-
ved an econometric model for the Finnish egg industry. This model can be used
in decision situations for assessing the impact of alternative target prices (and
those of the other exogenous variables, too) on the future development of the
egg industry. After that, we used the model for simulating the development of
the egg industry under the hypothesized price policies and the actual price
policy proceeding from the different situations of the egg industry in the past
in order to evaluate the effectiveness of certain policies/policy mixes in attaining
the major policy goals: producers’ income target and a reasonable level of
self-sufficiency in eggs.

1 As to the effectiveness of different support programs in terms of government expenditures;
see Gulbrandsen and Lindbeck (1973, p. 243 247).

2 We certainly recognize that one of the basic reasons is also the excess capacity of the
Finnish agriculture as a whole because of the mechanism of the target price system itself. We
may assume that the policy makers have tried to minimize the total costs of supporting agri-
cultural prices, and therefore measuring changes in the target price of eggs has also depended
on the market situation of the other target price products. We have not, however, paid much
attention to such considerations in this study mainly due to the fact that it is in practice possible to
analyze the comparative effectiveness of alternative price policies for the egg industry separately.



500

The specification of the basic model is given and discussed in Sections 2.3
2.4. It is an eight-equation model consisting of three blocks: a six-equation
simultaneous block and two single-equation blocks, which link together as shown
in Section 2.4. As to the production response, it was assumed that producers
are able to influence the total egg production via two major pioduction de-
cisions: purchasing chickens and culling layers. Accordingly, behavioral equa-
tions including the factors which influence the profitability of egg production
were specified for these activities. It was assumed that there is a lagged res-
ponse to price changes in purchasing chickens and, therefore, polynomial lag
formulations were applied in describing the relationships between the number
of chickens and the prices of eggs and feed. Our policy instrument the target
price of eggs was included in the producer price relation besides the variable
reflecting the discrepancy between domestic production and domestic con-
sumption of eggs (net exports). In addition, the basic model includes equations
for per capita consumption, retail prices and for net exports of eggs. It is a
dynamic model with lags of five periods (the model is built for semiannual data)
and nonlinear in variables (Section 2.4).

Due to its crucial importance, relatively much attention was paid to the
estimation of the basic model (Section 4). In addition to testing many kinds
of alternatives for the final specification through estimation experiments, the
stepwise regression method with regard to observations was applied to explore
possible time-related trends in the coefficients of the equations concerning
the determination of egg production (the recursive production response model
of four equations was derived for this purpose; see Section 2.5). A variant of
the 2 SLS-method was applied to the estimation of the six-equation simulta-
neous block (Section 3.3.1). Otherwise, the OLS-method was applied to the
estimation of the coefficients.

The validity of the basic model in simulating the actual behavior of the egg
industry was tested through deterministic simulations (Section 5) using two
types of tests: 1) historical simulation for the estimation period and 2) ex-post
simulation for period 1971—74/1. The time paths of the endogenous variables
were generated by giving only the values of the exogenous variables and the
starting values of the lagged endogenous variables to the model. Figures 5.1,
5.2 and 5.3 contain the graphical comparisons of the simulated time paths
of the key endogenous variables (See also Appendix 6). Theil’s inequality
coefficients (Uj and U 2) were also computed to evaluate the »goodness of fit»
(Tables 5.1 and 5.2).

We evaluated the comparative effectiveness of two alternative price policy
directions for the period 1956—70 in attaining the main goals of our agricultural
price policy: 1) the actual price policy the government has implemented under
the Price Acts and 2) the hypothesized equilibrium price policy (Sections 6.1
and 6.2). According to this alternative, the policy makers were assumed to
use two policy instruments: the target price system and a direct payment
scheme, in which payments to an individual producer are not related to the
production volume of his firm. We defined attaining producers’ income target
(as established by the Price Acts) and maintaining a reasonable level of self-
sufficiency in eggs (surpluses to be exported were defined as five percent of the



501

total consumption) as our price policy goals. Thus, the equilibrium price policy
alternative is consistent with the theory of economic policy; the necessary
condition for attaining two objective policy objectives is that we employ two
policy instruments. The policy mix with these instruments is clear: setting the
target price of eggs to a level which will maintain the domestic market balance,
and making direct cash payments to producers if and when the income target
cannot be achieved on the resulting producer price level.

The basic model was used for deriving the equilibrium target price path
for 1956—7O (Section 6.2) and for generating the development of the egg in-
dustry under these policy alternatives (Section 6.3) proceeding from the state
of the egg industry at the end of 1955. Our results suggest that, as a whole,
only a comparatively small downward adjustment of the actual price level
would have been required for maintaining the market balance of eggs in the
period 1956—7O assuming that the target price levels would have been set
consistently according to the market balance considerations through the entire
period. The criteria we used for ranking our policy alternatives were govern-
ment and consumer expenditures associated with these alternatives in attaining
our policy goals. Due to the complex nature of this kind of evaluation of diffe-
rent policy alternatives we had to make many simplifying assumptions, which
are described in Section 6.4. We assumed, for example, that the excess resources
which would not have been devoted to the egg industry in the case of the equilib-
rium price policy would have been employed in the pork sector. Given these
assumptions, our analysis suggested that the equilibrium price policy alternative
would have been a more effective policy mix than the actual price policy (Section
6.4). The net savings resulting from the adoption of the equilibrium price policy
alternative instead of the actual price policy is, however, relatively small in
the test period mostly due to our realistic assumptions of the alternative
use of the excess resources. On the other hand, we may argue that the net
savings through adopting this policy would have grown in the future (beyond
the year 1970). But due to many difficulties we did not estimate the future
savings of the equilibrium price policy.

We also conducted ex-post price policy simulations with the basic model
to evaluate the alternative price policies for the period 1971—74 (Section 7)
and, again, to give some idea of the type of the price policy analyses which
can be done by means of the basic model in real decision situations. In order
to define our policy alternatives, we assumed that the policy makers would
have been interested in adjusting the domestic production gradually to the
domestic consumption by the use of the price policy measures and simulta-
neously taking care to attain the income target of producers in the system by
paying direct cash payments to producers if and when necessary. The adjust-
ment alternatives were defined as attaining a reasonable balance between pro-
duction and consumption of eggs 1) in 5—6 years, 2) in 7—B years and 3) in
10—ll years. The basic model was used for generating the hypothesized devel-
opment of the egg industry under those policy alternatives for period 1971—74
proceeding from the state of the egg industry at the end of the year 1970 (Sec-
tion 7.1). The comparative effectiveness of the policy alternatives and the ac-
tual policy in attaining our two policy goals was then evaluated on the basis



502

of similar calculations as was used in evaluating the price policies in the period
1956—70 (the criterion for ranking alternatives was government and consumer
expenditures necessary to attain the policy goals). The income path generated
by the simulation of the actual price policy was regarded as a target income
development in this period.

The results implied that, proceeding from the situation at the end of the
year 1970, the target price changes ought to have been measured much smaller
compared with the actual ones in order to attain a reasonable balance between
production and consumption of eggs within 5—6 years; in the latter part of
the year 1974 the producer price of this alternative is some 0.50 Fmk lower
than that under the actual price policy. However, in terms of government and
consumer expenditure this adjustment policy alternative would have been
more effective in attaining the policy goals than the actual price policy under
the assumptions made in this study. Similar conclusions can be also drawn
from the results concerning price policy alternative 2 in this analysis. As to
price policy alternative 3, the results were inconclusive in the case when we
assumed that the excess feed grain would have been exported directly. The
final ranking of this alternative and the actual price policy would have also
required information about reductions in production costs in the system asso-
ciated with adopting alternative 3 instead of the actual price policy.

Thus, these computations suggest that the adjustment policy alternatives
would have been more effective than the actual price policy in period 1971
74. Moreover, the direct payment scheme would have provided the govern-
ment with a vehicle to equalize the income distribution within the producers
in the system. This is also an important aspect in evaluating the policy alter-
natives, because the target price system as implemented thus far has
its disadvantages in this respect.

As stated in Section 6, selecting an appropriate policy or policy mix is a
complex issue involving many political and economic considerations. In this
study we have only compared two different policies for attaining the objectives
of the agricultural policy from the economic standpoint. When writing these
conclusions, we already know that the government has introduced policy meas-
ures different from those discussed above to curb the further growth of egg
production1 (See Section 1.1). The policy mix the government has adopted
may be characterized by stating that the domestic price of eggs is kept at a
relatively high level under this policy and pait of the support costs is paid by
means of revenues from the marketing fees mentioned in Section 1.1. This may
be an effective policy program for curbing production increases, but there are
also negative impacts associated with this policy mix. First of all, the domestic
price level for eggs is artificially maintained at a higher level than would be the
case, for example, under the policy directions suggested in this study. On the
other hand, we may argue that measuring the levels of the policy instruments
is in such a policy program as this problematic, if attaining policy goals accu-
rately is regarded as important.

1 We wish to emphasize that all the analyses concerning alternative policies has been con-
ducted before the introduction of these policy programs.



503

REFERENCES

Agarvala, R. & Ball, R. J. 1970. Econometric Model Building for the Egg Market in the
U. K. Farm Economist 11: 537 555.

Almon, S. 1965. The Distributed Lag between Capital Appropriations and Expenditures.
Econometrica 33; 178 196.

Anon. 1966. Komiteanmietintö 1966: B 12. Maataloustuotteiden hinnanmuodostuskomitean
osamietintö 2. VIII -|- 89 s.

» 1969. Komiteanmietintö 1969; B 40. Maatalouskomitean mietintö 111. Xj-27 p.
* 1970. Massager ’7O. User’s Manual. Bank of Canada. 1970. Mimeograph.
* 1971 a. Komiteanmietintö 1971: B 104. Maataloustuotteiden hintamarginaaleja selvit-

televän neuvottelukunnan osamietintö I. 1964 1971. VIII 4- 55 p.
» 1971 b. Suomen Asetuskokoelma 1971; 977, 978.

—i> 1974. Eri tuotantosuuntaa harjoittavien kirjanpitotilojen tuloksia. Tilivuosi 1974.
The Agr. Econ. Res. Inst. Research Reports 35: 1 32.

» 1975. Suomen Asetuskokoelma 1975; 302, 354.
» 1976. Suomen Asetuskokoelma 1976; 360.

Bange, G, H. & Bender, F. E. 1969. An Estimate of Monthly Demand for Shell and Breaker
Eggs. Univ. of Maryland Agr. Exp. Sta. Miscel. Pubi. 703: 1—45.

Barker, R. & Stanton, B. F. 1965. Estimation and Aggregation of Firm Supply Functions.
J. Farm Econ. 47:701-712.

Box, G. E. P. & Jenkins, G. M. 1970. Time Series Analysis. Forecasting and Control. 542 p.
San Fransisco.

Bröcker, R. 1973. Marktprognosen fur Eier und Shlachtgefliigel, Bonner Hefte fiir Markt-
forschung 7:1 127. Stuttgart.

Chayat, M., Forker, O. D. & Padberg, D. I. 1974. An Econometric Determination of the
Welfare Impact of Giving Bargaining Power to Farmers; A Case Study ot the Egg
Industry. Search Agriculture 4; 4: 1 44. Cornell University Agricultural Experiment
Station.

Chen, D., Courtney, R. & Schmitz, A. 1972. A Polynomial Lag Formulation of Milk Pro-
duction Response. Am. J. Agr. Econ. 54: 77 83.

Cohen, K. J. & Cyert, R. M. 1961. Computer Models in Dynamic Economics. The Quarterly
Journal of Economics 75: 112 127.

Dardis, R. 1967. The Welfare Cost of Grain Protection in the United Kingdom. J. Farm Econ.
49: 597-609.

Day, R. H. 1961. Recursive Programming and Supply Prediction. In Agricultural Supply
Functions ed. by Heady et al. p. 108 125. Ames.

» 1962. An Approach to Production Response. Agr. Econ. Research 14 (1962); 134 148.
Dixon, W. J. & Massey, F. J. 1957. Introduction to statistical Analysis. 488 p. York,

Edgerton, D. L. 1972. Some Properties of Two Stage Least Squares as Applied to Nonlinear
Models. Int. Econ. Review 13: 26 32.

Fisher, M. R. 1958. A Sector Model. The Poultry Industry of the U.S.A. Econometrica 26;

37-66.
Flygare, B. 1972. Kananmunakaupan kansainvälinen markkinakehitys. Siipikarja 1972:

251-255.
Frick, G. E. & Andrews, R. A. 1965. Aggregation Bias and Four Methods of Summing Farm

Supply Functions. J. Farm Econ. 47: 696 700.
Gerra, M. J. 1959 a. The Demand. Supply and Price Structure for Eggs. U. S. Dep. Agr. Techn.

Bull. 1204: 1-157.
» 1959b. An Econometric Model of Egg Industry. J. Farm Econ. 41: 284 301.

Gulbrandsen, O. & Lindbeck, A. 1973. The Economics of the Agricultural Sector. 264 p.
Stockholm.

Hartman, D. G. 1974. The Egg Cycle and the Ability of Recursive Models to Explain it. Am.
J. Agr. Econ. 56: 254-262.

Heady, E. O. 1961. Uses and Concepts in Supply Analysis. In Agricultural Supply Functions
ed. by Heady et al. p. 3 25. Ames.

i> & Tweeten, L. 1963. Resource Demand and Structure of the Agricultural Industry,
515 p. Ames.



504

Holt, C. C. et. al. 1967. Program Simulate 11. A User’s and Programmer’s Manual. University
of Wisconsin. Programming Series 2. Mimeograph.

Howrey, E. P. & Kelejian, H, H. 1971. Simulation versus Analytical Solutions. In Computer
Simulation Experiments with Models of Economic Systems (by Naylor), p. 299 319.
New York.

Ihamuotila, R. 1972 a. Productivity and Aggregate Production Functions in the Finnish
Agricultural Sector 1950 1969. Pubi. of the Agr. Econ. Res. Inst. 25: 1 87.

» 1972 b. Leipäviljan tarjonnasta ja tarjontaan vaikuttavista tekijöistä Suomessa vuo-
sina 1951 1970. (Summary: On Bread Grain Supply Functions in Finland in 1951
1969). Pubi. of the Agr. Econ. Res. Inst. 26: I—6o.

» 1972 c. Maatalouden hintajärjestelmistä ja niiden merkityksestä Suomessa. Käytännön
Maamies 1972, 4: 10—ll and 52.

Johnston, J. 1972. EconometricMethods. 437 p. New York.
Josling, T. 1969. A Formal Approach to Agricultural Policy. J. Agr. Econ. 20: 175 191.
Kaarlehto, P. 1958. Kananmunien kysynnän joustavuudesta kokonaismarkkinatilastojen

valossa. Suomen Osuustoimintalehti 50: 299 308.
—* 1961. Tulotason vaikutuksesta elintarvikemenoihin ja kulutusmääriin. (Summary:

Income Elasticity of Food Expenditure and Consumption). Journal of the Scientific
Agricultural Society of Finland 33: 17 31.

—» 1964. Maataloustuotteiden markkinointi. Maanvilj. Tietok. 3: 359 400. Porvoo.
Kelejian, H. H. 1971. Two-Stage Least Squares and Econometric Systems Linear in Para-

meters but Nonlinear in the Endogenous Variables. Journal of the American Statistical
Association 66: 373 374.

Kersten, L. 1973. Ein Demographisches Modell Zur Vorausberechnung der Hidmerbestände.
Agrarwirtschaft 1973: 302 305.

» 1975. Mengen und Preisschwankungen am westdeutschen Eier Markt. Agrarwirtschaft
1975: 89-97.

Kettunen, L. 1968. Demand and Supply for Pork and Beef in Finland. Pubi. the Agr. Econ.
Res. Inst. 11: 1 93.

» 1972. Hintalait maatalouden tulopolitiikan välineenä. Kansantaloudellinen aikakaus-
kirja 1972: 2; 135-151.

Klein, L. R. 1965. An Introduction to Econometrics. 280 p. Englewood Cliffs.
Kmenta, J. 1971. Elements of Econometrics. 655 p. New York.
Koponen, V. 1964. Kananmunien kysynnästä ja tarjonnasta. (Summary; On Demand and

Supply for Eggs). Pubi. of the Marketing Research Institute of Pellervo-Society 8: 1 29.
Kukkonen, P. 1968. Analysis of Seasonal and Other Short-term Variations with Applications

of Finnish Economic Time Series. Bank of Finland Inst, for Econ. Res. Pubi. B; 28:
1-136.

Labys, W. C. 1973. Dynamic Commodity Models: Specification, Estimation, and Simulation.
351 p. Lexington.

Larson, A. B. 1964. The Hog Cycle and Harmonic Motion. J. Farm Econ. 46: 375 386.
Leuthold, R. M. 1975. On the Use of Theil’s Inequality Coefficients. Am. J. Agr. Econ. 57:

344-346.
Miller, B. R. & Masters, G. C. 1973. A Short Run Price Prediction Model for Eggs. Am. J.

Agr. Econ. 55: 484—489.
Naylor, T. H. 1970. Policy Simulation Experiments with Macroeconomic Models: The State

of the Art. Am. J. of Agr. Econ. 52: 263 271.
* 1971. Computer Simulation Experiments with Models of Economic Systems. 502 p.

New York.
Nerlove, M. 1958. Dynamics of Supply: Estimation of Farmers’ Response to Price. 267 p.

» 1961. Time-series Analysis of the Supply of Agricultural Products. In Agricultural
Supply Functions ed. by Heady et al. p. 31—60. Ames.

Nevala, M. 1973. Kananmunien tuotanto, kulutus ja hinnanmuodostus. Mimeograph. 87 p.
(Available at the Agr. Econ. Res. Inst.).

i> 1974. Prognosteknik för eggproduktion. Nordisk statistisk sekretariat. Tekniskerapporter
13. Kobenhagen.



505

Niitamo, O. E. & Soivio, A. 1964. Kokeita yksinkertaisilla simultaanisilla moniyhtälömalleilla.
(Summary: Experiments with Simple Simultaneous Equation Systems). Kansantalou-
dellisia tutkimuksia 25: 134 164 and 211 212.

Pihkala, K. U. 1941. Suomen kananmunahintojen muodostumisesta maailmansodan jälkeen.
(Summary: On the Formation of Egg Prices in Finland after the World War). Acta
Agralia Fennica 46: 1 277.

Pölkki, L. 1971. Naudan- ja sianlihanhintojen ja marginaalien lyhytaikaiset vaihtelut Suomessa
vuosina 1963 70. (Summary: The Short-Run Changes in Prices and Marketing Margins
for Beef and Pork in Finland 1963 1970). Pubi. of the Agr. Econ. Res. Inst. 24:
1-144.

Ray, D. & Tweeten, L. 1976. Alternative Agriculture and Food Policy Directions for the U. S.
With Emphasis on Continuation of Minimal Provisions of the 1973 Agriculture and

Consumer Protection Act. Mimeograph. 30 p.
Roy, S. K. & Johnson, P. N. 1973. Econometric Models for Quarterly Shell Egg Prices. Am.

J. Agr. Econ. 55: 209-213.
Sahi, R. K. & Graddock, W. J. 1974. Estimation of the flexibility Coefficients for Recursive

Programming Models Alternative Approaches. Am. J. Agr. Econ. 56: 344 350.
Saito, T. 1972. Comet User’s Manual. Nippon Univac Sogo Kenkyusho Inc. Mimeograph.
Sauli, L. 1971. Maatalouden hintapolitiikasta ja hintajärjestelmistä Suomessa. Talonpojan

asialla, Veikko Ihamuotila kuusikymmentä vuotta 17, 12.1971. Published by Maatalous-
tuottajien Keskusliitto. Helsinki.

Schaller, N. W. & Dean, G. W. 1965. Predicting Regional Crop Production: An Application
of Recursive Programming. U. S. Dep. Agr. Techn. Bull. 1329: 1 110.

Sharples, J. A. 1969. The Representative Farm Approach to Estimation of Supply Response.
Am. J. Agr. Econ. 51: 353 361.

Sheehy, S. J. & McAlexander, R. H. 1965. Selection of Representative Benchmark Farms
for Supply Estimation. J. Farm. Econ. 47: 681—695.

Sisler, D, G. 1966. Direct Payments as a Policy Tool in the Feed Grain Livestock Sector.
Cornell University Agr. Ex. Sta. Bull. 1011: 1 35.

Stojkovic, G. 1971. Market Models for Swedish Agriculture. 83 p. Uppsala.
Teräsvirta, T. 1971. Tillämpning av stegvis regression analys. Föredrag vid den fjärde nordiska

konferensen i matematisk statistik. Mimeograph. 11 p.
Theil, H. 1958. Economic Forecasts and Policy. 567 p. Amsterdam.

» 1966. Applied Economic Forecasting. 474 p. Chicago.
» 1971. Principles of Econometrics. 736 p. New York.

Tomek, W. G. 1973, R 2 in TSLS and GLS Estimation. Am. J. Agr. Econ. 55: 670.
» & Robinson, K. L. 1975. Agricultural Product Prices. 376 p. Ithaca.

Torvela, M. 1959. Kotieläintalouden tuotoksen riippuvuus käytetyistä tuotantovälineistä.
Maata). Tai. Tutk.lait. Tied.ant. 1/1959: 1 64.

Tweeten, L. G. & Tyner, F. H. 1966. The Utility Concept of Net Social Cost A Criterion
for Public Policy. Agr. Econ. Research 18, 2: 33 42.

» & Quance, C. L. 1969. Positivistic Measures of Aggregate Supply Elasticities: Some
New Approaches. Am. J. Agr. Econ. 51:342 352.

Waananen, M. & Kaarlehto. P. 1965. Marketing Margins for Eggs in Finland. Washington
Agr. Exp. Sta. Techn. Bull. 45: 1 26.

Wallace, T. D. 1962. Measures ofSocial Costs of AgriculturalPrograms. J. Farm Econ. 44:580-594.
Westermarck. N. 1972. Munantuotannonkannattavuustilivuonna 1971.Siipikarja 1972; 160-165.

* 1973. Munantuotannon kannattavuus kirjanpitokanaloissa tilivuonna 1972. Siipikarja
1973; 228-232.

» 1975. Mitä osoittaa kanatalouden kannattavuustutkimus neljännesvuosisadan ajalta.
Erip. Siipikarja 1975; 3. 8 p. Hämeenlinna.

Winfridson, O. 1974. Prognoser för den svenska äggmarknaden. Nordisk Statistisk sekretariat.
Tekniske rapporter 13. Kobenhagen.

Wipf, L. J. & Bawden, D. L. 1969. Reliability of Supply Equations Derived from Production
Functions. Am. J. Agr. Econ. 51: 170 178.

Wold, H. 1964. Forecasting by the Chain Principle. In Econoraetic Model Building. Contri-
butions to the Economic Analysis 36 ed. by Wold; 5 36.



506

SELOSTUS

Suomen kananmunasektoria kuvaava ekonometrinen malli

Markku Nevala
Maatalouden taloudellinen tutkimuslaitos, Rukkila, 00001 Helsinki 100

Käyttökelpoisen tiedon tuottaminen kananmuniin kohdistuvan hintapoliittisen päätöksen-
teon perustaksi tavoitehintajärjestelmää sovellettaessa on ollut tämän tutkimuksen perimmäi-
nen tavoite. Tätä tarkoitusta varten on tutkimuksessa rakennettu kananmunien tuotannon, ku-
lutuksen sekä hintojen määräytymistä kuvaava ekonometrinen malli ja käytetty sitä hinta-
poliittisiin analyyseihin. Tällaisen ekonometrisen mallin avullahan voidaan simuloida ja ana-
lysoida eri hintapoliittisten vaihtoehtojen vaikutuksia kananmunasektorin tulevaan kehityk-
seen olettaen, että mallin eksogeenisten muuttujien kehitysurat voidaan tavalla tai toisella en-
nakoida ja että sektorin reaktiot myös tulevaisuudessa noudattavat niitä lainalaisuuksia, joita
rakennettuun ekonometriseen malliin sisältyy. On perusteltua väittää, että tällainen apuväline
olisi hyödyllinen hintapolitiikan päätöksentekijöille erityisesti kananmunien kohdalla, jossa
maatalouspoliittisten tavoitteiden toteuttaminen kohtuullisin kustannuksin on viime aikoina
monesta eri syystä osoittautunut vaikeaksi.

Tässä tutkimuksessa johdettu perusmalli, jonka spesifikaatiota on esitelty tutkimuksen toi-
sessa luvussa on ns. sektorimalli, jokakäsittää kaikkiaan kahdeksan yhtälöä. Se on luonteeltaan
block-rekursiivinen jakautuen yhteen kuuden yhtälön simultaaniseen lohkoon jakahteen yhden
yhtälön lohkoon. Mallin perusajatuksena on se, että valtiovallan asettama tavoitehinta yhdessä
kotimaisen markkinatilanteen kanssa määräävät tuottajahintatason kullakin ajanjaksolla.
Tuottajien on oletettu reagoivan kananmunien tuotannon kannattavuuden muutoksiin. Mallin
muuttujia valittaessa on lähdetty siitä, että muutokset kananmunien tuottajahinnoissa, rehujen
hinnoissa sekä tuotantotekniikassa ovat tärkeimmät tekijättuotannonkannattavuuden muutok-
siin. Tuotannon vaihteluita selittäviä yhtälöitä spesifioitaessa on oletettu, että kananmunien
tuottajat voivat muuttaa yritystensä tuotostasoa vaihtelemalla kananpoikasten hankintamää-
riä ja karsittavia kanojen määriä. Tämän perusteella malliin on kehitetty yhtälöt selittämään
näiden suureiden vaihteluita ja kyseisissä yhtälöissä selittävinä muuttujina toimivat muun
muassa em. tuotannon kannattavuuteen vaikuttavat tekijät. Kananmunatuotos on mallissa
johdettu kanakannan, kananpoikasten määrän ja karsinnan perusteella määritelmäyhtälön ja
teknisiä riippuvuuksia sisältävän yhtälön avulla. Olettamukset tuottajien reaktioiden viiväs-
tymisestä hintojen muutoksiin johtivat jakautuneita viiveitä sisältäviin malleihin kananpoikas-
ten hankintaa selitettäessä. Kananmunien vähittäishinta on mallissa oletettu määräytyvän
tuottajahintatason ja kaupan marginaaleihin vaikuttavien tekijöiden perusteella. Lisäksi mal-
liin kuuluu kananmunien ulkomaankaupan määräytymistä selittävä määritelmäyhtälö sekä
kulutuksen vaihteluita selittävä yhtälö, jossa selittävinä muuttujina ovat tulotaso ja kanan-
munien vähittäishinta. Malli on rakennettu puolivuosittaista havaintoaineistoa varten. Se on
dynaaminen pisimmän viiveen ollessa 5 havaintokautta (Kuvio 2.3).

Perusmalli on estimoitu ajanjaksolta 1956 —7O soveltaen 2 SLS-metodia simultaaniseen
lohkoon ja muihin yhtälöihin vastaavasti OLS-metodia (Luku 4). Jakautuneita viiveitä si-
sältävät yhtälöt estimoitiin polynomiaalisia viiveformulointeja käyttäen. Kananmunien tuo-
tantoa selittävien yhtälöiden kertoimien ajallista vaihtelua pyrittiin tutkimaan myös estimoi-
malla kertoimet havaintojaksojen suhteen askeltaen. Tätä tehtävää varten johdettiin rekursii-
vinen tuotannon vaihteluita selittävä malli tutkimuksen perusmallista estimointityön helpotta-
miseksi. Suurin osa selittävistä muuttujista sai tilastollisesti merkitsevät kertoimet. Tuotantoa
selittävien yhtälöiden kertoimissa ei havaittu mitään systemaattista vaihtelua ajan suhteen,
joskin eräiden kertoimien lukuarvot vaihtelivat verrattain laajalla alueella, kun estimointi-
periodia vaihdettiin (Taulukko 4.1).

Mallin kykyä jäljittää kananmunien tuotannon, kulutuksen ja hintojen todellista kehitystä
on tutkimuksessa testattu kahdella kyseiseen malliin soveltuvalla testillä; 1) historiallisella
simuloinnilla, jossa endogeenisten muuttujien kehitysurat estimointiperiodille tuotettiin mallin
avulla antamalla siihen vain eksogeenisten muuttujien todelliset arvot ja alkuarvot viiväste-
tyille endogeenisille muuttujille, sekä 2) ex-post ennustesimuloinnilla havaintojaksoille 1971/1



507

1974/1 vastaavaa menettelyä käyttäen saadaksemme käsityksen siitä, millä tavalla malli simu-
loi todellista kehitystä, kun eksogeeniset muuttujat saavat arvoja, jotka poikkeavat niiden ar-
voista estimointikaudella. Tuotettuja endogeenisten muuttujien kehitysuria on sitten verrattu
todellisuudessa tapahtuneeseen kehitykseen graafisin tarkasteluin (Kuviot 5.1 —5.3) sekä
TheiTin erisuuruuskertoimien (Uj ja U 2) avulla. Näissä vertailuissa voitiin todeta, että malli
jäljitti endogeenisten muuttujien tasot eri ajanjaksoilla verrattain hyvin, jota voidaan pitää
eräänä ilmauksena mallin soveltuvuudesta kuvaamaan kananmunasektorin reaktioita ekse-
geettisten muuttujien vaihteluihin ja siten myös apuvälineeksi hintapoliittisiin analyyseihin.

Edellä mainitulla tavalla estimoitua ja testattua perusmallia on sitten käytetty hintapoliittis-
ten vaihtoehtojen analysointiin. Ensinnäkin on mallin avulla pyritty arvioimaan minkälaista
hintapolitiikkaa olisi estimointikaudella (1956—70) täytynyt harjoittaa, jotta kotimaiset ka-
nanmunamarkkinat olisivat pysyneet tasapainossa. Käyttämällä iteratiivista lähestymistapaa
on tutkimuksen perusmallilla tuotettu sellaiset ratkaisut, jotka vastaavat suunnilleen kotimai-
sen tuotannon ja kulutuksen tasapainotavoitetta havaintokausittain kyseisenä ajanjaksona
(Kuviot 6.1 ja 6.2). Saatujen tulosten mukaan tasapainoon tähtäävä tuottajahintaura poikkeaa
vain suhteellisen vähän todellisesta tuottajahintaurasta,mikä suurimmaksi osaksi johtuu tuo-
tannon reaktioiden herkkyydestä hintojen muutoksiin. Näin saatua hintapoliittista vaihto-
ehtoa on sitten verrattu todelliseen hintapolitiikkaan vuosina 1956—7O sen selvittämiseksi,
olisiko kyseinen hintapolitiikka yhdistettynä suoriin maksuihin tuottajille ollut yhteiskunnan
kannalta tehokkaampi vaihtoehto hintapoliittisten tavoitteiden toteuttamiseen kuin todelli-
suudessa harjoitettu hintapolitiikka. Tässä vertailussa hintapolitiikan tavoitteiksi määritettiin
tuottajien tulotason turvaaminen ja kananmunien omavaraisuuden säilyttäminen. Kriteerinä
tehokkuusvertailussa käytettiin valtiovallan ja kuluttajien menoja kussakin vaihtoehdossa.
Tässä analyysissa niiden todellisuudessa kananmunien tuotantoon sijoitettujen resurssien
osalta, joita ei tuotannon ja kulutuksen tasapainoon tähtäävän politiikan vallitessa olisi käytetty
kananmunien tuotantoon, on vaihtoehtoiseksi käytöksi oletettu sianlihan tuotanto. Laskelmien
tulokset osoittivat, että edellä määritelty hintapoliittinen vaihtoehto olisi ollut yhteiskunnan
kannalta tehokkaampi hintapoliittisten tavoitteiden saavuttamiseksi kuin todellinen hintapoli-
tiikka (Taulukko 6.1). Toisaalta tässä kohdin on huomautettava, että analyysin tulokset riippu-
vat tehdyistä olettamuksista ja hintatilanteesta maailmanmarkkinoilla.

Niinikään on tutkimuksessa suoritettu ex-post analyyseja erilaisista hintapoliittisista vaihto-
ehdoista lähinnä sen osoittamiseksi, minkälaisiin analyyseihin malli soveltuu todellisessa hinta-
päätöstilanteessa. Tätä tarkoitusta varten määriteltiin kolme erilaista vaihtoehtoa kotimai-
sen tuotannon ja kulutuksen sopeuttamiseksi paremmin toisiaan vastaavaksi lähtien tilanteesta
vuoden 1970 lopussa. Sopeuttamisajanjaksoiksi määriteltiin 5—6, 7 8 ja 10—ll vuotta ja
mallin avulla tuotettiin näitä vaihtoehtoja vastaavat ratkaisut vuosille 1971 —74 em. iteratii-
vista menettelyä käyttäen ja saatiin selville eri vaihtoehtoja vastaavat tuottajahintaurat.
Edellä esitetyn kaltaiset tehokkuusvertailut näiden vaihtoehtojen ja todellisen hintapolitiikan
välillä suoritettiin myös tässä analyysissa. Ylituotantoresurssien vaihtoehtoisesta käytöstä teh-
tiin kaksi erilaista olettamusta: sopeuttamisvaihtoehdoissa nämä resurssit oletettiin käytetyn
1) sianlihan tuotantoon ja 2) rehuvilja olisi suoraan viety ulkomaille jalostamatta sitä kotieläin-
tuotteiksi. Laskelmien tulokset viittaavat siihen, että nopea tuotannon ja kulutuksen sopeut-
taminen hintapoliittisin keinoin ja suorien maksujen käyttö tulotavoitteen toteuttamiseen olisi
ollut yhteiskunnan kannalta edullisempaa kuin harjoitettu hintapolitiikka kyseisenä ajanjak-
sona niiden olettamusten vallitessa, joita laskelmissa tehtiin (Taulukot 7.1 ja 7.2).



508

APPENDICES

Appendix 1. List of the Variables Used in the Analyses of This Study and Sources for the Data

The variables used in the analyses of this study are listed below. Also, the sources of the data
are shortly indicated for each variable. The data are predented in Appendix 2.

SP t = number of chickens at the end of period t, milj. pcs. Sources: for 1960 74, the Monthly
Bulletin of Agricultural Statistics published by Board of Agriculture. Since the pub-
lished numbers beginning from 1969 do not include chickens on farms under one
hectare of arable land, we transformed the official figures for 1969 —74 by the esti-
mated numbers of chickens on those farms (For more detail, see Nevala 1973, p. 44).
The data for 1952 —59 were taken directly from the study by Koponen (1964).

SK t =number of layers at the end of period t, mil. pcs. Data sources and the manipulation
of the data are the same as for variable SP.

KK, = number of culled layers in period t, mil. pcs. Since no official data were available for
this variable, we computed its values according to the following formula; KKt =

SK t_j + SP t_!
- SK t.

TS t = egg production in period t, mil. kg. Source; Board of Agriculture, Statistical Office.
Although based on the same sources, the total production figures of this study deviate
a little from those published by Board of Agriculture due to different methods in deriv-
ing the total egg production from the statistical sample in a given period (See Nevala
1973, p. 45).

CO t = total per capita consumption of eggs in period t, kg. The quantities of eggs consumed
have been derived by subtracting the net exports from the total production of eggs.
Thus, the consumption figures refer to the total domestic disappearance of eggs (i.e.
the figures include also nonfood utilisation, such as eggs used for hatchings).

UK t =net exports of eggs in period t, mil. kg. Source; the Statistical Office, Board of Cus-
toms.

PKT t producer price of eggs in period t, Fmk/kg. Source: the Agricultural Economics Re-
search Institute. For 1952—55 the data were obtained from the study by Koponen

(1964).

PKVt =retail price of eggs in period t, Fmk/kg. Source: the Bulletin of Statistics published
by the Central Statistical Office of Finland.

TAVt = target price of eggs fixed by the government, Fmk/kg. Since the target price has been
changed in different time points of the year, the semiannual values for this variable
have been computed as the simple averages of the monthly target price levels (= the
price which has been in force in a given month). Sources: Kettunen (1972) and the
Agricultural Economics Research Institute.

PR t = price of commercial feed, Fmk/kg. The wholesale price of one of the most common
commercial feed mix served as a proxy for farm prices of the commercial feed used
in the egg industry. Sources: the data for 1960 —74 were provided by the manu-
facturer of this feedmix, and the time series were extended to the previous observation
periods by assuming that the semiannual changes in this time series would have been
the same as in the corresponding time series developed by Koponen (1964).

RVS t = total availability of domestically produced feed grain (oats and barley). Total pro-
duction of oats and barley was used as a proxy for this variable. Source: the Monthly
Bulletin of Agricultural Statistics published by the Board of Agriculture.

Y t = consumers’ disposable incomes. The wage index of all the salary and wage earners
served as a proxy for this variable. Source: the Bulletin of Statistics published by
the Central Statistical Office of Finland.

YM t = wage index in commerce. Source: the Bulletin of Statistics published by the Central
Statistical Office of Finland.



509

EI t = consumer price index in period t. Source: the Bulletin of Statistics published by the
the Central Statistical Office of Finland.

H t = total population in Finland. Source: the Bulletin of Statistics, the Central Statistical
Office of Finland. Some minor adjustments were made to the published figures (See
Nevala 1973, p. 47).

T = trend. T = 9 for 1952/1 .

D 2 = seasonal dummy, D 2 = 1 for the second half of the year, otherwise D 2 =O.
ESK t = export subsidy paid by the government for eggs in period t, Fmk/kg. Source: the

Agricultural Economics Research Institute.

VHKt = export price of eggs in period t, Fmk/kg. Source: the Agricultural Economics Research
Institute.

PST t =producer price of pork in period t. Fmk/kg. Source: the Agricultural Economics Re-
search Institute.

ESS t = export subsidy paid by the government for pork in period t. Fmk/kg. Source; the Agri-
cultural Economics Research Institute.

VHSt = export price of pork in period t, Fmk/kg. Source: the Agricultural Economics Research
Institute.



510

Appendix 2. Observed Values of the Variables Used in the Analysis.

SP KK SK TS CO UK

1952 I 1.350.01 3.4615.25 3.76 -0.04
II 0.081.23 3.5815.77 3.86 -0.04

1953 I 1.730.41 3.2515.15 3.680.00
II 0.141.16 3.9217.09 4.120.00

1954 I 1.670.55 3.5217.30 4.140.08
II 0.071.01 4.1718.74 4.420.13

1955 I 1.720.70 3.5418.64 4.161.03
II 0.111.34 3.9218.35 4.190.60

1956 I 1.840.65 3.3917.90 3.861.40
II 0.171.14 4.0917.60 3.880.90

1957 I 2.060.82 3.4417.50 3.631.80
II 0,10 1.104.39 19.803.94 2.70

1958 I 2.020.87 3.6220.60 3.913.50
II 0.111.21 4.4321.40 4.163.20

1959 I 2.141.20 3.3420.70 3.863.70
II 0.091.35 4.1319.70 3.633.70

1960 I 2.290.76 3.4619.40 3.603.40
II 0.141.57 4.1821.19 3.844.20

1961 I 3.090.95 3.3720.54 3.714.00
II 0.322.15 4.3121.42 3,71 4.80

1962 I 2.830.95 3.6822.01 3.904.50
II 0,29 1.944.57 22.294.18 3.50

1963 I 2.970.99 3.8722.68 3.915.00
II 0.372.09 4.7623.94 4.284.50

1964 I 2.721.27 3.8724.41 4.673.10
II 0.191.79 4.7925.53 4.495.00

1965 I 2.790.90 4.0925.14 4.355.20
II 0.181.93 4.9526.36 4.724.70

1966 I 2.790.96 4.1726.42 4.555.50
II 0.211.85 5.1127.34 4.417.00

1967 I 2.790.82 4.5027.91 4.487.20
II 0.192.04 5.2528.55 4.318.60

1968 I 2.601.10 4.3427.61 4.277.80
II 0.241.77 5.1727.13 4.655.50

1969 I 3.55 1,33 4.0826.03 4.624.50
II 0.342.28 5.4028.12 4.746.10

1970 I 3.991.03 4.7129.27 4.578.10
II 0.201.67 6.8333.37 5.249.20

1971 I 3.211.50 5.5336.71 5.0213.60
II 0.411.99 6.5834.85 5.2010.90

1972 I 3.621.28 5.7134.52 4.9411.70
II 0.572.64 6.6937.60 5.4612.30

1973 I 3.931.38 5.8838.88 5.2314.62
II 0.562.83 6.9837.58 5.2113.16

1974 I 3.430.93 6.1137.51 4.9914.20
II 0.542.93 6.6139.24 5.7912.10



511

Appendix 2. (Continued). Observed Values of the Variables Used in the Analysis.

PKT PKV RVS PR Y YM

1952 I 2.032.36 1.020.26 -

II 2.252.80 1.060.29 - -

1953 I 1.892.39 1.060.29 - -

II 1.952.49 1.270.28 - -

1954 I 1.592.02 1.27 0,28 -
-

II 1.862.35 1.110.27 - -

1955 I 1.722.10 1.110.28 - -

II 2.022.49 0.960.28
1956 I 1.792.29 0.960.28 0.560.57

II 2.192.69 1.020.31 0.600.60
1957 I 1.912.40 1.020.33 0,61 0.60

II 2.262.77 1.010.36 0.610.61
1958 I 2.052.56 1.010.40 0.630.64

II 2.322.85 1.230.40 0.650.65
1959 I 2.112.64 1.230.41 0.660.67

II 2.442.96 1,04 0.420.68 0.67
1960 I 2.362.88 1.040.43 0.700.70

II 2.653.18 1.570.41 0.720.71
1961 I 2.453.00 1.570.41 0.750.74

II 2.332.89 1.340.42 0.770.75
1962 I 2.142.70 1.340.42 0.800.78

II 2.362.92 1.080.42 0.820.80
1963 I 2.493.08 1.080.44 0.86 0,87

II 2.393.04 1.360.45 0.910.88
1964 I 2.483.20 1.360.45 0.990.99

II 2.583.41 1.180.48 1.021.01
1965 I 2.703.44 1.180.50 0.081.07

II 2.643.47 1.630.50 1.091.07
1966 I 2.763.52 1.630.51 1.141.13

II 2.813.61 1.530.53 1.191.15
1967 I 2.903.72 1.530.55 1.251.21

II 2.853.78 1.760.61 1.291.24
1968 I 2.853.69 1.760.61 1.381.33

II 3.053.87 1.910.62 1.441.37
1969 I 3.204.02 1.910.63 1.501.44

II 3.084.07 1.920.63 1.531.44
1970 I 3.153.99 1.920.63 1.621.53

II 3.174.04 2.210.65 1.661.55
1971 I 3.163.93 2.210.66 1.811.69

II 3.324.16 2.450.67 1.901.76
1972 I 3.474.32 2.450.68 2.011.87

II 3.52 4,49 2.350.69 2.131.99
1973 I 3.634.61 2.350.77 2.292.13

II 3.875.00 2.160.81 2.502.34
1974 I 4.095.30 2.160.85 2.712.55

II 4.435.71 2.070.99 2.992.81



Appendix 2 (Continued). Observed Values of the Variables Used in the Analysis.

EI TAV H ESK VHK PST ESS VHS

1952 I ________

1953 I ________

1954 I ________

1955 I ________

1956 I 1.081.95 4.281.00 1.252.18
II 1.112.02 4.301.26 1.262.27 -

1957 I 1.212.05 4.321.22 1.072.48 -

II 1.262.18 4.341.00 2.132.15 1.121.49
1958 I 1.312.25 4.370.98 1.622.06 1.331.26

II 1.322.35 4.381.18 1.762.49 1.401.30
1959 I 1.332.40 4.401.30 1.342.30 1.211.60

II 1.352.47 4.411.52 1.672.56 1.281.77
1960 I 1.372.50 4.441.56 1.342.85 - -

II 1.392.57 4.431.53 1.872.87
1961 I 1.402.60 4.461.56 1.642.62 1.151.65

II 1.412.56 4.481.55 1.412.67 1.201.63
1962 I 1.452.55 4.491.55 1.202.72 1.251.59

II 1.492.48 4.501.39 1.562.68 1.251.59
1963 I 1.522.53 4.521.24 1.822.81 1.461.66

II 1.562.59 4.541.72 1.352.99 1.591.55
1964 I 1.682.60 4.562.08 1.132.99 1.411.91

II 1.722.67 4.572.02 1.333.02 1.771.74
1965 I 1.762.77 4.582.02 1.333.31 1.981.66

II 1.802.80 4.591.79 1.643.35 2.161.49
1966 I 1.832.80 4.601.96 1.493.40 1.711.70

II 1.872.93 4.612.08 1.513.32 1.741.84
1967 I 1.933.00 4.622.24 1.393.24 2.201.73

II 1.983.00 4.632.04 1.623.28 2.271.83
1968 I 2.103.00 4.641.91 1.713.44 2.531.74

II 2,14 3.154.65 1.851.97 3.852.64 1.81
1969 I 2.163.20 4.662.15 1.733.84 2.602.05

II 2.183.20 4.652.02 1.884.04 2.631.97
1970 I 2.213.27 4.642.33 1.724.00 1.822.76

II 2.243.35 4.612.42 1.694.01 1.732.85
1971 I 2.323.35 4.602.69 1.454.18 2.722.22

II 2.433.35 4.612.47 1.824.31 2.912.34
1972 I 2.493.42 4.622.72 1.684.37 2.832.33

II 2.593.50 4.632.79 1.714.69 3.112.22
1973 I 2.723.62 4.643.07 1.634.67 1.784.13

II 2.963.85 4.663.02 2.065.11 1.824.14
1974 I 3.204.05 4.673.02 2.355.34 2.184.31

II 3,47 4.404.68 3.542.30 5.984.25 4.23

512



513

Appendix 3. Estimation Method for the Finite Distributed Lag Models Used in This Study
(See ANON. 1970).

The distributed lag models of this study can be written as a relationship of the form (excluding
all the other explanatoryvariables):

m 1
(1) Yt -Z fii X t_i

i =0

In order to reduce the number of parameters to be estimated, the coefficients fa are cons-
trained to lie along a polynomial of specified degree that passes through zero at i = m. Thus,
we write:

o 2,. f 1"- 1) 3
Pi “ 2 bj ——■>

j-1 Sj
(2)

where q is the degree of a polynomial and Sj is a scaling factor defined by

m
Sj = X (m-i)j

i=o
(3)

Substituting (2) in (1), we obtain

m f <1 (m—i)j 1Y t = E 2 1 bj Xt_i; or
i-o L j=i si J

(4)

= Z bj (X)Dgj
j=i

(5)

where
m (m —i)j

(X)Dg, = E Xt_i
i=o sj

(6)

Coefficients bj for transformed variables (X)Dgj (j = 1 . . . q) are estimated by OLS. The
coefficients for individual lags fa are then computed according to formula (2).



Appendix
4.

OLS-estimates
for
the

Recursive
Production
Response

Model
Estimated

from
Different
Time

Periods.
1

Equation
(4.1);
SP
t

(

=

number
of

chickens)
as
the

dependent
variable Coefficients

for

Period

Constant
PKT
t

PKTt_1PKT
t_2PR
t

PR
t
_i

PR
t_2

T

D
2

T

D
a

R2d 1956/1
-68/1

1.086
0.435
0.290
0.145
-2.960
-1.973
0.987
0.067
-0.031
-1.487
0.980

1.37

1.6

1.6

1.6

-2.7
-2.7
-2.7
3.1

-2.6
-4.3

1958/1
-70/1
0.240
0.614
0.410
0.205
-2.460
-1.640
-0.820
0.054
-0.040
-1.299
0.971

1.37

1.7

1.7

1.7

-1.7
-1.7
-1.7
2.1

-2.6
-2.5

1960/1
-72/1
0.552
0.581
0.387
0.194
-2.258
-1.505
-0.753
0.045
-0.032
-1.564
0.973

1.69

1.6

1.6

1.6

-1.2
-1.2
-1.2
1.3

-2.1
-2.7

1962/1
-74/1
0.707
0.440
0.294
0.147
-2.068

-1.378
-0.689
0.055
-0.034

-1.464
0.975

1.59

1.0

1.0

1.0

-1.1
-1.1
-1.1

1.2

-2.3
-2.4

1956/1
-70/11
0.164
0.568
0.379
0.189
-2.223
-1.482
-0.741
0.056
-0.044
-1.164
0.974

1.69

1.8

1.8

1.8

-1.8
-1.8
-1.8

2.4

-4.2
-3.4

1957/11
—7O/11
0.297
0.589
0.393
0.196

-2.442
-1.628
-0.814
0.056
-0.040

-1.277
0.973

1,71

1.7

1.7

1.7

-1.8
-1.8
-1.8

2.2

-3.2
-2.9

1

Numbers
below
the

coefficients
are
Student
t-values.

Coefficients
for

PKT
t-i

and
PR
t

_j(i
=

0

...

2)
were

estimated
through
the

use
of
a
po-

lynomial
of
first
order
but
they
are

presented
above
in
an

unscrambled
form.

514



515

Appendix
4.

(Continued).
OLS-estimates
for
the

Recursive
Production
Response

Model
Estimated
from

Different
Time

Periods.
1

Equation
(4.2):
KK
t

(=

number
of
culled
layers)
as

the
dependent

variable Coefficients
for

Period
Constant
SP
t

_
2

SP
t_3SP
t_4SP
t

_s

PKf
t

PR
t

D2R 2d 1956/1
-68/1
0.992
-0.085
0.210

0.323

0.253

-0.771
2.594

0.308

0.858

1.67

-0.5

3.0

3.0

2.7

-2.3

2.0

0.8

1958/1
-70/1
0.811

—O.OBO
0.129

0.213

0.170
—0.544
2.483

0.435

0.838

1.79

-0.5

1.7

1.9

1.8

-1.5

1.8

1.1

1960/1
-72/1
0.758

0.097
0.076

0.053

0.028
—0.384

1.724

0.980

0.862

2.07

0.7

1.1

0.6

0.4

-1.0

1.2

2.8

1962/1
-74/1
0.478
-0.054
0.090

0.148

0.118
-0.224

1.643

0.636

0.813

2.23

-0.3

0.8

1.0

0.9

-0.3

0.6

1.2

1956/1
-70/11
0.825
-0.114
0.154

0.262

0.211

-0.605
2.669

0.270

0.825

1.59

-0.8

2.2

2.6

2.4

-1.7

2.0

0.8

1957/11
—7O/11
0.958

-0.144
0.141

0.256

0.213

-0.582
2.470

0.212

0.804

1.56

-0.8

1.8

2.8

2.2

-1.5

1.7

0.5

1

Numbers
below
the

coefficients
are

Student
t-values.

Coefficients
for

j

(i

=

0...3)
wereestimated

through
the

use
of
a

polynomial

of
second
order
but
they
are

presented
above
in
an

unscrambled
form.



516

Appendix 4. (Continued). OLS-estimates for the Recursive Production Response Model Estimat-
ed from Different Time Periods.

Equation (4.3): TS t ( total egg production) as the dependent variable

Coefficients for

Period Constant SKt—l SPt _i KK, T D R 3 d

1956/1 -68/1 -3.737 4.692 1.939 -1.775 0.217 0.914 0.979 1.28
5.7 1.8 -1.7 4.5 0.6

1958/1 -70/1 -5.035 6.093 3.590 -3.303 0.087 -0.668 0.977 1.12
7.7 4.0 -3.6 1.9 0.5

1960/1 -72/1 -2.807 5.128 1.968 -1.973 0.128 1.466 0.982 1.21
10.6 3.7 -2.6 2.6 1.0

1962/1 -74/1 -3.118 5.160 2.056 -1.138 0.111 0.417 0.976 1.39
8.1 2.9 -1.3 1.5 0.2

1956/1 —7O/11 -3.555 4.907 2.138 -1.674 0.169 0.567 0,978 0.98
6.1 4.8 -2.7 3.5 0.5

1957/11—7O/11 -3.940 5.301 2.227 -1.843 0.131 0.797 0.980 1.06
7.5 5.4 -3.4 3.1 0.8

1 Numbers below the coefficients are Student t-values.

Appendix 5. Linear Correlation Matrix of the Equation Residuals of the Basic Model Estimated
from Period 1956 70.

Equation (4.11) (4.5) (4.6) (4.7) (4.8) (4.9)
Dependent SP KK TS PKT PKV CO

Variable u ( u 2 u 4 u5 u 0 u 7

u, 1.00 0.36 -0.47 -0.20 -0.34 -0.25
u 2 1.00 -0.36 -0.39 -0.18 -0.30
u 4 1.00 -0.12 0.24 0.25
u, 1.00 0.64 0.10
u„ 1.00 0.10
u j 1.00

For n = 30 (n =number of observations), r has to be 0.46 at the 1 percent level of sig-
nificance and 0.36 at the 5 percent level of significance in order to be statistically different
from zero (See Dixon and Massey 1957, p. 468).



517

Appendix
6.

Ex-post
Forecast
for
the

Endogenous
Variables
and

Forecast
Errors
in
1971
1974/1

Generated
by
the

Basic
Model
(As
to

units
of

measurement;
see

Appendix
I).
1

SP
t

KK
t

SK
t

TS
t

PKT
t

PKVt

CO
t

UK
t

Period
Fore-
Error
Fore-
Error
Fore-
Error

Fore-
Error
Fore-
Error

Fore-
Error

Fore-
Error
Fore-
Error

cast

cast

cast

cast

cast

cast

cast

cast

1970
I

3.99

1.03

5.80

29.27

3.15

3.99

4.57

8.10

II

0.20

1.67

5.78

33.37

3.17

4.04

5.24

9.20

1971
I

3.39
-0.18
1.24

0.26
5.79
-0.26

35.79
0.92
2.98
0.18

4.04
-0.11
4.93
0.09

13.09
0.51

II

0.11
0.30
2.35
-0.35

6.84
-0.25
36.37
-1.52
3.08
0.24

4.21
-0.05
5.17
0.03

12.54
-1.64

1972
I

3.41
0.21

1.42
-0.14
5.52
0.19

35.55
-1.03
3.11
0.36

4.26
0.06
5.04
-0.10

12.25
-0.55

II

0.20
0.37
2.30
0.34

6.63
—0.05
35.65
1.95

3.32
0.20

4.56
—0.07
5.24
0.22

11.38
0.92

1973
I

3.54
0.39

1.29
0.09
5.55

0,33
35.53
3.35
3.37
0.26

4.65
-0.04
5.14
0.09

11.69
2.93

II

0.18
0.38
2.20
0.63

6.89
0.09
36.70
0.88
3.71

0.16

5.12
—0.12
5.34
—0.13

11.82
1.34

1974
I

3.70
—0.27
1.23

—0.30
5.84
0.27

37.19
0.32
3.84
0.25

5.31
—O.Ol
5.21
—0.22

12.83
1.37

1

Figures
for
the

year
1970

areactual
values,
which
are

presented
for
comparison

purposes.



518

Appendix 7. Simulated Time Paths for the Key Endogenous Variables under the Actual and
Hypothesized Equilibrium Price Policies. Variable Values Are Generated by the Basic Model
(As to the measurement units; see Appendix 1).

Total
Production (TS t ) Consumption (CO • H) t Net exports (CK t)

actual equilibrium actual equilibrium actual equilibrium
policy policy policy policy policy policy

1965 I 18.06 18.14 15.96 15.77 2.10 2.37
II 19.19 19.44 16.68 16.60 2.51 2.84

1957 I 18.51 18.99 16.60 16.42 1.91 2.57
II 18.06 18.74 16.67 16.78 1.39 1.96

1958 I 17.01 17.70 16.19 16,34 0.82 1.36
II 17.21 17.64 16.50 16,82 0.71 0.83

1959 I 16.95 16.86 15.97 16.28 0.98 0.58
II 17.94 17.16 16.58 16.95 1.36 0.21

1960 I 18.38 16.88 16.35 16.67 2.03 0.21
II 19.92 17.88 16.97 17.41 2.95 0.27

1961 I 20.88 17.90 16.83 17,24 4.05 0.66
II 22.66 19.03 17.93 18.10 4.73 0.93

1962 I 23.37 19.30 17.90 17.97 5.47 1.33
II 24.47 20.32 19.02 18.81 5.45 1.51

1963 I 24.20 20.33 18.59 18.53 5.61 1.80
II 24.47 20.96 19.43 19.58 5.04 1.38

1964 I 24.09 20.75 19.45 19.55 4.64 1.20
II 24.85 21.39 20.15 20.33 4.70 1.06

1965 I 25.05 21.18 19.68 19.98 5.37 1.20
II 26.26 21.83 20.49 20.66 5.77 1.17

1966 I 26.65 21.75 20.23 20.20 6.42 1.55
II 27.80 22.61 20.95 21.14 6.85 1.47

1967 I 28.11 22.72 20.56 20.77 7.55 1.95
II 29.03 23.44 21.56 21.61 7.47 1.83

1968 I 28.69 22.94 21.40 21.31 7.29 1.63
II 28.86 23.05 22.01 22.16 6.85 0.89

1969 I 28.34 22.46 21.44 21.61 6.90 0.85
II 29.16 23.14 22,31 22.37 6.85 0.77

1970 I 29.77 23.48 21.82 21.97 7.95 1.51
II 31.55 24.91 22.56 22.76 8.99 2.15



Appendix 7. (Continued). Simulated Time Paths for the Key Endogenous Variables under the
Actual and Hypothesized Equilibrium Price Policies. Variable Values Are Generated by the
Basic Model (As to the measurement units; see Appendix 1).

Producer price (PKT t) Retail price (PKVt )

Actual Equilibrium Actual Equilibrium
policy policy policy policy

1956 I 1.80 1.84 2.31 2.35
II 1.93 1.95 2.48 2.51

1957 I 1.93 1.98 2.44 2.48
II 2.18 2.15 2.71 2.69

1958 I 2.23 2.19 2.72 2.69
II 2.42 2.33 2.94 2.87

1959 I 2.40 2.31 2.88 2.81
II 2.53 2.43 3.05 2.96

1960 I 2.46 2.37 2.96 2.88
II 2.56 2.43 3.10 2.99

1961 I 2.47 2.36 2.99 2.89
II 2.46 2.41 3.04 3.00

1962 I 2.34 2.32 2.91 2.89
II 2.33 2.39 2.96 3.01

1963 I 2.31 2.33 2.94 2.96
II 2.48 2.44 3.14 3.10

1964 I 2.45 2.41 3.13 3.10
II 2.60 2.54 3.32 3.27

1965 I 2.61 2.51 3.32 3.23
II 2.70 2.64 3.45 3.40

1966 I 2.60 2.61 3.35 3.36
II 2.80 2.73 3.58 3.52

1967 I 2.78 2.70 3.55 3.48
II 2.85 2.83 3.69 3.67

1968 I 2.79 2.83 3.64 3.68
II 3.06 3.00 3.95 3.89

1969 I 3.05 2.98 3.94 3.87
II 3.12 3.10 4.05 4.03

1970 I 3.09 3.03 4.03 3.97
II 3.19 3.10 4.17 4.10

519



520

Appendix 8. Simulated Time Paths of the Most Important Endogenous Variables for 1971 —74
under the Actual Price Policy and Different Price Policy Alternatives as Specified in Chapter
7. Variable Values Are Generated by Ex-Post Simulations with the Basic Model.

Actual Alternative
policy 12 3

Production (TS t), mil. kg

1971 I 35.79 35.63 35.69 35.75
II 36.37 35.67 35.98 36.25

1972 I 35.55 34.00 34.76 35.40
II 35.65 33.13 34.44 35.51

1973 I 35.53 32.13 33.90 35.27
II 36.70 32.40 34.38 35.85

1974 I 37.19 31.57 33.61 35.14
II 38.56 31.04 33.18 35.01

Net Exports (UKt), mil. kg
1971 I 13.09 12.49 12.73 12.94

II 12.54 11.58 12.08 12.50
1972 I 12.25 10.45 11.37 12.13

II 11.38 8.61 10.03 11.14
1973 I 11.69 7.95 9.74 11.11

II 11.82 6.91 8.92 10.43

1974 I 12.83 6.56 8.71 10.48
II 13.23 5.00 7.37 9.41

Total Consumption (CO • H) t , mil. kg

1971 I 22.70 23.14 22.96 22.81
II 23.83 24.09 23.90 23.75

1972 I 23.30 23.55 23.39 23.27
II 24.27 24.52 24.41 24.37

1973 I 23.84 24.18 24.19 24.16
II 24.88 25.49 25.46 25.42

1974 I 24.36 25.01 24.90 24.66
II 25.33 26.04 25.81 25.60



521

Appendix 8. (Continued). Simulated Time Paths of the Most Important EndogenousVariables
for 1971 74 under the Actual Price Policy and Different Price Policy Alternatives as Specified
in Chapter 7. Variable Values Are Generated by Ex-Post Simulations with the Basic Model.

Actual Alternative
policy 12 3

Producer price (PKT t), Fmk/kg
1971 I 2.98 2.77 2.86 2.93

II 3.08 2.95 3.04 3.12
1972 I 3.11 2.99 3.07 3.13

II 3.32 3.19 3.24 3.26

1973 I 3.37 3,19 3.20 3.20
II 3.71 3.35 3.37 3.40

1974 I 3.84 3.42 3.49 3.65
II 4.31 3.82 3.97 4.12

Retail price (PKV t), Fmk/kg

1971 I 4.04 3.86 3.93 3.99
II 4.21 4.10 4.18 4.24

1972 I 4.26 4.15 4.22 4.27
II 4.56 4.45 4.50 4.52

1973 I 4.65 4.49 4.50 4.50
II 5.12 4.82 4.83 4.85

1974 I 5.31 4.95 5.01 5.15
II 5.93 5.52 5.65 5.77

Appendix 9. Producer Price, Export Subsidies and Export Price for Feed Grain used in Table
7.2, Fmk/kg.

Producer Export Export
Price Subsidy Price

Period Fmk/kg Fmk/kg Fmk/kg

1971 I 0.41 0.21 0.23
II 0.39 0.24 0.23

1972 I 0.40 0.25 0.22
II 0.39 0.25 0.22

1973 I 0.40 0.12 0.37
II 0.41 0.12 0.37

1974 I 0.48 - 0.52
II 0.48 - 0.54

Source: The Agricultural Economics Research Institute. The producer price of feed grain
refers to the average price of feed barley and oats.