A G R I C U LT U R A L A N D F O O D S C I E N C E Agricultural and Food Science (2022) 31: 24–36 24 https://doi.org/10.23986/afsci.112241 The role of subsidies in stabilising farm income: Evidence from Estonia Olha Aleksandrova1, Tetiana Zhmykhova2 and Ants-Hannes Viira1 1Estonian University of Life Sciences, Institute of Agricultural and Environmental Sciences, Fr. R. Kreutzwaldi 1, Tartu 51014, Estonia 2Donbas National Academy of Building Engineering and Architecture, Department of Higher and Applied Mathematics, Derzhavin str., 2, Makeyevka, Ukraine email: olha.aleksandrova@emu.ee This study examines the role of agricultural subsidies in the farm income (FI) variation in Estonia in the period 2006– 2019 using data from a balanced panel of Farm Accountancy Data Network (FADN) farms. FI variability is decomposed to three components: market income, total subsidies, and the cost of external factors. The results reveal that subsi- dies have a larger share in income of smaller farms. Variability of FI differs in farm size quartiles and is significantly lower in case of largest 25% of farms. Revenues are the largest source of variation of FI. While subsidies reduce FI variation, there is no evidence that they are counter-cyclical. We argue that in addition to subsidies, smaller farms need advice on farm management to increase market income, improve its stability, and expand farm size, as larger farms have more stable incomes despite relatively lower producer support estimate (PSE) levels. Key words: farm income volatility, farm management, agricultural subsidies, variance decomposition, Common Agricultural Policy Introduction Due to the changes in the global markets and attempts to liberalise trade, surplus production of agricultural products and associated budget constraints, enlargement towards Central and Eastern European Countries, and increasing public concerns about the environment, food safety and animal welfare, the European Union (EU) Com- mon Agricultural Policy (CAP) has seen several rounds of reforms and resulting implications on EU agriculture and food security (Moehlher 2008, Pe’er et al. 2019, Balezentis et al. 2020, Rac et al. 2020). Apart from the food secu- rity issues, the CAP and other EU policies are also expected to increasingly address environmental sustainability (Pe’er et al. 2019) and vitality of rural areas, though not always with clear results (Bakucs et al. 2018, Lillemets et al. 2022). The CAP for the period after year 2022 is expected to significantly contribute to achieving the goals of the EU’ arm to Fork Strategy, that ultimately aims to transform the EU food system and direct it to sustainable path (Shebesta and Candel 2020). Therefore, farmers involved in the production of agricultural products – one of the key commodities in global markets – are being caught in a difficult duel: they must confront the increasing demand for food, while having limited resources and increasing constraints on production practices (Fischer et al. 2012, Balezentis et al. 2020). Complexity of economic viability of agricultural production is exacerbated by differ- ent social, economic and sometimes environmental issues that cause the instability of FI. Therefore, one of the major objectives of the state intervention in agriculture and agricultural markets is FI stabilization (Meuwissen et al. 2008). Less volatile FI is also among the key objectives of the CAP. Starting in the 1980s and reinforced by the 1992 MacSharry and consequent reforms, the CAP has gradually drifted away from intervening in supply-demand mechanisms that used to favour higher prices, to the decoupled direct payments and other financial measures that help to ensure the financial stability of farmers (Tangermann 1998, Bojnec and Fertő 2019). Therefore, safe- guarding sufficient food production and ensuring a fair standard of living for people involved in agriculture still remains one of the main aims of the CAP. That is why, in recent years, a significant interest has emerged in ana- lysing the farm-level data to explore the role of agricultural subsidies in FI variability. While the existing academic literature includes studies on the role of direct payments in stabilizing FI, empirical evidence on this subject is limited. E.g., Mishra and Sandretto (2002) measured the variability in real net FI among the U.S. farms from 1933 to 1999, and found that variability of FI far exceeds the variability of the income of non- farm households. Vrolijk and Poppe (2008) analysed the volatility of FI in several EU countries and found that while FI variability is generally high, there are differences among the EU countries and types of farms, more specialized farms and smaller farms are often faced with relatively higher FI variability. Wąs and Malak-Rawlikowska (2011) found that the risk of low incomes is much higher in smaller dairy farms in Poland. Also, Agrosynergie (2011) found that FI variability is higher in smaller farms in most types of farming and across regions. The significant role of Received 17 November 2021 / Accepted 16 April 2022 The Scientific Agricultural Society of Finland ©This is an open access article under the CC BY 4.012 O. Aleksandrova et al. 25 direct payments in stabilizing FI in Switzerland was confirmed by El Benni et al. (2012), and for France and Italy by Enjolras et al. (2014). According to El Benni et al. (2012), the switch from market-based support to direct payments has decreased the variability of farm revenues and household income. Hadrich (2013) found that government payments and insurance payments increased farm revenue variation in crop and livestock farms in the Northern Great Plains. Severini et al. (2016a, 2016b) studied the role of CAP direct payments in stabilizing FI for Italian farms, using individual farm data of a decade. They established that small farms face a higher level of income variability than larger farms, but farms that did not benefit at all from direct payments had a relatively lower level of income variability. Van Asseldonk et al. (2008) conclude that farming is and will be a risky business, and arising from climatic conditions, substantial volatility remains despite risk management instruments. In Estonia, risks in agriculture have been previously studied from the perspectives of yield and price risk in crop production (Läänemes et al. 2011, Nurmet et al. 2016, Nurmet et al. 2018). According to our knowledge, there are no previous studies in the last five years on the role of agricultural subsidies in stabilization of FI in the Baltic States. The aim of the study is, therefore, to assess whether the agricultural subsidies stabilize FI in Estonia. The current study follows the approach applied by Severini et al. (2016a, 2016b, 2017) in case of Italy, and by Bojnec and Fertő (2019) in case of Slovenia and Hungary, but our analysis is applied to different country (Estonia) that has different climatic conditions and agricultural structure, and covers longer time period (2006‒2019). Further, in our study we analyse the role of subsidies in stabilizing FI also in different farm types (crops, livestock and mixed crops and livestock farms). Sectoral distinction has not been made in the previous studies. The analysis is based on a balanced data sample from the Estonian Farm Accountancy Data Network (FADN) from the period of 2006–2019. Dynamics of Estonian agriculture 2006‒2019 Estonia and the other Baltic States are known for their dynamic economies and agricultural sector that have been significantly influenced by the EU funds. After the EU accession, the sum and average rate per hectare of agricul- tural subsidies increased in Estonia (Viira et al. 2009, Olper 2011, OECD 2018). In the period 2006‒2019, in Estonia, the total amount of subsidies on products and other subsidies in production increased by 96%, while output of the agricultural industry increased by 75% (Fig. 1). At the same time increase in crop output (by 139%) outpaced the growth of animal output (41%), agricultural services and non-agricultural secondary activities (19%). Of these three output categories, crop output has the highest variation, suggesting that further increase in the share of crop output in total agricultural output may increase FI variation in the future. While the sum of agricultural subsidies has increased, the growth has not been stable. Figure 2 shows that in the period 2014‒2016, there was a decline in the total subsidies, which could be attributed mainly to two reasons: 1) the political decision not to top-up direct payments with transitional national aid in this period; 2) decline in the CAP II Pillar payments due to end of the programming period 2007‒2013, and beginning of the programming pe- riod 2014‒2020. The fact that agricultural subsidies show some instability further necessitates the study on their role in stabilizing FI. Fig. 1. Output of the agricultural industry, agricultural subsidies on products and production, and total intermediate consumption in Estonia, 2006‒2019 (current prices, 2006=100). Source: Statistics Estonia (2021) Agricultural and Food Science (2022) 31: 24–36 26 Material and methods Data Our study examines the role of agricultural subsidies in FI instability in Estonia. For that, the analysis employed individual farm-level data from the Estonian FADN database from the period 2006–2019. In 2019, Estonian FADN database included 658 sample farms that represented the population of 7610 farms, which standard output exceed- ed 4000 euros (Aamisepp et al. 2020). The FADN sample is rotating, therefore, for panel data analysis of 14-years period a limited number of farms could be included. From the whole dataset, a balanced panel data subset was made that covered the 14-year period and contained 3164 observations from 226 farms. Following Bojnec and Fertő (2019), we analysed the role of total agricultural subsidies instead of only direct payments, even though direct payments comprise more than 50% of the total subsidies targeted at FI enhancement and stabilization. Estonia and other Member States that accessed the EU in 2004 and later initially had quite low level of direct pay- ments. This is why payments from the CAP II Pillar have had relatively more important role in FI compared to the EU average or older EU Member States. Estonian farms have a relatively high share of subsidies in their FI (Table 1). Therefore, subsidies are important for the level of FI and its stability, which can also depend on the type, economic size and specialization of the farm (Bojnec and Fertő 2019). In the FADN methodology, the type and economic size of farm households are determined based on the value of the standard output (SO) of the holding in Euros. In Estonia, the minimum threshold for inclusion in the FADN farm population is the SO of 4000 Euros (Aamisepp et al. 2020). The farm’s SO determines its membership in the corresponding economic size classes that range between 3 (standard output 4000‒8000 Euros) and 14 (standard output more than 3000000 Euros). In the analysis the whole balanced data sample was divided in sub-samples according to: (i) farm economic size classes: (a) σ 1 ; (b) σ 2 ; (c) σ 3 ; (d) σ 4 ; (ii) relative share of subsidies in total FI, which is denoted as a PSE level: (a) σ 1 ; (b) σ 2 ; (c) σ 3 ; (d) σ 4 ; (iii) three farm type groups: crops, livestock and mixed crops and livestock (Table 1). These main farm types were selected according to the EU classification of types of farming (EC 2015). In the group crops, we included farm types 15 (specialist cereals, oilseeds and protein crops), 16 (general field cropping), 21 (spe- cialist horticulture indoor), 22 (specialist horticulture outdoor), 36 (specialist fruit and citrus fruit), 38 (various permanent crops combined), 61 (mixed cropping). In the group livestock, we included farm types 45 (specialist dairying), 46 (specialist cattle – rearing and fattening), 47 (cattle – dairying, rearing and fattening combined), 48 (sheep, goats and other grazing livestock), 51 (specialist pigs), 52 (specialist poultry), 73 (mixed livestock, mainly grazing livestock), 74 (mixed livestock, mainly granivores). Mixed crops and livestock includes farm types 83 (field crops – grazing livestock combined) and 84 (various crops and livestock combined). Fig. 2. Paid agricultural subsidies in Estonia, million Euros, 2006‒2019. Source: Estonian Ministry of Rural Affairs (2022) 31 44 51 61 73 82 92 99 110 106 123 124 135 144 31 40 50 29 35 36 33 21 19 18 15 20 8 19 54 61 105 113 126 152 167 162 126 93 115 132 148 140 0 50 100 150 200 250 300 350 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 Pa id s ub si di es , m ill io n Eu ro s I Pillar - Direct payments I Pillar - Transitional national aid I Pillar - Market intervention II Pillar - Total National payments O. Aleksandrova et al. 27 To reduce the effect of price inflation on the results of the analysis, the values of all variables were deflated by the total agricultural output price index, which was obtained from the database of the Statistics Estonia (2021). 2006 was considered as the base year. The number of observations in our sample and subsamples must be divisible by 14 (as we use a balanced panel for 14 years). Due to this, the division of observations is slightly uneven in the sub-samples (3164/14=791 obser- vations, but 791/14=56.5 farms). Therefore, sub-samples (i) and (ii) include observations of 14 years from either 56 or 57 farms. Sub-sample (iii) is classified into farm type groups, and farms are not evenly distributed between these groups. Methodology In this study, FI is defined as: (1) where FI is farm income as given by FADN variable gross farm income (FADN code SE410). REV is revenues meas- ured by the FADN variable coded total output (SE131), and defined as the total value of the output of crops and crop products, livestock and livestock products and other output, including other gainful activities of farms. EC is cost of external factors measured by the FADN variable total intermediate consumption (SE275), and defined as the total specific cost, including inputs produced on the holding and overheads arising from production, includ- ing machinery costs. TS is total subsidies, measured by the FADN variable total subsidies for current operations linked to production, excluding investments (SE605). Investment subsidies were excluded since not all farms apply and receive those, and for those farms who do, the flow of investment subsidies is very unstable. Moreover, the investment subsidies compensate certain proportion (usually 40–50%) of costs of certain investments and should not be considered as part of FI. Therefore, MI represents market income (FI–TS). First, we analysed the relative importance of TS in FI using the following two indicators: , (2) Where PSE is a proxy of producer support estimate. Strictly speaking, the PSE in this study does not fully correspond to the PSE as suggested originally by Josling in the 1970s (Cahill and Legg 1990) and used by the Organization of Economic Cooperation and Development (OECD) for decades. However, the proxy PSE as suggested Table 1. The share of total subsidies in FI (full sample including observations with negative MI) Number of observations Number of farms Agricultural area (mean), ha PSE (mean) TS/FI (mean) Quartiles of economic size σ 1 784 56 40.9 0.339 0.794 σ 2 798 57 113.7 0.295 -2.715 σ 3 798 57 225.6 0.275 0.798 σ 4 784 56 733.2 0.202 0.536 Quartiles of PSE σ 1 784 56 520.5 0.131 0.328 σ 2 798 57 256.5 0.229 0.632 σ 3 798 57 179.3 0.307 0.823 σ 4 784 56 155.2 0.441 1.172 Farm types Crop 1092 78 277.4 0.264 -1.835 Livestock 1470 105 307.8 0.296 0.874 Mixed crops and livestock 602 43 204.6 0.274 0.477 Total sample 3164 226 251.1 0.270 0.687 Source: authors’ calculations based on FADN data 𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹 = 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 − 𝑅𝑅𝑅𝑅𝐸𝐸𝐸𝐸 + 𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇 = 𝑀𝑀𝑀𝑀𝐹𝐹𝐹𝐹 + 𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇, 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃 = 𝑇𝑇𝑇𝑇𝑃𝑃𝑃𝑃 𝑇𝑇𝑇𝑇𝑃𝑃𝑃𝑃 + 𝑅𝑅𝑅𝑅𝑃𝑃𝑃𝑃𝑅𝑅𝑅𝑅 Agricultural and Food Science (2022) 31: 24–36 28 by Severini et al. (2016a) and Bojnec and Fertő (2019) represents the relative importance of TS in the whole farm receipts (TS+REV), and: (3) that represents the share of subsidies in FI. Table 1 presents the results for both indicators (2) and (3). Second, we use the coefficient of variation (CV) for measuring variability of FI. CV measures the extent of data variability in the sample relative to the mean and it was calculated to measure of variability of FI over the 14-year period for each individual farm using Equation (4): , (4) where σ is a standard deviation, and μ is the mean of FI of each individual farm over 14-year period. Our sample includes farms with negative values of FI for some years (there are no farms with negative average FI over the 14-year period). Therefore, the average value of FI can be close to zero due to this. However, the sample includes farms with negative mean value of market income (MI). There are also farms with zero-values of TS for the whole period under analysis. The CV indicates the size of standard deviation compared to the mean. Comparing the CVs is therefore equivalent to comparing data on a logarithmic scale. If the mean values of variables are nega- tive, close to zero or zero, then the CV becomes negative, very large or does not exist. Therefore, if the mean is zero, close to zero or negative, such comparison becomes meaningless (Grimmett 2020). To avoid this problem, for analysing the volatility of FI over time, we restricted our sample to farms that had a positive mean value of MI and non-zero values of TS for the whole period under analysis (81% of the whole sample) (Tables 2, 4, 5). The second caveat with the data is related to the fact that in general, from 2006‒2019, agricultural output, subsidies and cost of inputs in current prices increased (Fig. 1). This generally increasing trend suggests that analysing variance of FI via CV might result with overestimated variation. Theoretically, one of the solutions in this situation could be de- trending the data. However, this should be done at the individual farm level, because different farms are in the different stage (growth, maturity, or decline) of their life cycle (Viira at al. 2013). Also, since the values of all vari- ables were deflated by total agricultural output price index, the price inflation effect was already removed from the time-series data. Another reason for not detrending the data was related to variance decomposition analysis used in this paper. According to Offutt and Blandford (1986) and Severini et al. (2016a), using detrended data in the variance decomposition analysis leads to biased results. The significance of the differences in CV(FI) among the quartiles of farms’ economic size and PSE level was statis- tically tested using the non-parametric Conover-Iman test (Conover and Iman 1979) (Table 2). The null hypoth- esis (H 0 ) was that, the median values of CV(FI) are equal for farms in different size quartiles, for different PSE level quartiles, and also for different farm type groups. The results of the Conover-Iman test were presented as a p-value and it was calculated at significance level α. H 0 was rejected, if p ≤ α/2 (Table 2). Third, variance and covariance of FI were assessed. Variance is a measure of dispersion, i.e. it is a measure of how far a set of values is spread out from their average. Covariance is a measure of the joint variability of two random variables. If the greater values of one variable mainly correspond with the greater values of the other variable, and the same holds for the lesser values (that is, the variables tend to show similar behaviour), the covariance is positive. In the opposite case, when the greater values of one variable mainly correspond to the lesser values of the other (that is, the variables tend to show opposite behaviour), the covariance is negative (Grimmett 2020). According to the literature (e.g., El Benni and Finger [2013], Severini et al. [2016a, 2017], Bojnec and Fertő [2019]), the importance of the three components of Equation 1, (i.e. revenues, costs of external factors and subsidies) on FI variability was evaluated by employing variance decomposition of income components, relying on multiplica- tive or additive identities. The variance was decomposed, using the general relationship (Hadrich 2013): . (5) When this general relationship (5) is applied to the FI (Equation 1), the result is a variance decomposition of FI (Equation 6): 𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇 𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹⁄ 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 = 𝜎𝜎𝜎𝜎 𝜇𝜇𝜇𝜇 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉(𝑥𝑥𝑥𝑥 ± 𝑦𝑦𝑦𝑦) = 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉(𝑥𝑥𝑥𝑥) + 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉(𝑦𝑦𝑦𝑦) ± 2𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶(𝑥𝑥𝑥𝑥, 𝑦𝑦𝑦𝑦) O. Aleksandrova et al. 29 . (6) Dividing the result of Equation (6) by the sum of the first three variance terms gives a normalized form that is use- ful for interpretation (Equation 7): , (7) where p 1 , p 2 , p 3 , are the direct effects, and p 12 , p 13 , p 14 , are the indirect effects of three considered income components: . In Equation 7 the three direct effects sum to unity and an increase in the variance of any of these components increases the variability of FI. In the empirical analysis, variance and covariance of FI, REV, TS and EC (Equation 6) and its normalized version (Equation 7) were calculated for each individual farm over the 14-year period. The percent contribution of each variance and covariance source were calculated for each individual farm and it was averaged over all farms (Table 3). Fourth, if variation of TS as FI component is lower than variability of remaining part of FI, i.e. MI, then subsidies stabilise FI. To understand this, we calculated the corresponding coefficients of variance, PSE and TS/FI for the restricted sample (with MI>0) (Table 4). Wilcoxon rank-sum test (Mann and Whitney 1947) was used to test the significance of the differences between median values CV(FI), CV(MI) and CV(TS) (Table 4). Fifth, following Severini et al. (2016a) to explore whether agricultural subsidies have a countercyclical role against fluctuation of MI over time, we calculated Pearson’s correlation between TS and MI based on the 14 years of data for each farm. Table 5 presents the average of these correlations between TS and MI for the whole sample and each sub-sample. To assess whether the share of agricultural subsidies in total revenues stabilise FI, Pearson’s correlation between PSE and CV(MI) was computed. Since we had only one value of CV(MI) for each farm over the 14-year period, in this case the correlation was calculated at the whole sample and subsample levels. Table 5 shows these results as well. The data was analysed using the program R (package “lmtest”) and MS Excel. Results and discussion Share of agricultural subsidies in FI On average total subsidies accounted for 27% of total farm receipts (PSE) and 69% of farm income (TS/FI) in Estonia in the period 2006‒2019 (Table 1). In Italy, direct payments accounted for 13% of total farm receipts (PSE), and 43% of farm income (TS/FI) (Severini et al. 2016a). For Slovenia and Hungary, the average of PSE was 27% and 23% respectively and TS/FI was 17% and 44% respectively (Bojnec and Fertő 2019). In livestock farms the PSE was 30% and TS/FI was 87%; in crop farms, the PSE as 26% and TS/FI was -184%; in mixed crops and livestock farms the PSE was 27% and TS/FI was 48%. This implies that for livestock farms, the subsidies were slightly more impor- tant source of income compared to crop and mixed crop and livestock farms. In Estonia, the PSE was higher in smaller farms. In the 1st economic size quartile the PSE was 34%, and in the 4th size quartile it was equal to 20%. In the PSE quartiles, the PSE level increases from the 1st quartile to the 4th PSE quar- tile accordingly from 13 to 44%. In the 1st PSE quartile, the farms have above average agricultural area. However, the differences in average agricultural area in 2nd, 3rd and 4th PSE quartile are not that marked. The TS/FI measure has a high volatility (ranging from negative to positive values) in the 14 years considered in the analysis. TS/FI in size quartiles ranged from –2.715 to 0.798 and did not have a clear trend related to economic size. For the 2nd size quartile, the TS/FI was negative due to negative FI, and it was lowest in the 4th size quartile. According to quartiles of the PSE, the TS/FI level ranges from 0.328 to 1.172, and it is higher in those quartiles that have higher PSE and smaller agricultural area. Such strongly scattered average values of TS/FI (ranging from negative to positive) can be explained by high volatility of FI in the 14-years period under analysis. 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉(𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹) = 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉(𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑉𝑉𝑉𝑉) + 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉(𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇) + 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉(𝑅𝑅𝑅𝑅𝐸𝐸𝐸𝐸) + 2𝐸𝐸𝐸𝐸𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶(𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑉𝑉𝑉𝑉, 𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇) − 2𝐸𝐸𝐸𝐸𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶(𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑉𝑉𝑉𝑉, 𝑅𝑅𝑅𝑅𝐸𝐸𝐸𝐸) − 2𝐸𝐸𝐸𝐸𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶(𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇, 𝑅𝑅𝑅𝑅𝐸𝐸𝐸𝐸). 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉(𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹) 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉(𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑉𝑉𝑉𝑉) + 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉(𝑅𝑅𝑅𝑅𝐸𝐸𝐸𝐸) + 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉(𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇) = 𝑝𝑝𝑝𝑝1 + 𝑝𝑝𝑝𝑝2 + 𝑝𝑝𝑝𝑝3 + 𝑝𝑝𝑝𝑝12 − 𝑝𝑝𝑝𝑝13 − 𝑝𝑝𝑝𝑝23, 𝑝𝑝𝑝𝑝1 = 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉(𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑉𝑉𝑉𝑉) 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉(𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑉𝑉𝑉𝑉)+𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉(𝑅𝑅𝑅𝑅𝐸𝐸𝐸𝐸)+𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉(𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇) , 𝑝𝑝𝑝𝑝2 = 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉(𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇) 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉(𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑉𝑉𝑉𝑉)+𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉(𝑅𝑅𝑅𝑅𝐸𝐸𝐸𝐸)+𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉(𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇) , 𝑝𝑝𝑝𝑝3 = 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉(𝑅𝑅𝑅𝑅𝐸𝐸𝐸𝐸) 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉(𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑉𝑉𝑉𝑉)+𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉(𝑅𝑅𝑅𝑅𝐸𝐸𝐸𝐸)+𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉(𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇) , 𝑝𝑝𝑝𝑝12 = 2𝐸𝐸𝐸𝐸𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶(𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑉𝑉𝑉𝑉,𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇) 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉(𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑉𝑉𝑉𝑉)+𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉(𝑅𝑅𝑅𝑅𝐸𝐸𝐸𝐸)+𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉(𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇) , 𝑝𝑝𝑝𝑝13 = 2𝐸𝐸𝐸𝐸𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶(𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑉𝑉𝑉𝑉,𝑅𝑅𝑅𝑅𝐸𝐸𝐸𝐸) 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉(𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑉𝑉𝑉𝑉)+𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉(𝑅𝑅𝑅𝑅𝐸𝐸𝐸𝐸)+𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉(𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇) , 𝑝𝑝𝑝𝑝23 = 2𝐸𝐸𝐸𝐸𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶(𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇,𝑅𝑅𝑅𝑅𝐸𝐸𝐸𝐸) 𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉(𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑉𝑉𝑉𝑉)+𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉(𝑅𝑅𝑅𝑅𝐸𝐸𝐸𝐸)+𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉𝑉(𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇) Agricultural and Food Science (2022) 31: 24–36 30 Concerning PSE indicators, similar results were observed for Slovenian farms, where PSE declined from the 1st eco- nomic size quartile to the 4th size quartile but not for Hungary (Bojnec and Fertő 2019). For Italian farms PSE de- creased from small to large farms (Severini et al. 2016a). TS/FI was highest for the second economic size quartile (0.691) and lowest for the third quartile (0.076) in Hungary, and highest for the third quartile (0.945) and lowest for the first quartile (–1.543) in Slovenia (Bojnec and Fertő 2019). For Italian farms the TS/FI increased from 7% in the 1st to 100% in the 4th PSE quartile, but for quartiles of economic size classes it declined from 56% in the small- est to 31% in the largest farms’ size class. Variability of FI For the analysis of FI volatility, restricted sample was used to avoid the interpretation difficulties that arise from the effects of negative MI on the calculation of the volatility measure (CV). According to the results (Table 2), the volatility of FI for Estonian farms is lower than reported for Italian farms and for Hungarian and Slovenian farms. On average, the median of CV(FI) was 0.34, while it was assessed to be 0.64 for Italian (Severini et al. 2016a), and 0.41 and 0.37 for Hungarian and Slovenian farms respectively (Bojnec and Fertő 2019). Even though, smaller farms had higher PSE level, variability of FI is highest for the smallest farms and then de- creases for the 2nd, 3rd and 4th quartiles of economic size. Vrolijk and Poppe (2008), Severini et al. (2016a, 2016b), and Bojnec and Fertő (2019) also found a converse relationship between variability of FI and farm size. The null hypothesis was that the medians of CV(FI) in different economic farm size quartiles are equal. A Conover-Iman test (Conover and Iman 1979) rejected the validity of the null hypothesis as p-values indicate significant differences between 1st and 4th, and 2nd and 4th size quartiles at the 5% significance level. Therefore, the farms in the 4th size quartile had significantly lower FI variability compared to farms in the 1st and 2nd size quartiles. The finding that in Estonian farms, CV(FI) increased with increasing PSE level is consistent with the earlier results for Slovenian farms, but not for Hungarian farms (Bojnec and Fertő 2019). In the Hungarian study, except for the 4th quartile of PSE level, variability of FI slightly declines with an increase in PSE. Variability of FI for Hungarian farms was the highest for the 1st quartile of PSE, and then decreases (Bojnec and Fertő 2019). For Italian farms, Severini et al. (2016a, 2016b) report only very limited differences between farms with different PSE levels without Table 2. Median of the CV(FI), results of the Conover-Iman test for equality of median values of CV(FI) in different economic size and PSE level quartiles (restricted sample without negative MI observations) Number of observations Number of farms Agricultural area (mean), ha CV(FI) median Conover-Iman test (p-value) Total sample 2534 181 313.9 0.338 Quartiles of economic size σ 2 σ 3 σ 4 σ 1 476 34 34.8 0.477 0.760 0.160 0.000 σ 2 588 42 114.2 0.431 - 0.290 0.010 σ 3 714 51 217.4 0.400 - - 0.060 σ 4 756 54 736.3 0.338 - - - Quartiles of PSE σ 1 770 55 529.9 0.352 0.080 0.020 0.010 σ 2 798 57 256.5 0.400 - 0.360 0.140 σ 3 644 46 199.4 0.421 - - 0.640 σ 4 322 23 168.9 0.479 - - - Farm types Livestock Mixed crops and livestock Crop 952 68 261.2 0.431 0.210 0.070 Livestock 1148 82 362.3 0.353 - 0.670 Mixed crops and livestock 434 31 301.7 0.341 - - the difference between medians of CV(FI) is not statistically significant at the 5% level; the difference between medians of CV(FI) is statistically significant at the 5% level. Source: authors’ calculations based on FADN data O. Aleksandrova et al. 31 a clear link between PSE level and variability of FI. A Conover-Iman test rejected the validity of the null hypothesis for equality of median values of CV(FI) between the 1st and 3rd, and the 1st and 4th PSE quartiles. The null hypoth- esis could not be rejected for other pairs of quartiles. Therefore, in the 1st quartile of PSE level that includes larger farms, the variability of FI was significantly lower than in the 3rd and 4th quartiles of PSE that included smaller farms. The variability of FI for crops farms was higher than for livestock and mixed crop and livestock farms but the dif- ferences were not statistically significant. Components of the FI variation To identify the FI components that contribute the most to the FI instability in different economic size and PSE quartiles as well as in different farm type groups, the variance decomposition of FI was used as suggested by Severini et al. (2016a), and Bojnec and Fertő (2019). Table 3 shows that most of the variance of FI (62%) is due to variance in revenues (p 1 ). Total intermediate consumption (p 3 ), and the subsidies (p 2 ) comprise a smaller part, respectively 31% and 7% in the FI variation. Variance in REV is higher for farms in the 4th (largest) size quartile, and 1st (smallest) PSE quartile. For Italian farms the variance in revenues comprised 65%, the variance of EC comprised 30% and the variance of direct payments comprised 5% of the FI variation. For Hungary, these indicators were 91% (REV), 6% (EC) and 3% (TS) respectively, and in Slovenia 84% (REV), 15% (EC) and 2% (TS) respectively. In comparison of farm types, the variance in TS and EC is higher for livestock farms and these indicators are 10% (TS) and 33% (EC) respectively (Table 3). Indirect effects, related to covariance between income components show that covariance between revenues (REV) and intermediate consumption (EC) (p 13 ) provided the highest contribution (56%, Table 3) to the total variability of FI. This also confirms that often, increase in revenues is correlated with increase in intermediate consumption, and together they increase overall variation of FI. Covariance between revenues and intermediate consumption was higher for farms in the 4th size quartile and 1st PSE quartile, i.e. for larger farms. Covariance between revenues (REV) and total subsidies (TS) (p 12 , 8%) and covariance between total subsidies (TS) and intermediate consumption (EC) (p 23 , 10%) contributed less to the overall variation of FI. For different types of farms, the result was similar: the covariance between revenues (REV) and intermediate consumption (EC) had the highest contribution to farm income variability for all farms types (Table 3). Table 3. Results of variance decomposed of FI (full sample including observations with negative MI) Variance decomposition Relative importance of income sources Direct effect Indirect effect p 1 p 2 p 3 p 12 p 13 p 23 REV/FI TS/FI EC/FI Quartiles of economic size σ 1 0.612 0.110 0.278 0.081 0.497 0.126 1.404 0.794 1.245 σ 2 0.620 0.065 0.315 0.042 0.283 0.046 -1.081 -2.715 -4.793 σ 3 0.616 0.061 0.323 0.037 0.258 0.047 1.992 0.798 1.788 σ 4 0.642 0.042 0.316 0.039 0.328 0.041 2.204 0.536 1.733 Quartiles of PSE level σ 1 0.656 0.033 0.311 0.060 0.684 0.052 2.279 0.328 1.592 σ 2 0.654 0.031 0.315 0.066 0.541 0.080 2.172 0.632 1.759 σ 3 0.620 0.077 0.303 0.087 0.547 0.104 1.900 0.823 1.716 σ 4 0.559 0.137 0.304 0.101 0.465 0.157 1.415 1.172 1.565 Farm types Crop 0.683 0.038 0.278 0.051 0.520 0.073 -0.459 -1.835 -3.281 Livestock 0.568 0.098 0.334 0.084 0.580 0.143 2.104 0.874 1.968 Mixed crops and livestock 0.581 0.096 0.323 0.105 0.559 0.121 1.637 0.477 1.173 Total sample 0.622 0.069 0.309 0.078 0.558 0.098 1.981 0.687 1.640 Source: authors’ calculations based on FADN data Agricultural and Food Science (2022) 31: 24–36 32 For Italian farms, the covariance effect between revenues and direct payments was negative suggesting that direct payments could play a countercyclical role in comparison to fluctuations of revenues over time (Severini et al. 2016a). We cannot draw the same conclusion for Estonian farms. Agricultural subsidies and FI stabilization Table 4 presents the results about PSE, TS/FI, CV(FI), CV(MI) and CV(TS) for restricted sample with only non-nega- tive MI values. The CV(TS) ranges from 0.169 for large farms to 0.262 in case of smaller farms; for crop farms the CV(TS) was 0.173 while for livestock farms the respective value was 0.210. As expected, in average, CV(TS) was lower than CV(FI) and CV(MI). Therefore, one can conclude that, due to lower variability of CV(TS), subsidies have a role in stabilising FI. The income-stabilizing role of subsidies increases as one moves from the 1st to the higher quartiles of economic size. This finding is in line with findings for Hungarian and Slovenian farms (Bojnec and Fertő 2019) and with findings for Italian farms (Severini et al. 2016a). The CV(TS) decreases when moving from the 1st to 2nd quartile of PSE level, increases from 2nd to 3rd quartile of the PSE level and decreases in the 4th quartile of the PSE level. Wilcoxon rank-sum tests (p-values< 0.000) confirmed that the differences between the median val- ues of CV(TS), CV(FI) and CV(MI) are always statistically significant at 1%. The variability of FI in general is around 56 percentage points lower than the variability of MI. This confirms total subsidies help farms to cope with a high level of MI variability (El Benni et al. 2012). Table 5 shows the results of correlation analysis between the TS and MI and CV(MI) and PSE. If TS is negatively correlated with the MI, the income stabilizing role of TS is higher (Severini et al. 2016a). Our empirical results suggest that subsidies have played a limited countercyclical role in terms of reduction of fluctuations of MI over the last 14 years. For the whole sample, the correlation between TS and MI was positive but small (0.03). In the 1st and 3rd economic size quartile, the correlation between TS and MI was negative but in the 2nd and 4th size quar- tile it was positive. In the quartiles of the PSE level, the correlation between TS and MI was positive in the 1st quartile, and negative in other quartiles. For field crop types the correlation between TS and MI was negative (–0.028) but in other types it was positive with highest value (0.174) in case of livestock farms. Therefore, we can conclude that if subsidies were designed in a way that they were negatively correlated with the MI over time, the agricultural payments could have a stronger income stabilising effect, especially so in case of larger farms in the 4th size quartile and for livestock farms. Such subsides could be, e.g. counter-cyclical payments for farmers when Source: authors’ calculations based on FADN data Table 4. The share of total subsidies in farm revenues and income, and the variation of subsidies and farm income measures (restricted sample without observations with negative MI) Importance of subsidies (mean) Median Wilcoxon rank-sum test N Agricultural area (mean), ha PSE TS/FI CV(TS) CV(FI) CV(MI) Quartiles of economic size σ 1 476 34.8 0.272 0.657 0.262 0.477 1.596 *** σ 2 588 114.2 0.258 0.513 0.209 0.431 1.095 *** σ 3 714 217.4 0.250 0.696 0.200 0.400 0.928 *** σ 4 756 736.3 0.186 0.495 0.169 0.338 0.564 *** Quartiles of PSE level σ 1 770 529.9 0.132 0.329 0.277 0.352 0.471 *** σ 2 798 256.5 0.229 0.632 0.163 0.400 0.836 *** σ 3 644 199.4 0.309 0.816 0.205 0.421 1.320 *** σ 4 322 168.9 0.415 1.084 0.168 0.479 3.971 *** Farm types Crop 952 261.2 0.234 0.484 0.173 0.431 0.896 *** Livestock 1148 362.3 0.258 0.730 0.210 0.353 0.906 *** Mixed crops and livestock 434 301.7 0.219 0.614 0.187 0.341 0.661 *** Total sample 2534 313.9 0.237 0.592 0.199 0.388 0.881 *** O. Aleksandrova et al. 33 the marketing year average price received for a covered commodity is less than the target price. Bojnec and Fertő (2019) obtained similar results for small Hungarian and Slovenian farms. The overall average of the correlation coefficient between PSE and CV(MI) was 0.325, i.e. positive and with a me- dium strength. This suggests that farms with higher PSE levels have higher volatility of market income. Among the economic size quartiles, the correlation between CV(MI) and PSE was highest in the 4th quartile. This may indicate that larger farms that have higher PSE level take more risks and therefore have more volatile MI. However, if we divided the whole sample in quartiles according to the PSE level, it appeared that the correlation between CV(MI) and PSE was negative in farms with the lowest and highest PSE level. Among the farm types, the correlation be- tween CV(MI) and PSE varies from 0.270 in case of crop farms to 0.511 in case of mixed crops and livestock farms. Conclusions This research examined the farm income stabilizing role of the CAP subsidies in Estonia, during the period 2006‒2019. For the analysis, we used a balanced data sample from the Estonian FADN database, which was subsequently divided into quartiles based on farm economic size and relative share of subsidies in FI, which was denoted as a producer support estimate (PSE) level. Additionally farm income figures and their variation was studied in crop, livestock and mixed crops and livestock farms. For the analysis of FI variation, the farms with negative MI values were removed, to assure that the results are meaningful and interpretable. In the studied farms, the average PSE level over the 14-year period was 26%, and total subsidies comprised 69% of the FI. Therefore, in Estonia, the CAP subsidies have a more significant role on FI compared to Italy, Hungary and Slovenia. The PSE level was higher in the smallest farm size quartile and increased in the subsequent size quartiles. While the first PSE quartile included larger farms, and in general the average farm size decreased in the following PSE quartiles, the differences were not marked. Among different farm types, livestock farms had highest PSE level. Coefficient of variation was used to measure variability of the FI. On average, the CV(FI) was 0.34 for the 14-year period, which was lower than in Italy, Slovenia and Hungary. Lower income variation in Estonia may be related to higher PSE and TS/FI levels in Estonia, compared to the above-mentioned countries. We found that despite higher PSE levels, smaller farms had higher variation in FI. This suggest that smaller farms experience significantly higher volatility of MI compared to larger farms, and even higher PSE level cannot offset the arising effects on FI volatility. Table 5. The results of Pearson’s correlation between MI and TS, and between CV(MI) and PSE (restricted sample without observations with negative MI) Sample size number Agricultural area (mean), ha MI and TS (mean) CV(MI) and PSE Quartiles of economic size σ 1 476 34.8 -0.104 0.349 σ 2 588 114.2 0.051 0.298 σ 3 714 217.4 -0.014 0.267 σ 4 756 736.3 0.153 0.624 Quartiles of PSE level σ 1 770 529.9 0.162 -0.229 σ 2 798 256.5 -0.008 0.325 σ 3 644 199.4 -0.049 0.101 σ 4 322 168.9 -0.002 -0.127 Farm types Crop 952 261.2 -0.028 0.270 Livestock 1148 362.3 0.174 0.311 Mixed crops and livestock 434 301.7 0.009 0.511 Total sample 2534 313.9 0.034 0.325 Source: authors’ calculations based on FADN data Agricultural and Food Science (2022) 31: 24–36 34 Our results showed that most of the variance of FI is due to variance in revenues. Variance of revenues was high- est in the 4th size quartile and smallest in the 1st size quartile, and in crop farms. Analysis of indirect sources of in- come variation revealed relatively high covariance between revenues and intermediate consumption, especially so for the farms in the 4th size quartile and 1st PSE quartile, and in livestock farms. This suggests that for larger farms that are more market oriented and have lower level of PSE, increased stability in use of inputs would reduce variation in revenues and FI. It appears also from Figure 1 that in 2009, 2014 and 2016, which, due to crises, were difficult years for agricultural sector, both intermediate consumption and total output decreased. Covariance be- tween revenue and total subsidies was small but positive. Therefore, according to our results, total subsidies did not have countercyclical effects on variation of FI. Nevertheless, variation in total subsidies was lower than variation in FI and MI. Therefore, total subsidies are more stable than market income and total income, and could be considered as stabilizing element of FI. This is also con- firmed by the fact that in average, CV(FI) was 56 percentage points lower than CV(MI). It also appeared that the income stabilizing role of subsidies was higher in the 4th farm size quartile, i.e. in case of larger farms, and also in case of crop and mixed crops and livestock farms. Among the economic size quartiles, the correlation between CV(MI) and PSE was highest in the 4th quartile and in case of mixed crops and livestock farms, suggesting that larger farms, and also mixed crops and livestock farms with higher PSE level take more risks and therefore have more volatile MI. Policy implications The results suggest that subsidies have played only a limited countercyclical role in terms of reduction of fluctu- ations of MI. Therefore, policy makers should make efforts to design subsidies in a way that they are negatively correlated with the MI over time. That would increase the income stabilizing effect of the agricultural payments, especially so in case of larger, and presumably more market-oriented farms. The results of our study show that even though farms in smaller size quartiles generally have higher PSE level, variation in FI is still larger in smaller farms. Therefore, in smaller farms, the subsides were not sufficient to offset high volatility of MI. This raises questions about the appropriate policy for improving incomes and reducing income variation in smaller farms. Theoretically, increasing the subsidies for smaller farms would result in reduced variation in their FI. However, it does not help to reduce volatility of MI. Therefore, the extension and knowledge transfer programmes should also provide smaller farms with adequate expertise on how to improve farm management, and increase farm size to improve its financial stability. Also, more counter-cyclical farm payments would help to improve FI stability, especially so for farms in the larg- est size quartile, where the correlation between market income and total subsidies was positive and larger than in other quartiles. This may be related to additional subsidies paid in times of crises, which are often delayed, so that the farmers receive actual support next year. Therefore, the crisis payments should be paid with as small de- lay as possible. While theoretically appealing, introduction of countercyclical elements to the CAP either in the form of counter- cyclical direct payments or revenue insurance payments, would require establishing EU-wide harmonized farm accounting system for participating farms. In Estonia, where most of the production comes from relatively large farms and national registry of invoices, and online tax declarations already exists, this might be technically more easily feasible than in the EU member states that have large number of very small farms. Therefore, it should be kept in mind that these new instruments would not bring about additional differentiation between larger and smaller farms and member states with different farm structures. Limitations As the limitations of our study, we can note the following: (i) the analysis disregards off-farm income that has been found to reduce the variability of the overall income of farm households (see e.g. Mishra and Sandretto 2002); (ii) in the analysis of volatility of FI we restricted our sample to farms with non-negative MI, thus the analysis omits almost 20% of farms. This shortcoming should be addressed by future studies. Future studies could also address more thoroughly the relationship between total intermediate consumption and farm revenues, which had a positive covariance in our study, as our data was not sufficient to analyse the causes of variation in total intermediate consumption. O. Aleksandrova et al. 35 Acknowledgments We thank Ms. Marju Aamisepp from the Estonian Agricultural Research Centre for providing the FADN data, and Ms. Katre Kirt from the Estonian Ministry of Rural Affairs for providing the agricultural subsidies data. References Aamisepp, M., Varendi, A., Persitski, H., Kaio, U., Matveev, E. & Kärner, M. 2020. FADN Farm Return 2019. Agricultural Research Centre, 131 p. https://pmk.agri.ee/sites/default/files/2020-12/2020-FADN-majandusnaitajad-2019-aasta-kohta-MMAO-eraa- mat_20-12-18.pdf. Agrosynergie 2011. Evaluation of income effects of direct support. Final report, Bruxelles, May. https://op.europa.eu/en/publi- cation-detail/-/publication/1ab836ce-38f1-4bed-873e-fe2a3477b0d2. Bakucs, Z., Fertő, I., Varga, Á.& Benedek, Z. 2018. Impact of European Union development subsidies on Hungarian regions. Euro- pean Planning Studies 26: 1121–1136. https://doi.org/10.1080/09654313.2018.1437394 Balezentis, T., Ribasauskiene, E., Morkunas, M., Volkov, A., Streimikiene, D. & Toma, P. 2020. Young farmers’ support under the Common Agricultural Policy and sustainability of rural regions: Evidence from Lithuania. Land Use Policy 94: 104542. https://doi.org/10.1016/j.landusepol.2020.104542 Bojnec, Š. & Fertő, I. 2019. Do CAP subsidies stabilise farm income in Hungary and Slovenia? Agricultural Economics - Czech 65: 103–111. https://doi.org/10.17221/190/2018-AGRICECON Cahill, C.& Legg, W. 1990. Estimation of Agricultural Assistance Using Producer and Consumer Subsidy Equivalents: Theory and Practice. OECD Economic Studies, No. 13. Conover, W.J., & Iman, R.L. 1979. On Multiple-Comparisons Procedures. University of California. Los Alamos Scientific Laboratory. https://doi.org/10.2172/6057803 El Benni, N., Finger, R. & Mann, S. 2012. Effects of agricultural policy reforms and farm characteristics on income risks in Swiss agriculture. Agricultural Finance Review 72: 301–324. https://doi.org/10.1108/00021461211277204 El Benni, N. & Finger, R. 2013. Gross revenue risk in Swiss dairy farming. Journal of Dairy Science 96: 936–948. https://doi.org/10.3168/jds.2012-5695 Enjolras, G., Capitanio, F., Aubert, M. & Adinolfi, F. 2014. Direct payments, crop insurance and the volatility of farm income. Some evidence in France and Italy. New Medit 1: 31–40. Estonian Ministry of Rural Affairs 2022. Paid agricultural subsidies in Estonia, million Euros, 2006-2019. Data received on 14.03.2022. EC 2015. Commission Implementing Regulation (EU) 2015/220 of 3 February 2015 laying down rules for the application of Council Regulation (EC) No 1217/2009 setting up a network for the collection of accountancy data on the incomes and business opera- tion of agricultural holdings in the European Union. Fischer, G., Hizsnyik, E., Prieler, S., van Velthuizen, H. & Wiberg, D. 2012. Scarcity and abundance of land resources: competing uses and the shrinking land resource base. In: Worlds Within Reach: From Science to Policy - IIASA 40th Anniversary Conference, 24-26 October 2012, Hofburg Congress Center: Vienna and IIASA, Luxemburg Austria. Grimmett, G. 2020. Probability and Random processes. Oxford University Press. 688 p. Hadrich, J.C. 2013. Quantifying the sources of revenue variation in the Northern Great Plains. Agricultural Finance Review 73: 493–506. https://doi.org/10.1108/AFR-09-2012-0046 Lillemets, J., Fertő, I.& Viira, A.-H. 2022. The socioeconomic impacts of the CAP: Systematic literature review. Land Use Policy 114: 105968. https://doi.org/10.1016/j.landusepol.2021.105968 Läänemets, O., Viira, A.-H. & Nurmet, M. 2011. Price, Yield, and Revenue Risk in Wheat Production in Estonia. Agronomy Research 9(Special Issue II): 421–426. Mann. H.B. & Whitney, D.R. 1947. On a test of whether one of two random variables is stochastically larger than the other. The Annals of Mathematical Statistics. 18: 50–60. https://doi.org/10.1214/aoms/1177730491 Meuwissen, M.P.M., van Asseldonk, M.A.P.M. & Huirne, R.B.M. 2008. Income Stabilisation in European Agriculture. Wageningen Academic Publishers: Wageningen. 224 p. https://doi.org/10.3920/978-90-8686-650-2 Mishra, A.K. & Sandretto, C.L. 2002. Stability of Farm Income and the Role of Nonfarm Income in U.S. Review of Agricultural Eco- nomics 24: 208–221. https://doi.org/10.1111/1058-7195.00014 Moehler, R. 2008. The Internal and External Forces Driving CAP Reforms. In: Swinnen, J.F.M. (ed.): The Perfect Storm. The Political Economy of the Fischler Reforms of the Common Agricultural Policy. Centre for the European Policy Studies. Brussels. p. 76–82. Nurmet, M., Lemsalu, K. & Põder, A. 2016. Agricultural insurance in Estonia - Current situation and farmers’ willingness to use crop insurance. Science and Studies of Accounting and Finance: Problems and Perspectives 10: 122−128. https://doi.org/10.15544/ssaf.2016.12 Nurmet, M., Lemsalu, K.& Põder, A. 2018. Yield risk and the potential for insurance in agriculture. Proceedings of the19th In- ternational Scientific Conference “Economic Science for Rural Development 2018”, 9-11 May 2018, Jelgava, Latvia. p. 147–153. OECD 2018. Innovation, Agricultural Productivity and Sustainability in Estonia, OECD Food and Agricultural Reviews, OECD Publishing, Paris. https://doi.org/10.1787/9789264288744-en Agricultural and Food Science (2022) 31: 24–36 36 Offutt, S.E.& Blandford, D. 1986. Commodity market instability: empirical techniques for analysis. Resources Policy 12: 62–72. https://doi.org/10.1016/0301-4207(86)90049-8 Olper, A. 2011. Constraints and Causes ofthe 2003 EU Agricultual Policy Reforms. In: Swinnen, J.F.M. (ed.): The Perfect Storm. The Political Economy of the Fischler Reforms of the Common Agricultural Policy. Centre for the European Policy Studies. Brussels. p. 83–102. Pe’er, G., Zinngrebe, Y., Moreira, F., Sirami, C., Schindler, S., Müller, R., Bontzorlos, V., Clough, d., Bezák, P., Bonn, A., Jürgens, B.H., Lomba, A., Möckel, S., Passoni, G., Schleyer, C., Schmidt, J.& Lakner, S. 2019. A greener path for the EU Common Agricultural Pol- icy. Science 365: 449–451. https://doi.org/10.1126/science.aax3146 Rac, I., Erjavec, K., & Erjavec, E. 2020. Does the proposed CAP reform allow for a paradigm shift towards a greener policy?. Spanish Journal of Agricultural Research 18: e0111. https://doi.org/10.5424/sjar/2020183-16447 Severini, S., Tantari, A. & Di Tommaso, G. 2016a. Do CAP direct payments stabilise farm income? Empirical evidences from a con- stant sample of Italian farms. Agricultural and Food Economics 4. https://doi.org/10.1186/s40100-016-0050-0 Severini, S., Tantari, A. & Di Tommaso, G. 2016b. The instability of farm income. Empirical evidences on aggregation bias and het- erogeneity among farm groups. Bio-based and Applied Economics 5: 63-81. https://doi.org/10.13128/BAE-16367 Severini, S., Tantari, A. & Di Tommaso, G. 2017. Effect of agricultural policy on income and revenue risks in Italian farms. Agricul- tural Finance Review 77: 295–311. https://doi.org/10.1108/AFR-07-2016-0067 Schebesta, H. & Candel, J.J.L. 2020. Game-changing potential of the EU’s Farm to Fork Strategy. Nature Food 1: 586–588. https://doi.org/10.1038/s43016-020-00166-9 Statistics Estonia. 2021. http://www.stat.ee/ Tangermann, S. 1998. An Ex-post Review of the 1992 MacSharry Reform. In: Ingersent K.A., Rayner A.J., Hine R.C. (eds) The Re- form of the Common Agricultural Policy. Palgrave Macmillan, London. https://doi.org/10.1007/978-1-349-26101-7_2 Van Asseldonk, M.A.P.M., Majewski, E., Meuwissen, M.P.M, Was, A., Guba, W., Dalton, G.,Landmesser, J., Berg, E.& Huirne, R.B.M. 2008. Economic impact of prospective risk management instruments under alternative policy scenarios. Paper prepared for pres- entation at the 108st EAAE Seminar ‘Income stabilisation in a changing agricultural world: policy and tools’, Warsaw, Poland, 8-9 February, 2008. Viira, A.-H., Põder, A.& Värnik, R. 2009. 20 years of transition - institutional reforms and the adaptation of production in Estonian agriculture. Agrrwirtschaft 58: 294-303. Viira, A.-H., Põder, A.& Värnik, R. 2013. The Determinants of Farm Growth, Decline and Exit in Estonia. German Journal of Agri- cultural Economics 62: 52–64. Vrolijk, H.C.J. & Poppe, K.J. 2008. Income variability and income crisis in the European Union. In: Meuwissen M.P.M., van Assel- donk M.A.P.M., Huirne R.B.M. (ed.): Income Stabilisation in European Agriculture. Wageningen Academic Publishers: Wagenin- gen. p. 33–54. Wąs, A., & Malak-Rawlikowska, A. 2011. Policy impact on production structure and income risk on Polish dairy farms. The Common Agricultural Policy after the Fischler Reform: National Implementations, Impact Assessment and the Agenda for Future Reforms. p. 183–193. The role of subsidies in stabilising farm income: Evidence fromEstonia Introduction Dynamics of Estonian agriculture 2006‒2019 Material and methods Data Methodology Results and discussion Share of agricultural subsidies in FI Variability of FI Components of the FI variation Agricultural subsidies and FI stabilization Conclusions Policy implications Limitations Acknowledgments References