
































Microsoft Word - 132-COPYEDIT-FINAL.docx


http://journals.sfu.ca/abr  ADVANCES IN BUSINESS RESEARCH 
2016, Volume 7, pages 51-65 

	
	

51 
	

Evaluating the role of income in  
national software piracy 

 
Eric Kyper 

Lynchburg College 
 

Gerald Prante 
Lynchburg College 

 
Lee Schimmoeller 

Lynchburg College 
 

This study investigates the proper role of income in predicting national 
software piracy rates. We run regressions isolating various measures of income, 
including GDP per capita and median household income and variations thereof, to 
predict national software piracy rates. Then we also run multivariate regressions 
incorporating GDP and other non-income predictors such as corruption in a manner 
consistent with previous studies. This topic is of importance due to the multi-billion 
dollar losses incurred globally each year due to pirating. The results show that a 
square root version of median household income is the best measure of national 
income and consistent with what economic theory predicts.  

 
Keywords:  Software piracy; income; culture 
 

Introduction 
 
Software piracy is a well-studied topic with global losses in excess of $62 billion (BSA, 2014). 

However, the determinants of software piracy vary from study to study. One variable in common is 
some measure of national income, typically GDP per capita. Economic theory dictates that, ceteris 
paribus, higher income leads to more discretionary spending. Following that lead we conclude that 
wealthier nations should have lower national software piracy rates. Indeed this is generally true, but 
does not eliminate variation in software piracy rates among wealthy nations. The proper role of 
income is of importance because virtually every study regarding national piracy includes income, and 
only a nuanced understanding of income will flush out real differences between countries with 
similar incomes. Previous studies report conflicting results regarding income. For example, the 
statistical significance of GDP varies across studies and year of study. GDP is also sometimes used 
as a proxy for other data such as number of internet users. This study makes an important 
contribution to current literature by elucidating the role of GDP. We clarify that GDP is typically 
only significant in studies where more illuminating variables are not included. This amounts to GDP 
being used as a variable of convenience. While informative, if GDP is being used as a proxy for 
better variables, we should recognize this and work towards analyzing those variables. From this, we 
also ask the question whether or not a different income measure is superior to GDP per capita in 
predicting software piracy. 



INCOME AND NATIONAL SOFTWARE PIRACY	

52 
	

In addition, a number of variables beyond income have been found to influence piracy. 
These include culture and intellectual property protection. This study aims to determine the exact 
role GDP plays in determining national software piracy rates first in isolation, then in concert with 
non-income variables.  

 
Literature Review 

 
Software piracy is a huge and growing issue in the global economy. The BSA Software 

Alliance Survey in 2013 estimated the commercial value of unlicensed software totaled $62.4 billion 
globally. Furthermore, the global rate at which software was installed without proper licensing 
increased from 42% in 2011 to 43% in 2013 (BSA, 2014). This increase was attributed to the growth 
in emerging economies where unlicensed software is prevalent. There has been a large body of 
research concerning this topic. Some of the earlier studies of national software piracy rates included 
economic, cultural, and temporal variables. Table 1 below reviews a sample of studies from 1998 to 
2014, summarizes the findings and identifies the variables included. Particular attention is given to 
GDP as an independent variable (IV).  

This review of the literature indicates that some measure of economic performance is the 
most common determinant analyzed in the research of national software piracy rates. Often, this 
economic indicator is not at the country level (GDP) but as some type of proxy for individual 
wealth. Many such as Goel and Nelson (2012) use GDP per capita, while other such as Marron and 
Steel, (2000) use per capita income. A full list of variables used is illustrated in Table 1 above. 
Additionally, Harbi, Grolleau and Bekir (2012) believe there is a Kuznets curve in that as poor 
people acquire wealth and then technology, the piracy rates increase until the wealth is at a level 
where piracy rates begin to decrease. All of these nine studies used some type of economic indicator 
of either country or personal wealth, suggesting that income is a critical factor.  

In addition to income, many of the previous studies incorporate cultural factors in their 
models. Hofstede’s (1980) cultural model is the most frequently referenced framework for 
describing national cultural differences. These five dimensions attempt to quantify the key factors 
that determine a country’s specific national culture and include power distance, individualism, 
masculinity, uncertainty avoidance, and long term orientation. Much of this research (4 of 9) focused 
on individualism versus collectivism and often found individualism to be a statistically significant 
predictor of national software piracy.  

To clarify, Hofstede defines individualism as a preference for a loosely knit social framework 
in which individuals take care of themselves and their immediate families only; they are not much 
concerned with the greater society. Collectivism is a preference for a closely knit social framework in 
which individuals are emotionally integrated into an extended family, clan or other group which will 
protect them in exchange for loyalty. Analysis has shown that members of a collective, while 
extremely loyal to the collective, are less likely to be concerned about others outside the collective, 
this includes being more likely to copy software from others outside their collective. A collective is 
associated with a nation in this study. Collectivism data is included in the study.  

 
Theoretical framework of national software piracy rates: Cost-benefit analysis 

From an economic theory perspective, software piracy is a simple cost-benefit decision for 
individuals and firms. On the benefit side, pirating software instead of paying for it simply increases 
an individual’s disposable income. And economic theory would predict that higher individual 
incomes would reduce the need to engage in software piracy due to the concept of diminishing 
marginal utility of income. This concept states that the first dollar of income adds more to an 



Kyper	

53 
 

individual's utility than the second dollar, which adds more to utility than the third dollar, and so 
forth. Put simply, a poor person making $7,000 who currently has a low utility level saving $100 
dollar by pirating a piece of software is adding more to his/her utility than a rich person making 
$100,000 doing so, ceteris paribus.  
 
Table 1. Summary of previous studies of Software Piracy Rates (SPR) where SPR is dependent 

variable (DV) 
 
Authors Sample size IVs Key Findings  Method 
(Chavarria & 
Morrison, 2014) 

42 countries GDP per Capita, 
Hofstede 5 cultural values 

Cultural values of collectivism and 
performance orientation positively 
related to SPR. Log of GDP per capital 
also positively related to SPR. 

OLS 
Regression 

(Hamister & 
Braunscheide, 2013) 

105 
countries 

GDP per Capita, 
Hofstede 5 cultural values 
& Intellectual Property 
Rights Protection 

GDP per capita and IPPR are negatively 
correlated with SPR. Also, high power 
distance is positively associated with 
SPR. Other cultural dimensions were 
not significant, including collectivism. 

OLS  
Regression 

(Goel & Nelson, 
2012) 

100 
countries 

GDP per capita shadow 
economy &  internet 
diffusion 

Underground economy has a positive 
relationship with SPR. Negative 
relationship with GDP per capita and 
internet diffusion.   

OLS  
Regression 

(Harbi, Grolleau, & 
Bekir, 2012) 

100 
countries 

GDP per capita Studied 100 countries over 15 years and 
showed piracy first increase with level of 
GDP per capita, reaches maximum, 
then decreases at higher income levels. 

OLS  
Regression 

(Yang, Sonmez, 
Bosworth, & Fryxell, 
2009) 

59 countries GNI per capita, 
Individualism & 
technology development 

All 3 IV’s are negatively correlated with 
SPR. 

OLS  
Regression 

(Gopal & Sanders, 
Sep 2000) 

65 countries Per capita GNP An inverse relationship between 
software piracy rates and per capita 
GNP with a break at $6,000 with those 
below the break being affected more. 

OLS  
Regression 

(Marron & Steel, 
April 2000) 

77 countries, Per capita income, 
individualism, R&D, 
Education, strength of 
economicinstitutions 

The inverse relationships of 
individualism, per capita income and 
strength of economic institutions 
explain the greatest amount of variance 
in software piracy rates. 

OLS  
Regression 

(Husted, 2000) 39 countries, Per capita GNP, 
individualism, income 
inequality, masculinity, 
power distance, 
uncertainty avoidance 

Included cultural variables finding that 
GNP per capita, income inequality, and 
individualism are significantly related to 
software piracy (all relationships are 
inverse). 

Correlation 

(Gopal & Sanders, 
December 1998) 

13 countries GDP per capita, domestic 
software industry size 

Provides support for the proposition 
that a government’s incentive to enact 
and enforce copyright protection laws is 
closely related to the size of the 
domestic software industry. Also 
presents a model for ethical intentions 
based on ethical predisposition and 
demographics. This information can be 
used to create better prevention and 
deterrent controls for software piracy. 

OLS  
Regression 



INCOME AND NATIONAL SOFTWARE PIRACY	

54 
	

 
Of course, utility is not directly observed or quantifiable. A standard form for writing a 

utility-income function is  
 
Eq. 1: U = cIα 

 
where c is a constant, I is income and α is the key parameter that determines the marginal effect of 
more income on utility. This form is derived from the standard Cobb-Douglas utility function. 
Mathematically, starting from the following Cobb-Douglas utility function, which expresses utility as 
a function of the quantities of goods X and Y consumed, where a and b are preference parameters: 
 

Eq. 2: ! = !!!! 
subject to the budget constraint (I = income, assuming all income is spent on X and Y): 
 
Eq. 3: ! = !!! +  !!! 
we can derive the following utility-income function: 
 

Eq. 4: ! =  !(!!!)

!!! (!!!!)! !!! (!! !!)!
 

 
The denominator of this fraction, which consists of the prices of goods X and Y and parameters, is 
a constant as it relates to income. If a+b < 1, then diminishing marginal utility of income holds 
because increasing income by a factor (a+b) would increase utility by less than that factor (a+b). The 
speed at which utility is increasing as income increases depends on the value of the factor (a+b). If 
a+b is close to 1, then utility is increasing as income increases at a rate that is relatively fast. If a+b is 
close to 0, then utility is increasing as income increases but at a relatively slow rate. For the purposes 
of this paper, we use a simple square root of income to approximate the utility-income function. 
Specifically, we assume that the denominator of the utility-income function above is 1 and a+b = 
0.5. (Eq. 1: α = 0.5 and c = 1) The hypothesis is that using the square root of income instead of the 
simple dollar value is likely to be a better proxy for the current utility level of a country's average (or 
median) citizen. This works because the square root captures the effect of diminishing marginal 
utility on income. A square root is adopted as a parsimonious choice in deriving diminishing 
marginal utility. 

In addition to the question of whether or not to use simple income or a utility proxy that 
assumes diminishing marginal utility of income, there is also the question of which measure of 
central tendency for income in a country should be used to calculate the marginal utility from 
software piracy. Because income tends to be highly skewed, median income is often used to better 
gauge the economic position of a country's typical citizen than is mean income. That is the 
hypothesis in this paper. However, it should be noted that it is technically possible that using mean 
income could serve to better represent the economic position of software users (weighted by the 
amount of software they use) as compared to median income. That is because in many low-income 
countries a large fraction of the population does not use software in the first place. Therefore, as 
their inclusion at the low end of the distribution pushes down both the median and mean, it could 
lead to a situation where mean income for all citizens is actually closer to the median software user's 
income than is the median income of all citizens. Unfortunately, country-by-country data does not 
exist on the incomes of software users only, which would be the ideal income metric. 



Kyper	

55 
 

While the benefit of increasing disposable income (and thereby utility) from software piracy 
exists, engaging in software piracy is not costless, and the costs vary across countries. Like any crime, 
there is always the possibility of getting caught and facing civil or criminal consequences. Therefore, 
in those countries with greater enforcement of piracy laws and tougher penalties for piracy the 
expected cost of piracy is higher. In order to account for this, the study includes a measure of 
intellectual property protection across countries, as perceived by global business leaders. 

Just like any type of theft, there are personal and social costs of piracy. Regardless of 
whether or not an individual believes he/she will get caught by the authorities, an individual may 
simply feel personally guilty for stealing software. Those who believe that piracy is the immoral 
stealing of another’s work are likely to place a higher value on the individual contribution of that 
person or firm. In terms of an individualist-collectivist spectrum, these people would lean towards 
the individualist side. On the other hand, those who place little emphasis on the past contributions 
of the original piracy designer and instead place a greater emphasis on the usefulness of that 
software to overall society would lean towards the collectivist side (this is represented by Hofstede’s 
indices). Even beyond the question of one’s own personal viewpoint on the ethics of software 
piracy, that person may also be influenced by the views of his/her peers on whether or not it is 
acceptable to engage in software piracy. Therefore, he/she would be influenced by the overall 
country’s attitude towards individualism-collectivism. The existence of these personal moral and 
social peer-pressure costs of piracy is the theoretical justification for using Hofstede’s individualism-
collectivism index in the analysis. 

Similar to the individualist-collectivist question, a country’s level of corruption may also 
influence a person’s cost-benefit analysis for whether or not to engage in software piracy. If a 
country is perceived by its citizens as having a culture of corruption, there is likely to be lower peer-
pressure costs from engaging in piracy. This is likely to be highly correlated with intellectual property 
protection (IPP), which is a legal issue, but it does add some information. Suppose two countries 
have similar degrees of IPP but one is rated as more corrupt than the other. The individual in the 
corrupt country may feel as though engaging in illegal activity generally comes with less social 
backlash than the individual in the less corrupt country even if the expected legal costs are the same. 

Even though both income and cultural variables theoretically impact software piracy rates 
directly, there is obviously a high degree of correlation between income and the cultural variables 
themselves. The causation likely goes both ways in some cases, such as with corruption and income. 
This issue is discussed in the results section with a collinearity matrix of all the variables used. 

 
Research Questions and Accompanying Hypotheses 

Do alternative income measures better predict national software piracy rates than standard 
GDP per capita? (Appendix B provides the regression models used to calculate R in the equations 
following H1-H3) 

 
Ho1: Mean GDP per capita and square root of mean GDP per capita have the same 
predictive power of national software piracy rates 

i.e. R (GDP per capita) – R (sq. root of mean GDP) = 0 
Ho2: Mean GDP per capita and median household income have the same predictive power 
of national software piracy rates 

i.e. R (GDP per capita) – R (median household income) = 0 
Ho3: Mean GDP per capita and square root of median household income have the same 
predictive power of national software piracy rates 

i.e. R (GDP per capita) – R (sq. root of median household income) = 0 



INCOME AND NATIONAL SOFTWARE PIRACY	

56 
	

 
How does the best alternative measure of income behave in the presence of standard software 
piracy predictors?  

Ho4: Square root of income does not predict software piracy 
Ho5: Corruption does not predict software piracy 
Ho6: Collectivism does not predict software piracy 
Ho7: Intellectual property protection does not predict software piracy 

Hypotheses 4-7 are tested using the following regression model 
Software Piracy = B0 + B1 Corruption Index + B2 IPP Competitiveness Index + B3 Collectivism Index + 
B4 Square Root of Median Household Income + ε      
 

Data Collection / Analysis 
To examine the role of income in understanding national software piracy rates we collected 

data from 2006 to 2013 for the following variables: BSA’s national software piracy rate (dependent 
variable), World Bank (GDP per capita by country PPP adjusted, Hoefstede’s cultural indices 
(collectivism), Transparency International’s corruption index, the World Economic Forum’s Global 
Competitiveness Report (the intellectual property protection component), and median household 
income from Gallup by country (surveyed between 2006 and 2012). 

There are 108 countries for which software piracy data from BSA exists for 2009, which is 
the year of reference. Of those 108 countries, 98 (91%) have data for both median income and GDP 
per capita (PPP basis). Complete data of all the explanatory variables exists for 61 countries (see 
Appendix A). The 37 country reduction in full country coverage is primarily due to a lack of 
coverage in Hoefstede’s cultural indexes. 

The dependent variable is national software piracy rate, collected from the BSA Global 
Software Survey. This variable represents the percentage of pirated software installed out of total 
software installed. Roughly this is calculated as the difference between total software installed 
(demand) and total legal software shipped (supply). The Software Alliance (BSA) is an advocacy 
group for software producers that seeks to promote the policy interests of software producers in 
governments throughout the world. One such area of public policy is their push to combat software 
piracy, which is why they commission this study on software piracy. 

The independent variables used in the study fall into two categories: economic (income) and 
cultural/legal (collectivism, corruption, and intellectual property protection). 

GDP per capita by country is derived using World Bank data on gross domestic product, 
which is then converted to international dollars using purchasing power parity (PPP) rates also 
provided by the World Bank. An international dollar has the same purchasing power over GDP as 
the U.S. dollar has in the United States; this controls for the difference of purchasing power among 
countries. GDP was the most commonly included economic variable in previous studies. 

Median household income data comes from a Gallup report of country-by-country median 
household incomes, the results of which were released in 2013. Gallup surveyed the populations of 
131 countries over a six-year period of 2006 through 2012. Like the study does with GDP, Gallup 
adjusted each country’s income measure using the purchasing power parity rates provided by the 
World Bank. Because the Gallup figures are averages over a six year window from 2006 through 
2012, we chose to use 2009 as the year of record in the analysis. 

Our measure of intellectual property protection (IPP) comes from the World Economic 
Forum’s Global Competitiveness Report, which it describes on its website as follows: 

The Global Competitiveness Report 2014-2015 assesses the competitiveness landscape of 
144 economies, providing insight into the drivers of their productivity and prosperity. The report 



Kyper	

57 
 

remains the most comprehensive assessment of national competitiveness worldwide, providing a 
platform for dialogue between government, business and civil society about the actions required to 
improve economic prosperity. Competitiveness is defined as the set of institutions, policies and 
factors that determine the level of productivity of a country. The level of productivity, in turn, sets 
the level of prosperity that can be earned by an economy. 

The IPP component of the Global Competitiveness Report is part of the institutions sub index of 
the GCR and is derived based on responses of business leaders to the following question in the 
Executive Opinion Survey: 

How would you rate intellectual property protection, including anti-counterfeiting measures, 
in your country? [1 = very weak; 7 = very strong] 

The two cultural variables used in this study were chosen based on the findings of previous 
researchers seeking to explain variation in piracy rates across countries. Many of these appear in the 
literature review. The measure for collectivism comes from Hofstede’s seminal study on cultures 
across nations. As explained by Hofstede (1983): 

“Individualism collectivism index: individualism, which stands for a preference for a loosely 
knit social framework in which individuals are supposed to take care of themselves and their 
immediate families only; as opposed to Collectivism, which stands for a preference for a tightly knit 
social framework in which individuals are emotionally integrated into an extended family, clan, or 
other in-group which will protect them in exchange for unquestioning loyalty. The word 
Collectivism in this sense carries no political connotations and does not assume any positions as to 
the role of the state; it operates at a much smaller scale of social integration (Hofstede, 1983).” 

Merritt (2000) has shown that Hofstede’s cultural indices are replicable across different time 
periods and subject groups. In particular collectivism showed to be durable across time and subjects. 
This provides the current study with the necessary empirical and theoretical foundations to continue 
using Hofstede’s cultural indices. 

Finally, the measure of corruption comes from Transparency International’s Corruption 
Perceptions Index. The study rates countries on perception of public corruption on a scale of 0-10 
where 0 is high corrupt countries and 10 is low corrupt countries. These ratings are based on surveys 
of surveys. That is, Transparency International uses multiple surveys that been conducted by 
organizations such as the World Bank, IHS Global Insight, Freedom House, and others in order to 
come up with a rating for each country. The respondents of these surveys are typically international 
experts and business persons. 

We first address the question of what income measure best predicts software piracy by 
running bivariate OLS regressions for each income measure with the dependent variable for the 98 
countries that have coverage of both income measures and software piracy. After running the four 
bivariate regressions, we then run four additional regressions, each containing a different income 
measure along with all of the non-income explanatory variables for the 61 countries that have full 
coverage. All hypotheses are tested at the 95% significance level (α = 0.05).  

 
Results 

 
In understanding the relationship between software piracy and income we examined 98 

countries using correlation tests. (See Table 2 below). We found that the difference between r for 
median income is statistically significant than r for mean income (alpha = 0.05), thus rejecting H2 
(mean GDP vs. median household income). We find no statistically significant differences within 
means nor medians between standard values and square root values, thus NOT rejecting H1 (mean 
GDP vs. square root of mean GDP) and H3 (mean GDP vs. square root of median household 



INCOME AND NATIONAL SOFTWARE PIRACY	

58 
	

income). However, model fit (R2) improves consistently as we move from mean GDP to the square 
root of median income. This suggests that capturing the diminishing marginal utility of income is an 
important step in understanding software piracy.  

Although H1 and H3 were not rejected using conventional hypothesis testing assuming a 
sample size of 98, recall there are only 108 countries with data on software piracy rates from BSA. 
Therefore, if one considers BSA countries to be the population of interest, we have near-population 
level coverage in this study. There are only 196 independent countries in the world. The 108 
countries selected for inclusion in the BSA study is not a random sample from those 196. The BSA 
study focuses on the largest and richest countries given that BSA member companies would have 
the most at stake in those countries and given that those countries tend to have better data 
availability. Together, the 98 countries included in the bivariate analysis account for over 94% of the 
world’s GDP.  

 
 
 

Table 2.  Pearson Correlation Coefficients and R2 ( ) for different income measures with software 
piracy 

 
Income Measure All countries with data 

coverage of piracy and 
income   (n = 98) 

Countries with full data coverage 
of piracy, income, corruption, 

collectivism, and IPP        (n = 61) 

Mean GDP 
 

0.674 
(0.454) 

0.825 
(0.681) 

Square Root of Mean GDP 0.764 
(0.584) 

0.851 
(0.724) 

Median Household Income 0.836 
(0.699) 

0.847 
(0.717) 

Square Root of Median 
Household Income 

0.851 
(0.724) 

0.873 
(0.762) 

 
 
 
Table 3. Regression results for square root of median household income 
 

Independent Variable Coefficient P-value Hypothesis 
Tested 

Constant 84.553 0.000* N/A 
Corruption Index -0.954 0.453 H5 
IPP  
(Competiveness Index) 

-5.346 0.016* H7 

Collectivism Index 0.213 0.000* H6 
Square Root of  Median HH 
income 

-0.136 0.002* H4 

  
Sample Size 61 
Adj. R2 0.856 
F Statistic 89.891 

 



Kyper	

59 
 

Using the square root of median household income in Table 3 below highlights the multiple 
regression findings which allow us to further reject hypotheses H4 (square root of income), H6 
(collectivism), and H7 (intellectual property protection). It shows that when we use the square root 
of median household income as the income metric along with all other non-income explanatory 
variables, the R2 value is 0.856.  Note that the universe of countries included is smaller because of 
the lack of coverage for certain non-income variables. All of the variables are significant except for 
corruption. Corruption (H5) is not significant largely because of collinearity with income as shown 
and discussed below in Table 4. 

In Table 4 we present results for four different regressions with each including a different 
income measure. The universe of countries included in these regressions is the 61 countries with full 
coverage of all the variables. These 61 countries make up 90% of world GDP. Regressions #1 and 
#2 use mean GDP per capita and its square root respectively. Regressions #3 and #4 use median 
household income and its square root respectively.  

Regression #1 in the table shows that mean GDP per capita is not significant when the 
other explanatory variables are included. The other income variables are significant in their 
respective regressions. The relationship between mean GDP per capita and software piracy is simply 
not strong enough to overcome the effect of collinearity on statistical significance. That is not the 
case for the other income measures, which as we saw earlier have stronger relationships with 
software piracy than mean GDP per capita. We again find that in all regressions, corruption is not 
statistically significant. Although intellectual property protection (IPP) and collectivism are also 
correlated with the other explanatory variables, both of these variables were significant in each 
regression at a 95% confidence level.  

 
Table 4. Coefficients and P-values ( ) from Regressions Predicting BSA Piracy Rates across 

Countries in 2009, by Income Metric Used 
 

Independent Variable Regression #1 
Mean GDP 

Regression #2 
!"#$!"# 

Regression #3 
Median Income 

Regression #4 
!"#$%&!"#$%& 

Constant 76.928 
(0.000) 

85.421 
(0.000) 

76.465 
(0.000) 

84.553 
(0.000) 

Corruption Index -1.962 
(0.137) 

-1.445 
(0.255) 

-1.512 
(0.255) 

-0.954 
(0.453) 

IPP  
(Competiveness Index) 

-5.409 
(0.025)* 

-5.177 
(0.024)* 

-5.541 
(0.018)* 

-5.346 
(0.016)* 

Collectivism Index 0.2500 
(0.000)* 

0.231 
(0.000)* 

0.234 
(0.000)* 

0.213 
(0.000)* 

GDP per capita ($1,000) -0.254 
(0.087) 

   

Square Root of 
GDP per capita 

 
 

-0.117 
(0.006)* 

  

Median HH income ($1,000)  
 

 -0.351 
(0.027)* 

 

Square Root of Median HH 
income 

 
 

  -0.136 
(0.002)* 

     
Sample Size 61 61 61 61 
Adj. R2 0.836 0.849 0.842 0.856 
F Statistic 77.334 85.241 80.675 89.891 

 



INCOME AND NATIONAL SOFTWARE PIRACY	

60 
	

While there is strong collinearity between each of the non-income explanatory variables and 
the four income measures, collinearity is strongest between corruption and the income measures. 
The correlation coefficients shown in Table 5 explain why corruption is not a significant predictor 
of software piracy in the presence of an income variable. Table 5 also explains why the R2 values 
only increase modestly when the non-income variables are included compared to the simple income-
piracy regression.  

One of the key takeaways from the analysis is that the high degree of statistical collinearity 
between the theoretical determinants of software piracy make it difficult to assess the individual 
magnitude of each determinant. For example, by itself, higher income leads to lower piracy. But the 
statistically strong relationship between income and piracy is, in part, masking the ability of 
corruption (and other variables) to explain piracy differentials across countries. 

 
Table 5. Correlation coefficients between income and other explanatory variables  

 
 Corruption IPP Collectivism 

GDP per Capita 0.856 0.844 -0.627 

Square Root of GDP 
per capita 0.850 0.830 -0.647 

Median HH Income 0.866 0.839 -0.659 

Square Root of 
Median HH income 0.866 0.835 -0.674 

n = 61 (full coverage countries only) 
Note: All correlation coefficients are statistically significant at 95% 
confidence level. 

 
Sensitivity Analysis 

Our findings are robust under various alternative specifications. Specifically, we ran 
regressions for years other than 2009 to verify that the research conclusions do not change based on 
the year chosen. In addition, we performed listwise regressions for year 2009 and concluded that are 
results do not change. The listwise regression included all 91 countries instead of just the 61 
countries that have full coverage of all of the explanatory variables. Under listwise, for those 
countries that are missing one of the explanatory variables, the regression ignores that variable but 
includes the information from the variables for which data is present. We do not replace the missing 
data with the variable’s mean value. We simply treated missing data as missing. Results under these 
alternative specifications are available to readers upon request from the authors. 

 
Conclusion 

 
The role of income in predicting national phenomena such as software piracy has been 

frequently studied. Average per capita GDP is an understandably common measure of individual 
income. However, failure to recognize the diminishing marginal utility of income can be an 
egregious error when comparing countries with dramatically different income levels. This creates the 
potential for the misinterpretation of the role of income. Furthermore, because income is highly 
skewed within countries, using median in place of the average has significant advantages. The 



Kyper	

61 
 

findings show applying the square root of median income captures the diminishing role of income 
and provides the best income measure to use in software piracy studies. 

Whatever income measure is chosen, the study shows that collinearity between income and 
other variables is a major issue in trying to assess the predictive power of non-income variables on 
software piracy rates. The problem stems from two competing realities: (1) Income should 
theoretically be part of any software piracy model, and (2) Income is highly correlated with other 
variables – such as intellectual property protection, corruption, and collectivism – that should 
theoretically be part of any software piracy model. Ideally these findings will inform future studies 
using income to predict national phenomena.  

 
References 

 
Bagchi, K., Kirs, P., & Cerveny, R. (2006). Global software piracy: Can economic factors alone 

explain the trend? Communications of the ACM, 49(6), 70-76.  
BSA. (2003). Expanding global economies: The benefits of reducing software privacy. Business Software 

Alliance. Retrieved, from the World Wide Web: 
http://global.bsa.org/idcstudy/pdfs/White_Paper.pdf 

BSA. (2014). The Compliance Gap: BSA Global Software Survey, 2013. Business Software Alliance. 
Retrieved, from the World Wide Web: 
http://globalstudy.bsa.org/2013/downloads/studies/2013GlobalSurvey_Study_en.pdf  

Chavarria, J., & Morrison, R. (2014). Revisiting Software Piracy Using Globe Cultural Practices. 
Academy of Information and Management Sciences Journal, 17(1), 101-121.  

Chen, C., Chen, C., & Yeh, C. (2010). Determinants of Software Piracy: Evidence from Far East 
Countries. Journal of International and Global Economic Studies, 3(2), 53-62.  

Cheng, H., Sims, R., & Teegen, H. (1997). To Purchase or to Pirate Software: An Empirical Study. 
Journal of Management Information Systems, 13(4), 49-60.  

Cobb, C., & Douglas, P. (1928). A Theory of Production. American Economic Review 18 (Supplement), 
139-165. 

Dewan, S., & Kraemer, K. (2000). Information Technology and Productivity: Evidence from 
Country-Level Data. Management Science, 46(4), 548-562.  

Hofstede, G. (1983). The cultural relativity of organizational practices and theories. Journal of 
International Business Studies, 14(2), 75-89. 

Gallup (2006-2012). Worldwide median household income. Retrieved from 
http://www.gallup.com/poll/166211/worldwide-median-household-income-000.aspx  

Goel, R., & Nelson, M. (2012). Shadow economy and international software piracy. Applied Financial 
Economics, 22, 1951-1959.  

Gopal, R., & Sanders, G. (2000). Global Software Piracy: You Can't Get Blood Out of a Turnip. 
Communications of the ACM, 43(9), 83-89.  

Gopal, R., & Sanders, G. (1998). International Software Piracy: Analysis of Key Issues and Impacts. 
Information Systems Research, 9(4), 380-397.  

Hamister, J., & Braunscheidel, M. (2013). Software Piracy and Intellectual Property Rights 
Protection. Academy of Information and Management Sciences Journal, 16(1), 15-35.  

Harbi, S., Grolleau, G., & Bekir, I. (2012). Is There A Piracy Kuznets Curve? Review of Law & 
Economics, 8(2), 433-456.  

Husted, B. (2000). The Impact of National Culture on Software Piracy. Journal of Business Ethics, 
26(3), 197-211.  



INCOME AND NATIONAL SOFTWARE PIRACY	

62 
	

Kariithi, N. (2011). Is the Devil in the Data? A Literature Review of Piracy Around the World. The 
Journal of World Intellectual Property, 14(2), 133-154.  

Kovacic, Z. (2007). Determinants of Worldwide Software Piracy. 127-150.  
Marron, D., & Steel, D. (2000). Which countries protect intellectual property? The case of software 

piracy. Economic Inquiry, 38(2), 159-174. 
Merritt, A. (2000). Culture in the cockpit: Do Hofstede’s measures replicate? Journal of Cross-Cultural 

Psychology, 31, 283-301. 
Olkin, I. and Finn, J.D. (1995). Correlations Redux. Psychological Bulletin, 118(1), pp. 155-164. 
Swinyard, W., Rinne, H., & Kau, A. (1990). The Morality of Software Piracy: A Cross-Cultural 

Analysis. Journal of Business Ethics, 9, 655-664.  
The World Bank, World Development Indicators (2006-2013). GDP per capita (current US$). [Data 

file]. Retrieved from http://data.worldbank.org/indicator/NY.GDP.PCAP.CD  
The World Economic Forum, Competitiveness Rankings (2006-2013). Retrieved from 

http://reports.weforum.org/global-competitiveness-report-2014-2015/rankings  
Thong, J., & Yap, C. (1998). Testing an Ethical Decision-Making Theory: The Case of Softlifting. 

Journal of Management Information Systems, 15(1), 213-237.  
Transparency International, (2006-2013). Corruption perceptions index. Retrieved from the World 

Wide Web: http://www.transparency.org/cpi2013/results  
Yang, D., Sonmez, M., Bosworth, D., & Fryxell, G. (2009). Global Software Piracy: Searching for 

Further Explanations. Journal of Business Ethics, 87, 269-283.  
 

 
Eric Kyper,	Gerald Prante, and Lee Schimmoeller work at Lynchburg College.  
  



Kyper	

63 
 

Appendix A 
 
List of countries included in multiple regression analyses 
Australia Guatemala Peru 

Austria Hong Kong Poland 

Bangladesh Hungary Portugal 

Belgium India Romania 

Brazil Indonesia Russia 

Bulgaria Ireland Serbia 

Canada Israel Singapore 

Chile Italy Slovakia 

China Japan Slovenia 

Colombia Latvia South Africa 

Costa Rica Lithuania South Korea 

Croatia Malaysia Spain 

Czech Republic Malta Sweden 

Denmark Mexico Thailand 

El Salvador Morocco Turkey 

Estonia Netherlands United Kingdom 

Finland New Zealand United States 

France Norway Uruguay 

Germany Pakistan Venezuela 

Greece Panama Vietnam 

 
 
 
Additional countries included in the bivariate regressions of income and software piracy 

Albania Egypt Nicaragua 

Algeria Georgia Nigeria 

Armenia Honduras Paraguay 

Azerbaijan Iraq Qatar 

Bahrain Jordan Saudi Arabia 

Belarus Kazakhstan Senegal 

Bolivia Kenya Sri Lanka 
Bosnia and 
Herzegovina Kuwait Tunisia 

Botswana Lebanon Ukraine 

Cameroon Libya Yemen 

Cyprus Luxembourg Zambia 
Dominican 
Republic Moldova  

Ecuador Montenegro   
 



INCOME AND NATIONAL SOFTWARE PIRACY	

64 
	

The following 10 countries have BSA piracy rates but are lacking data for either GDP per capita or 
median household income: 

Argentina Mauritius Taiwan 

Brunei Oman UAE 

Iceland Puerto Rico  

Ivory Coast Switzerland  
 
  



Kyper	

65 
 

Appendix B 
 
Do alternative income measures better predict national software piracy rates than standard GDP per 
capita? 
  
Ho1: Mean GDP per capita and square root of mean GDP per capita have the same predictive 
power of national software piracy rates 
  
(Regression #1):            Software Piracy = B0 + B1 Corruption Index + B2 IPP Competitiveness Index + B3 

Collectivism Index + B4 GDP per capita + ε   
  
vs. 
  
(Regression #2):            Software Piracy = B0 + B1 Corruption Index + B2 IPP Competitiveness Index + B3 

Collectivism Index + B4 Square Root of GDP per capita + ε 
  
  
  
Ho2: Mean GDP per capita and median household income have the same predictive power of 
national software piracy rates 
  
(Regression #1):             Software Piracy = B0 + B1 Corruption Index + B2 IPP Competitiveness Index + B3 

Collectivism Index + B4 GDP per capita + ε 
  
vs. 
  
(Regression #3):             Software Piracy = B0 + B1 Corruption Index + B2 IPP Competitiveness Index + B3 

Collectivism Index + B4 Median Household Income + ε 
  
  
  
Ho3: Mean GDP per capita and square root of median household income have the same predictive 
power of national software piracy rates 
  
(Regression #1):             Software Piracy = B0 + B1 Corruption Index + B2 IPP Competitiveness Index + B3 

Collectivism Index + B4 GDP per capita + ε 
  
vs. 
  
(Regression #4):             Software Piracy = B0 + B1 Corruption Index + B2 IPP Competitiveness Index + B3 

Collectivism Index + B4 Square Root of Median Household Income + ε 
 

	


