The Relationship of Economic
Policies, Human Capital and
Gender with Entrepreneurship:
An Empirical Approach
MSc thesis International Economics and Business
Segura Soldevilla, Nicolás
Abstract
The aim of this study is to analyse the relationship between entrepreneurship and the burden of taxes and corruption, together with human capital and gender variables. Using panel data from around 130 countries from the Quality of Government (QoG) Standard Data between 2004 and 2013, we estimate ordinary least squares regressions with fixed effects to infer the impact of the main independent variables of interest and other covariates on entrepreneurship. The results suggest that lower taxes benefit new businesses creation. Furthermore, a stronger anticorruption policy positively leads to higher entrepreneurial activity. Evidence shows that higher rates of women in the labour force and a greater number of schooling years, both secondary and tertiary, also report a positive relationship with the outcome. Nevertheless, some minor differences are observed in the sensitivity analysis.
Keywords: entrepreneurship, economic policy, burden of taxes, corruption,
CONTENTS
List of tables ……….….…….. 3
List of figures ……….….… 3
1. Introduction ……….……….…... 4
2. Hypotheses ………..……….... 6
3. Data and estimation technique …...……….………...… 10
3.1 Study sample ……….……….….…….. 10
3.2 Variables description ….……….……….………… 10
3.2.1 Outcome variable: entrepreneurship ……….……….…. 10
3.2.2 Independent variables ……….……….………. 11
3.2.3 Control variables ……….……….…………. 12
3.3 Methodology framework ……….………….………... 12
3.3.1 Bivariate analysis between the dependent and independent variables .………. 13
List of tables
Table 1: Descriptive statistics, 2004 – 2013 ………..…. 17
Table 2: Linear regression results. Fixed effects ………..……….... 18
Table 3: Hausman test for random effects vs fixed effects estimation …………...…………. 19
Table 4: Linear regression results with the interaction term between fiscal freedom and anticorruption policies. Fixed effects ………..……….22
Table 5: Linear regression results with additional control variables. Fixed effects ... 23
Appendix B: List of countries included in the analysis ……….... 35
Appendix C: Correlation matrix ………. 36
Appendix E: Linear regression. Random effects ………..……… 38
Appendix F: Testing multicollinearity: variance inflation factors (VIF) ……… 39
List of figures Figure 1: Histogram of new business density and its logarithmic form……….………… 11
Figure 2: Scatter plots of the dependent variable and the independent variables of interest ………..……… 14
1. INTRODUCTION
To the best of our knowledge, no previous literature has tried to analyse the relation-ship between the economic policy, human capital and gender with entrepreneurrelation-ship using an explanatory set of variables included at the same time in the regression analy-sis. Hence, we think that the current analysis is of great relevance since we bring togeth-er these three groups of indicators: economic policy, human capital and gendtogeth-er. We be-lieve that this work could be useful for entrepreneurs, who could be wondering where to establish their own companies, and policy makers, who need to rule with the aim of achieving the common wealth, and high entrepreneurial rates are a good symptom of prosperity. Therefore, the fact that the effect of these determinants on entrepreneurship has been widely analysed as single factors but not jointly highlights the importance of the present study.
The literature shows that a higher number of years of education leads to higher skilled jobs (Sluis et al., 2005 and 2008). Taking into account the level of managerial knowledge required, running a new business should be considered a highly skilled job. Moreover, according to the existing analyses, women seem to be less prone to establish a new business by themselves, probably due to their lack of expertise (Fischer et al., 1993), their lower interest in doing so (Kourilsky and Walstad, 1998) or their fear to failure (Noguera et al., 2013).
In this study, we use country level data to explore how taxes and corruption affect entrepreneurship, as well as the impact of human capital and gender indicators, which are barely analysed when assessing the entrepreneurial activity in any country. We do so by using data from more than 130 countries with different characteristics in terms of the variables included in our analysis and looking at a recent time period, controlling for potential different trends between countries and time framework.
2. HYPOTHESES
In order to establish the hypotheses of our study, we firstly conducted a literature review, following the search strategy found in Appendix A.
We expect positive results in terms of entrepreneurial rates in countries with lower corporate taxes. The classical Economic Theory has asserted that the lower the constraints for companies, such as taxes, the better the companies will do and thus, the more efficiently resources will be used. For making profit, entrepreneurs firstly need to pay taxes set by politicians in each economic jurisdiction. Therefore, it seems plausible that entrepreneurs will sometimes have the possibility to move their companies or future companies to other countries to develop their activities in low-taxes economies. This will, of course, have a different effect on each individual according to the diverse elasticity of each entrepreneur in a change in corporate taxes.
Thus, we expect:
Hypothesis 1: The higher the burden of taxes, the lower the number of new businesses will be set up.
Nowadays, corruption is still present in many parts of the world. Corruption includes activities such as bribery, theft, fraud, embezzling, extortion or black mailing among others, and it is thus expected to negatively affect new businesses creation. One of the aims of the present study is to empirically evaluate if this is true and to which extent.
In this line, the literature has already examined how corruption affects entrepreneurship. When including different kinds of taxes with corruption, corporate taxes are no longer significant for the entrepreneurial activity when corruption is low, as Chowdhury et al. (2015) highlighted in their model where they add corruption, time and documents to export to different kinds of taxes in order to explain the percentage of total early stage businesses with a heterogeneous sample of 48 countries. Matching data from different sources, they conclude that corruption seems to be a double-edge sword, having both advantages and caveats for international entrepreneurship. Therefore, corruption could also worsen the burden of regulations, especially the ones with financial cost elements.
According to this two-faced effect of corruption, Bologna and Ross (2015) use data from Brazil municipalities to assess how corruption decreases the entrepreneurial activity, especially in the long run. Nevertheless, it also increases the number of new businesses when the institutional quality is poor. Despite the fact that their model is based on how corruption and the institutional environment affect the total number of establishments per industry in a municipality, they do not identify the net number of new companies per year.
institutions channel effort into productive entrepreneurship.
It seems pretty clear that corruption is expected to reduce the odds of entrepreneurial activity, as business people are less confident in institutions. Therefore, according to the literature on the topic, we expect:
Hypothesis 2: The higher the degree of corruption, the lower the number of enterprises will be set up.
Human capital variables such as the educational system quality and the average schooling years are usually positively associated with the entrepreneurial activity. People with higher educational level are better trained and, thus, more likely to achieve highly skilled jobs. It goes without saying that running your own company is expected to be a high-skill job.
Justin van der Sluis et al. (2008) demonstrate the positive impact that education has on people to set up their own companies. In addition, as they empirically show, this effect is positive and significant but the effect of education on earnings is smaller for entrepreneurs than for employees. Moreover, Justin van der Sluis et al. (2005) present a model to explain the impact of education in developing economies. They assert that the higher the number of schooling years in developing economies, the higher the number of enterprises are, on average, set up. They conclude that the most educated workers typically achieve highly remunerated jobs and prefer non-farm entrepreneurship to farming.
Therefore, we can say that education increases the odds of being an entrepreneur and the income of such companies. Robinson and Sexton (1994) modelled entrepreneurship outcomes using education and working experience, controlling for farmers and professionals. They reach similar conclusions to Justin van der Sluis et al. in 2008 as they demonstrate the positive association between education and entrepreneurship in terms of self-employment and success. Given the results found in the literature and the expected positive relation that education has on setting up a new company, we propose our third hypothesis:
Hypothesis 3: A higher number of schooling years of the population of a country will have a positive effect on the entrepreneurial rates of that country.
the ones men get. Women also play a key role in this aspect, since they sometimes prefer to give up some of their working time to look after their families (Alvarez et al., 2012).
Many authors have already dealt with this in the past, Alvarez et al. (2012) argue that women are in clear disadvantage when taking social networks and resources into account, reaffirming what Robinson and Stubberud (2009) previously supported. While women’s source of advice are usually their relatives and friends, men normally receive guidance from professional acquaintances and consultants, who are likely to be better trained. These two distinctions affect the way each gender performs. As it was also analysed by Noguera et al. (2013) and Alvarez et al. (2012), the perception women have on their own capabilities also plays a key role in their attitude towards setting up their own business. Additionally, Kourilsky and Walstad (1998) analyse the differences in knowledge between men and women on entrepreneurship. Not only were women found to know less on the topic, but they also showed a lower degree of interest in setting up their own companies. Likewise, Fischer et al. (1993) show the lower degree of expertise of women managing employees or helping start-ups. Attributes like company size, income growth and sales are also found to be smaller for women. This could be explained by the lower experience women have in running and working in firms. Taking all this into account, we propose the fourth and last hypothesis:
3. DATA AND ESTIMATION TECHNIQUE 3.1 Study sample
We have used the Quality of the Government (QoG) Standard Data, which includes around 2500 variables from merging more than 100 data sources, such as FAO Statistics or Eurostat Datasets. It has data from countries that are current or previous members of the United Nations which it is provided by the QoG Institute, an independent research organization within the Department of Political Science at the University of Gothenburg. The QoG Standard dataset offers information on a wide range of variables, such as quality of the government, conflicts, education, labour market, health and welfare. Both cross-sectional and time series (TS) databases are available, but we will work with the TS one, being the unit of analysis country-year.
The selected independent and control variables are guided by the literature. Although our principal aim was to work with a time period analysis of at least 15 years, data availability constrains of our chosen variables reduced the time period of our analysis to 2004-2013 for the 133 countries included in our analysis. The sample of countries is heterogeneous in terms of development and inequality. While there are undeveloped countries such as Sierra Leone, rich countries like The Netherlands are also present in the regressions. A complete list of the countries contained in the sample can be found in appendix B.
3.2 Variables description
3.2.1. Outcome variable: entrepreneurship
et al., 2004), we used the natural logarithm of the new business density as our
dependent variable.
Figure 1: Histogram of new business density and the logarithm of new business density
3.2.2 Independent variables
In order to measure the burden of taxes, we use a variable that assesses the fiscal freedom of the countries in the dataset. Fiscal freedom includes the top tax rate on individual income, the top tax rate on corporate income and the total tax revenue as a percentage of GDP, each of them being equally weighted as one-third of the final variable. The country’s fiscal freedom ranges between 0 and 100, where 100 represents the maximum degree of fiscal freedom.
Since we are also interested in the association between corruption (or freedom from corruption) and entrepreneurship, we include the anticorruption policy variable available in our dataset. This variable is categorical, ranges from 0 to 10 and measures the extent to which governments successfully contain corruption. Value 0 refers to no control whatsoever, whereas value 10 denotes success in containing corruption and placement of the effective mechanisms.
Another issue regarding our research interest is how entrepreneurship is affected by different women rates in the labour force. Thus, we select the ratio of women to men in labour force.
3.2.3. Control variables
Guided by the literature, GDP growth in percentage is used as control variable for the main regressions of the present work.
Furthermore, as part of the sensitivity analysis, we include new control variables one by one and altogether in our regressions guided by the previous literature on the topic. These variables are: annual population growth in percentage, annual population growth squared also in percentage, days required to start a business, unemployment rate and the number of changes in government per year.
3.3 Methodology framework
In this work we use panel data observations of 131-133 countries (depending on the regression) from 2004 to 2013. In the case of some variables, we had to extrapolate the data from previous years until the next year where the data was available. This was done for the variables secondary and tertiary schooling years since the dataset only had this data every five years.
Firstly, fixed and random effects are obtained for the sample and then a Hausman test is done to check for the best way to deal with the available sample. There are several differences between fixed and random effects, each of them have different assumptions and they treat the data in a different way.
Nevertheless, the assumption of the random effects that the random error cannot be correlated with any of the explanatory variables is actually rejected. When that is the case, the parameters of the random effects become biased and inconsistent. We assume that entrepreneurship across countries depends on a set of economic policies, human capital and gender variables, together with GDP growth as a control variable.
3.3.1 Bivariate analysis between the dependent and independent variables
We present five bivariate scatter plots of the different independent variables with the natural logarithm of new business density per 1000 people. What can be appreciated from the graphs below (Figure 2) and from the correlations matrix (appendix C) is that all of the independent variables are positively associated with the outcome, except fiscal freedom, with anticorruption policies and the average years of secondary schooling being the strongest relations with a correlation of 0.64 and 0.62, respectively.
Some outliers are found in anticorruption policies and in both of the human capital variables. Nevertheless, these outliers are more clearly seen in the box graphs available in appendix D. In the case of anticorruption policies, the outliers appear in the middle values when the anticorruption policies are moderate. Extreme values can be also seen in the higher years of secondary education and in the lower number of tertiary schooling years. We also performed univariate regressions to test for the individual significance of each independent variable with respect to entrepreneurship. Results will be explained later in this document.
3.3.2 Multivariate analysis
According to our assumption, to assess the independent explanatory power of our variables of interest, we model a first regression with one of the economic policies parameters, fiscal freedom (Model A). We then add the second variable related to such category, anticorruption policy (Model B), and our control variable, GDP growth, as it is also an economic indicator in a later model (Model C).
Figure 2. Scatter plots of the dependent variable and each independent variable of interest
Finally, we model the natural logarithm of new business density with all our explanatory parameters as follows:
ln(𝑁𝐵𝐷)𝑖𝑡 = 𝛽0+ 𝛽1𝑓𝑖𝑠𝑐𝑎𝑙𝑓𝑟𝑒𝑒𝑑𝑜𝑚𝑖𝑡+ 𝛽2𝑎𝑛𝑡𝑖 − 𝑐𝑜𝑟𝑟𝑢𝑝𝑡𝑖𝑜𝑛𝑖𝑡+ 𝛽3𝐺𝐷𝑃𝑔𝑟𝑜𝑤𝑡ℎ𝑖𝑡
+ 𝛽4𝑤𝑜𝑚𝑒𝑛𝑖𝑡 + 𝛽5𝑠𝑒𝑐𝑜𝑛𝑑𝑎𝑟𝑦𝑠𝑐ℎ𝑜𝑜𝑙𝑖𝑡+ 𝛽6𝑡𝑒𝑟𝑡𝑖𝑎𝑟𝑦𝑠𝑐ℎ𝑜𝑜𝑙𝑖𝑡+ 𝜇𝑖 + 𝜔𝑡 + 𝜀𝑖𝑡
ln(𝑁𝐵𝐷)𝑖𝑡 is the logarithm of new business density in country i and time t, 𝑓𝑖𝑠𝑐𝑎𝑙𝑓𝑟𝑒𝑒𝑑𝑜𝑚𝑖𝑡 refers to the existing fiscal freedom in country i and time t, 𝑎𝑛𝑡𝑖 − 𝑐𝑜𝑟𝑟𝑢𝑝𝑡𝑖𝑜𝑛𝑖𝑡 denotes the corruption policies in country i and time t,
𝐺𝐷𝑃𝑔𝑟𝑜𝑤𝑡ℎ𝑖𝑡 is included as a control variable, 𝑤𝑜𝑚𝑒𝑛𝑖𝑡 refers to women in labour force, ratio to men in country i and time t, and 𝑠𝑒𝑐𝑜𝑛𝑑𝑎𝑟𝑦𝑠𝑐ℎ𝑜𝑜𝑙𝑖𝑡 and 𝑡𝑒𝑟𝑡𝑖𝑎𝑟𝑦𝑠𝑐ℎ𝑜𝑜𝑙𝑖𝑡 are average years of secondary and tertiary education, respectively, in country i and time t. 𝜇𝑖 denotes country-specific fixed-effects and 𝜔𝑡includes time dummies for each year within the sample to control for time-fixed effects. 𝜀𝑖𝑡 is the error term, which is independent and identically distributed with mean 0 and variance σε2.
3.3.3 Sensitivity analysis
In the sensitivity analysis, as Chowdhury et al. (2015) reported that the effect of taxes disappears when corruption and taxes are included together in the regression model, we also add an interaction term between fiscal freedom and anticorruption policy to test whether that assumption holds in our sample.
4. EMPIRICAL RESULTS 4.1 Descriptive statistics
Our first analysis of the study has been a univariate analysis which is included in Table 1 together with the descriptive statistics. All the independent and control variables are statistically significant at univariate level.
Concerning the descriptive statistics, the mean of new businesses creation per 1000 people within 14 and 64 years is 3.11 for the overall sample, the minimum (0.002) and maximum (44.13) show that there is a huge difference between the country with highest entrepreneurial rates, and the one with the lowest number of companies set up per person. The economic policy variable “fiscal freedom” reports a mean for the sample of 74.31 out of 100, which shows that the average of the countries in the sample have high levels of fiscal freedom, in other words, low top tax rates on individual and company income and a low total tax revenue as a percentage of GDP. The minimum level of fiscal freedom in the sample is 32 while there is a country close to the maximum feasible fiscal freedom quote (99.92). Concerning the anticorruption policy, the mean for the overall sample is close to the middle of the scale (5.12), but there is at least one country with no control whatsoever on corruption, and at least another with the maximum possible control, being the minimum and maximum values 0 and 10, respectively.
Regarding the women in labour force ratio, in the sample there is a higher percentage of men working as the number is below 1 (0.73), with a country in the sample having almost 5 men working for each woman (0.21). However, there is a country where there are more women working than men (1.03). The data also report much higher average of secondary schooling (3.24) than tertiary (0.56).
4.2 Regression results
Table 1: Descriptive statistics, 2004 - 2013
Variable Unit obs. No. Mean SD Min Max Univariate analysis
Dependent variable
New business density Density 1091 3.11 4.41 0.002 44.13 New business density Ln log 1091 0.22 1.61 -6.16 3.79
Economic policy variables Fiscal freedom 0 - 100 1091 74.31 12.95 32 99.92 *** Anticorruption policies 1 - 10 1077 5.12 1.92 0 10 *** Gender variable Women in labour
force, ratio to men 0 - 1 1069 0.73 0.17 0.21 1.03 ***
Human capital variables
Average years of
secondary schooling Number of years 1091 3.24 1.43 0.02 6.90 *** Average years of
tertiary schooling Number of years 1091 0.56 0.35 0 1.76 ***
Control variables
GDP growth % 1084 4.13 4.73 -46.08 54.16 *
Additional control variables
Population growth % 1081 64.04 6.40 47.63 85.96 ***
Days required to start
a business Number of days 1091 40.93 105.90 0.5 1596.88 *** Unemployment rate 0 - 1 1091 0.08 0.27 0 1 ***
Changes in
government per year
Number of
changes 1089 0.51 0.62 0 3
Note: Table 1 reports summary statistics for 10 years and more than 130 countries measured annually.
The Table reports the univariate significance of each independent variable with regards to the dependent variable. *** p<0.01, ** p<0.05, * p<0.1.
These results support hypothesis 1 and are in line with the classical economic theory showing that a higher degree of fiscal freedom policies can lead to more companies being set up.
Table 2: Linear regression results with the logarithm of new business density as the dependent variable. Fixed effects VARIABLES M1 M2 M3 M4 M5 M6 Fiscal freedom 0.0136*** 0.0149*** 0.0143*** 0.00903*** (0.00258) (0.00252) (0.00251) (0.00234) Anti-corruption policies 0.0286*** 0.0326*** 0.0299*** (0.00962) (0.00958) (0.00878) GDP growth, in % 0.00641** 0.00607** (0.00297) (0.00271) Women in labour force,
ratio to men 0.416(0.101) *** 0.441(0.0988) *** 0.470(0.0958) *** Average years of secondary
schooling
0.0605*** 0.0539***
(0.0127) (0.0128) Average years of tertiary
schooling
0.0932** 0.0888**
(0.0452) (0.0452) Observations 1,091 1,077 1,070 1,069 1,069 1,049 Number of countries 133 133 132 132 132 131 Country FE YES YES YES YES YES YES Year effects YES YES YES YES YES YES F-value 27.74 21.96 15.83 16.93 23.48 16.77 Prob > F 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
The women in the labour force ratio always shows a significant and positive effect, and it increases with the inclusion of more variables. These results not only contradict hypothesis number 4, but they also show that the higher the number of women in the labour force, the more companies are set up.
Both education variables show similar positive results with the outcome. A greater number of years of secondary or tertiary education in the population have positive results in entrepreneurial rates. An increase of one year of the average of secondary schooling in the population leads to a 0.054% increase in new company establishments per person in the last model, while in the case of the tertiary education this number increases to 0.089%.
4.3 Econometric tests
Firstly, to allow heterogeneity across individuals and to take advantage of the main benefit of panel data, a Hausman test is used to reject the null hypothesis of using a random effects model, whose results can be found in appendix E. This test allows to determine whether the error component 𝜀𝑖𝑡 is correlated with the regressors in a random effects model.
Table 3: Hausman test for random effects vs fixed effects estimation
chi2 68.07
Prob > chi2 0.0000
Looking at table 3, the p-value is 0.000, lower than our threshold p-value of 0.05 and, thus, we have evidence to think that the random effect model can be affected by an incorrect specification. In conclusion, a fixed effects model is more appropriate.
All the models fit the data well as the p-value of all of them is 0 (table 2), showing that the variables are correctly selected and are good determinants of our outcome variable.
4.4 Sensitivity analysis
In order to check the robustness of our results, we perform two more analyses using fixed effects. The first one includes the same variables and models than our principal specification, but with the inclusion of the interaction between fiscal freedom and anticorruption policies in models 2, 3 and 6 (table 4), where these two variables are separately included. In the second robustness check, more control variables one by one and altogether are added to model 6 (table 5) according to additional controls considered in the literature. Every model in both sensitivity tables is valid and fits the model well since the p-value of every model is always 0.000.
The inclusion of the interaction between fiscal freedom and anticorruption policies has different effects on each of both variables in model 2 if we compare them to the main analysis (Table 2). This interaction, which negatively affects new businesses creation in a significant way, increases the effect of fiscal freedom from 0. 0149 to 0.0208 comparing the model 2 of table 2 and table 3. The opposite happens to anticorruption policies, which reduces its effect on the outcome. Moreover, the inclusion of GDP growth (model 3) does not change much the numbers of the two explanatory variables, although it makes the interaction lose its significance.
Nevertheless, the inclusion of all the variables together with the interaction of the economic policy variables shows interesting results. Although none of the variables change its sign, not only does the interaction lose its significance like in model 3, but anticorruption policies also becomes insignificant. This can stem from the fact that when the interaction variable is included, the single variable losses part of its explanatory power.
Some interesting results are obtained when we control for the number of changes in the government per year. In this model, not only does anticorruption policies lose its significance, but it changes its sign. Apart from this, the human capital variables also lose their significance and even the variable secondary years of schooling turns negative. Similar results to the ones of the last model appear when all the control variables are integrated in the analysis in table 4.
One of the reasons of the inclusion of the interaction term between fiscal freedom and anticorruption policies in the sensitivity analysis was to test what Chowdhury et al determined in 2015. They reported that the effect of taxes disappears when corruption and taxes are included together in the regression model. However, our results show he opposite, it is anticorruption policies the variable which becomes insignificant when the interaction term is integrated in the analysis. These opposite results obtained may differ to the ones of Chowdhury et al (2015) due to variable or sample selection differences.
Table 4: Linear regression results with the logarithm of new business density as the dependent variable and the interaction term between fiscal freedom and anticorruption policies. Fixed effects
VARIABLES M1 M2 M3 M4 M5 M6 Fiscal freedom 0.0136*** 0.0208*** 0.0194*** 0.0110*** (0.00258) (0.00434) (0.00435) (0.00409) Anti-corruption policies 0.114** 0.106** 0.0573 (0.0522) (0.0521) (0.0482) Fiscal freedom # anticorruption policies -0.00113* -0.000976 -0.000362 (0.000680) (0.000679) (0.000627) GDP growth, in % 0.00650** 0.00612** (0.00297) (0.00272) Women in labour force,
ratio to men
0.416*** 0.441*** 0.472***
(0.101) (0.0988) (0.0959) Average years of secondary
schooling
0.0605*** 0.0536***
(0.0127) (0.0128) Average years of tertiary
schooling
0.0932** 0.0907**
(0.0452) (0.0453) Observations 1,091 1,077 1,070 1,069 1,069 1,049 Number of countries 133 133 132 132 132 131 Country FE YES YES YES YES YES YES Year FE YES YES YES YES YES YES F-value 27.74 15.59 12.40 16.93 23.48 14.41 Prob > F 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
Table 5: Linear regression results with the logarithm of new business density as the dependent variable and the inclusion of additional control variables. Fixed effects
VARIABLES population M6 with growth M6 with days required to start a business M6 with unemployment M6 with stability of government M6 with all controls included Fiscal freedom 0.00880*** 0.00959*** 0.00810*** 0.0110*** 0.0103*** (0.00231) (0.00235) (0.00223) (0.00371) (0.00350) Anti-corruption policies 0.0261*** 0.0307*** 0.0323*** -0.0122 -0.00674 (0.00856) (0.00876) (0.00837) (0.0108) (0.00973) GDP growth, in % 0.00681** 0.00593** 0.00576** 0.00989*** 0.00815*** (0.00266) (0.00271) (0.00259) (0.00320) (0.00295) Women in labor force, ratio to men 0.410*** 0.477*** 0.424*** 0.744*** 0.592***
(0.0946) (0.0957) (0.0915) (0.178) (0.166) Average years of secondary schooling 0.0453*** 0.0521*** 0.0671*** -0.0146 0.0167 (0.0125) (0.0128) (0.0123) (0.0186) (0.0175) Average years of tertiary schooling 0.0755* 0.0873* 0.0743* 0.00253 -0.0535
(0.0439) (0.0451) (0.0431) (0.0529) (0.0485) Population growth, in % 0.592*** 0.0250
(0.0920) (0.773) Population growth squared -0.00401*** 0.000211
(0.000676) (0.00574) Number of days required to start
business 0.000699(0.000311) ** (0.000710) -0.00176** Unemployment rate 0.670*** 0.473***
(0.0701) (0.0658) Number of changes in government per
year
Year FE YES YES YES YES YES F 20.91 15.16 28.87 14.83 13.29 Prob > F 0.0000 0.0000 0.0000 0.0000 0.0000
5. FURTHER DISCUSSION
In this analysis we aim to assess the relationship between different economic policies, human capital, gender and entrepreneurship. As a measure of entrepreneurial activity we have used the density of new businesses per one thousand people aged between 15 and 64 years old using a sample of around 130 countries in ten years 2004-2013. Given the assumptions explained in the second section, we have performed several statistical analyses which are now going to be discussed.
Hypothesis 1 has tested the positive relationship between fiscal freedom and new business creation. We have used a continuous variable which takes into account tax rates on individual income, on corporate income and the total tax revenue as a percentage of GDP, each of them being equally weighted as one-third of the final variable. Eventually, this hypothesis has been supported in the main regressions and in the following sensitivity analysis. The results of Djankov et al. in 2008 and Cullen and Gordon (2002) are confirmed as the countries with higher degrees of fiscal freedom in the sample have shown higher entrepreneurial rates. This is a clear symptom that entrepreneurs take advantage of low-tax jurisdictions to set up their companies. This is explained by the fact that revenues are expected to be higher in those countries due to the reduced amount of taxes companies are required to pay to the government, increasing the number of companies in the area.
We have also evaluated the negative association between corruption and our outcome variable. Instead of measuring corruption itself, we have used a categorical variable with higher values denoting better anticorruption policies. The second hypothesis is confirmed by the data as it has been seen that a stronger can act as a shield against embezzlement, theft, fraud and bribery which negatively affect entrepreneurship. This supported hypothesis goes in line with Bologna and Ross' assessment of corruption and its effect on entrepreneurship from 2015. Nevertheless, they measure entrepreneurship as the number of establishments per municipality, while what we measure here is the total number of new business per year. In any case, in some models of the sensitivity analysis anticorruption policies was not significant. This reduction in the explanatory power of the variable may be due to the interaction variable “absorbing” part of the effect.
average schooling years also seems to increase the number of new businesses per person. Both of them have a significant effect, showing that, if the average years of education is increased, the number of new business establishments rises. Although these variables were not found significant in all the models of the sensitivity analysis, they were significant in most of them. These results provide evidence that there is a certain level of education people need to acquire before being able to set up their own company. Since it is not easy to create your own business, many things should be taken into account and managing certain data and some calculations are usually a must. These results confirm the findings of Justin van der Sluis et al. (2008) and (2005). Nevertheless, we cannot confirm their point that in developing countries the highest percentage of entrepreneurial activity is not found in the highest skilled people, but in the middle skilled ones, as in the present study we do not divide the sample by the level of development of the countries.
The effect of the ratio of women in the labour force is also assessed in the present analysis, being our fourth hypothesis and expecting a negative relationship between both variables, maybe due to sexual discrimination or their family roles in some societies. Although this last hypothesis is not supported, it shows some interesting results. Namely, a higher percentage of women in the labour force has been found to increase entrepreneurship. We provide evidence that the countries where a greater percentage of working people are women, the most developed and modern societies in the sample, have more positive outcomes in terms of new businesses in all of our models.
corruption is still common, as Hellman et al. (2000) demonstrate that between 45% and 95% of the companies in twenty transition economies accepted having paid bribes. The expenditure in such policies can be paid off by the taxes derived from the new business ventures. Secondly, governments should foster the modernization of their socioeconomic societies in gender equality terms, as a higher percentage of women in the labour force has a positive impact on entrepreneurship. Promoting a more important role of women in the business environment could be essential to reduce this gap, and hence, modernizing an economy. Thirdly, governments should also consider prioritising the education of their citizens as it is a fundamental inversion for any economy. Not only should the secondary education be promoted but also the tertiary one. These highly skilled workers will create a greater number of businesses in the future, lowering, in turn, the unemployment and contributing to enlarge the GDP of the country.
There are also a number of recommendations for entrepreneurs which can lead to higher succeeding rates in their activity. Firstly, entrepreneurs and corporations willing to establish a new enterprise should search for low-tax countries were their revenues can be higher. These low-tax places will attract a high number of companies, which in turn, will improve the overall economy in the area for the companies settled. Secondly, entrepreneurs should choose a country without high corruption or one with a strong anticorruption policy. This would prevent them from having to pay bribes or being defrauded, and thus increase their revenues. Finally, entrepreneurs should also invest in socioeconomic modern societies, with an adequate ratio of women and men in the labour market, and acceptable educational standards. Actually, according to our results, businesspersons should dedicate more time to have an outstanding secondary education level, rather than focus on tertiary education.
are not the real values. Another limitation concerns the dependent variable. What we measure in the present work is the gross entrepreneurial effect, not the net effect. What can be observed is the density of new business per 1000 people, but not the business that, at the same time, go bankrupt or stop their activity. Finally, another limitation regards the taxes. We have used fiscal freedom, which although it contains corporate taxes in it, it is also composed by individual income taxes. Had we had a variable just for corporate taxes, our results could have been more accurate. Consequently, further research should be done improving these aspects.
6. CONCLUSIONS
In this study we have assessed the relationship between the entrepreneurial activity and a set of economic policies, human capital and gender. Our first hypothesis is that lower taxes increase new business density. The results support the preliminary hypothesis and the work of previous authors on the topic such as Djankov et al. (2008) and Cullen and Gordon (2002). Dia z Casero et al. (2015) also reach similar conclusions than the ones of this study, without distinguishing between personal and corporate taxes. Our available data did not allow us to distinguish such issues either. Therefore, further research distinguishing between both would be recommended.
We had also expected to find a positive link between anticorruption and the outcome. This hypothesis is also supported by our analyses, showing that the greater the anticorruption policy is, the more new businesses are created. Aidis et al. (2010) also point out that freedom from corruption is positive and significantly related to entrepreneurship. Nevertheless, we build on their work as when the interaction between fiscal freedom and anticorruption policies is included in the analysis, both the interaction and anticorruption policies become insignificant, denoting that the effect of these two variables alone might be mediated by the effect of them together
We also conclude that a better education fosters the entrepreneurial activity in a significant way for both more years of secondary and tertiary schooling. This also supports what Sluis et al. asserted in 2005 and later in 2008. They highlight the point that in developing countries a higher percentage of companies are set up by medium skilled workers rather than the highly skilled ones. This cannot be proven by the present analysis as our sample includes heterogeneous nations and we have not divided our sample by the development level of the countries.
Our results finally contradict our last assumption regarding the negative effect of higher numbers of women in the labour force and the entrepreneurial rates. However, the positive effect of a higher number of women in the labour force can be explained by the expected positive outcomes of a modern socioeconomic environment, as the ratio women to men in the labour force can be reflection of the cultural progress of the countries.
REFERENCES
Aidis, Ruta, Saul Estrin and Tomasz Mickiewicz, 2010, Size matters: Entrepreneurial entry and government, IZA Discussion paper no. 5052.
Alvarez, C., Noguera, M. & Urbano, D. (2012): Condicionantes del entorno y emprendimiento femenino. Un estudio cuantitativo en España. Economía Industrial, 368, 213-225.
Bemol, William J., 1990, Entrepreneurship: Productive, unproductive, and destructive,
Journal of Political Economy, vol. 98, no. 5, part 1, p. 893-921.
Bologna, Jamie and Amanda Ross, 2015, Corruption and Entrepreneurship: Evidence from a Random Audit Program, College of Business and Economics West Virginia
Universi-ty Working Paper no. 15-05.
Bruce, Donald and Mohammed Mohsin, 2006, Tax Policy and Entrepreneurship: New
Time Series Evidence, Small Business Economics, vol. 26, issue 5, p. 409-425.
Chowdhury, Farzana; David B. Audretsch and Maksim Belitski, 2015, Does corruption matter for international entrepreneurship? International Entrepreneurship and
Man-agement Journal, vol. 11, issue 4, p. 959-980.
Cullen, Julie Berry and Roger H. Gordon, 2006, Taxes and entrepreneurial risk-taking: Theory and evidence for the U.S., Journal of Public Economics volume 91, issues 7–8, p. 1479–1505.
Cullen, Julie Berry and Roger H. Gordon, 2002, Taxes and Entrepreneurial Activity: Theory and Evidence for the U.S, NBER working paper 9015.
Da Rin, Marco; Marina Di Giacomo, Alessandro Sembenelli, 2011 Entrepreneurship, firm entry, and the taxation of corporate income: Evidence from Europe, Journal of
Pulbic Economics vol. 95, issues 9–10, p. 1048–1066.
Darnihamedan, Ptheirya; Jörn. Block; Jolanda A. Hessels and Aram Simonyan, 2015 Start-up Costs, Taxes and Innovative Entrepreneurship, Tinbergen Institute Discussion
Paper -013/VII.
Díaz Casero, Juan Carlos; Ángel Manuel Díaz Aunión, María de la Cruz Sánchez Esco-bedo and R. Hernández Mogollón, 2015, Size of government and entrepreneurship. Analysis of three groups of countries with different economic development, FAEDPYME
Djankov, Simeon ;Tim Ganser; Caralee McLiesh; Rita Ramalho and Andrei Shleifer, 2010, The Effect of Corporate Taxes on Investment and Entrepreneurship, American
Economic Journal: Macroeconomics, American Economic Association, vol. 2(3), p. 31-64.
Dreher, Axel and Martin Gassebner, 2007, Greasing the wheels of entrepreneurship? Impact of regulations and corruption on firm entry, KOF Working Paper No. 166.
Fischer, Eileen M.; A. Rebecca Reuber and Lorraine S. Dyke, 1993, A theoretical over-view and extension of research on sex, gender, and entrepreneurship, Journal of Business
Venturing volume 8, issue 2, p. 151-168.
Gentry, William M. and R. Glenn Hubbard, 2005, Success taxes, entrepreneurial entry, and innovation, Chapter of the NBER book Innovation Policy and the Economy, volume 5, p. 87 – 108.
Heij, Christiaan Paul de Boer, Philip Hans Franses, Teum Kloek, and Herman K. van Dijk, 2004, Econometric methods with Applications in Business and Economics, Oxford
University Press, first edition.
Hellman, Joel S.; Geraint Jones, Daniel Kaufmann and Mark Schankerman, 2000, Measuring Governance, Corruption, and State Capture How Firms and Bureaucrats Shape the Business Environment in Transition Economies, Policy Research working Paper, No. 2312.
Hill, Lissa, 2013, Conceptions of political corruption in ancient Athens and Rome,
His-tory of Political Thought, 34(4), 565 –587.
Hill, R. Carter, Griffiths, William E., and Lim, Guay C., 2012, Principles of Econometrics, John Wiley & Sons, fourth edition
Keuschnigg, Christian and Soren Bo Nielsen, 2003, Taxation and venture capital backed entrepreneurship. University of St. Gallen Economics Discussion Paper No. 2003-17.
Kourilsky, Marilyn L and William B Walstad, 1998, Entrepreneurship and female youth: knowledge, attitudes, gender differences, and educational practices, Journal of
Business Venturing volume 13, issue 1, p. 77–88.
Noguera, N., Alvarez, C. & Urbano, D., 2013, Socio-cultural factors and female entrepreneurship. International Entrepreneurship and Management Journal, 9(2): 183-197.
Robinson, Sherry and Hans Anton Stubberud, 2009, Sources of advice in entrepre-neurship: gender differences in business owners´ social networks, International Journal
of Entrepreneurship, p. 83-101.
Robinson, Peter B and Edwin A. Sexton, 1994, The effect of education and experience on self-employment success, Journal of Business Venturing volume 9, issue 2, p. 141–156.
Rose-Ackerman, S., 2006. International Handbook on the Economics and Corruption. Edward Elgar Publishing Limited
Russell S. Sobel, 2008, Testing Baumol: Institutional quality and the productivity of entrepreneurship, Journal of Business Venturing, vol. 23, issue 6, p. 641–655.
Tonoyan, Vartuhi; Robert Strohmeyer; Mohsin Habib and Manfred Perlitz, 2010, Cor-ruption and entrepreneurship: How formal and informal institutions shape small firm behavior in transition and mature market economies, Entrepreneurship Theory and
Prac-tice, vol. 34, issue 5, p. 803–831.
Van der Sluis, Justin; Mirjam Van Praag and Wim Vijverberg, 2008, Education and en-trepreneurship selection and performance: a review of the empirical literature, View
issue TOC volume 22, issue 5 December 2008, p. 795–841.
Van der Sluis, Justin; Mirjam Van Praag and Wim Vijverberg, 2005, Entrepreneurship Selection and Performance: A Meta-Anlalysis of the impact of Education in Developing Economies, The World Bank Economic Review (Impact Factor: 1.13). 02/2005; 19(2):225-261.
Van Stel, André; David J. Storey and A. Roy Thurik. 2006; The effect of business regu-lations on nascent and young business entrepreneurship, Small Business Economics, vol. 28, p. 171-186.
Wooldridge, J. M., 2010, Econometric Analysis of Cross Section and Panel Data, 2nd
APPENDIX
APPENDIX A: SEARCH STRATEGY
A1 - Search terms for the literature review on the association between entrepreneurship and corruption and taxes
a. Entrepreneurship b. Entrepreneurs c. New business d. Corruption e. Bribes f. Taxation g. Taxes h. Corporate taxes
A2 - Search terms for the literature review on the association between entrepreneurship and education
a. Entrepreneurship b. Entrepreneurs c. New business d. Education e. Educational level f. Training
A3 - Search terms for the literature review on the association between entrepreneurship and gender differences
APPENDIX B: LIST OF COUNTRIES INCLUDED IN THE ANALYSIS
LIST OF COUNTRIES, PERIOD 2004 – 2013
Afghanistan Cyprus Italy Namibia South Sudan Albania Czech Republic Jamaica Nepal Spain
Algeria Denmark Japan Netherlands Sri Lanka Antigua and
Barbuda Dominica Jordan New Zealand St Kitts and Nevis Argentina Republic Dominican Kazakhstan Niger St Lucia Armenia Egypt Kenya Nigeria St Vincent and the
Grenadines Australia El Salvador Kiribati Norway Suriname Austria Estonia Korea, South Oman Sweden Azerbaijan Ethiopia Kyrgyzstan Pakistan Switzerland Bangladesh Finland Laos Panama Syria
Belarus France Latvia Peru Tajikistan Belgium Gabon Lesotho Philippines Thailand Belize Georgia Liechtenstein Poland Timor-Leste Bhutan Germany Lithuania Portugal Togo
Bolivia Ghana Luxembourg Qatar Tonga Bosnia and
Herzegovina Greece Macedonia Romania Tunisia Botswana Grenada Madagascar Russia Turkey Brazil Guatemala Malawi Rwanda Uganda Bulgaria Guinea Malaysia Samoa Ukraine Burkina Faso Haiti Maldives Sao Tome and Principe United Arab Emirates Cambodia Hungary Malta Senegal United Kingdom Canada Iceland Mauritius Serbia Uruguay Chile India Mexico Sierra Leone Uzbekistan Colombia Indonesia Moldova Singapore Vanuatu Congo,
Democratic
Republic Iraq Mongolia Slovakia Zambia Costa Rica Ireland Montenegro Slovenia
APPENDIX C: CORRELATION MATRIX
Correlations of dependent and independent variables Logarithm of
new business density
Fiscal
freedom Anticorruption policies labour force Women in
Average years of secondary school Average years of tertiary school GDP growth (annual %) Logarithm of new business density 1.00 Fiscal freedom -0.11* 1.00 Anticorruption policies 0.64* -0.19* 1.00 Women in labour force 0.39* -0.25* 0.25* 1.00 Average years of secondary school 0.62* -0.04 0.38* 0.22* 1.00 Average years of tertiary school 0.53* -0.05 0.27* 0.24* 0.69* 1.00 GDP growth (annual %) -0.18* 0.23* -0.16* -0.13* -0.15* -0.13* 1.00
APPENDIX D: BOX GRAPHS
Anti-corruption policies
APPENDIX E: LINEAR REGRESSION RESULTS WITH THE LOGARITHM OF NEW BUSINESS DENSITY AS THE DEPEND-ENT VARIABLE. RANDOM EFFECTS
VARIABLES M1 M2 M3 M4 M5 M6 Fiscal freedom 0.0125*** 0.0115*** 0.00837*** 0.00342 (0.00364) (0.00285) (0.00266) (0.00230) Anti-corruption policies 0.0239** 0.0235** 0.0253*** (0.0105) (0.00987) (0.00842) GDP growth, in % -0.000769 0.0127*** (0.00323) (0.00301) Women in labour force,
ratio to men
0.384*** 0.415*** 0.391***
(0.0989) (0.0972) (0.0919) Average years of secondary
schooling
0.0564*** 0.0452***
(0.0124) (0.0122) Average years of tertiary
schooling
0.0658 0.0568 (0.0444) (0.0432) Observations 1,091 1,077 1,070 1,069 1,069 1,049 Number of countries 133 133 132 132 132 131 Country RE YES YES YES YES YES YES Year RE YES YES YES YES YES YES Wald chi2 129.2 146.1 159.6 162.4 183.5 232.7 Prob > chi2 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
APPENDIX F: TESTING MULTICOLLINEARITY: VARIANCE INFLATION FACTORS (VIF)
VARIABLE VIF 1/VIF
