Social benefits

3.3 Model: Model Specification

Social benefits in cash and in kind are closely interdependent. Putting both of them in the model is very likely to cause multicollinearity, which will make regression coefficients insignificant. Therefore, I will provide two model specifications: one without social benefits in kind and one without social benefits in cash. I change the variables specifications by including lags rather than by differencing or taking logarithms. I did not find any non-stationary processes in the explanatory variables, therefore differencing is not required. Logarithms are usually used to make the effect of large variables on coefficients more influential. However I use percent values on the right and on the left side of the equation and thus applying logarithms is unnecessary and is unlikely to bring any value.

First of all I specified the model by taking all the appropriate lags and putting them in one regression of three types: pooled, fixed effects and random effects. The pooled model (Model 4) has significant social benefits in kind. Fixed effects model (Model 5) has low F-statistics, indicating that it is a better fit than pooled model. Random effects model (Model 6) is a better fit than fixed effects model (Test 9), as p-value in Hausman test is 49%.

Secondly, I noticed that this model is not the best possible fit and started searching for better specification. Education and research and development have high correlation and I decided to try their fit in the model separately, as together they could result in multicollinear relationship. Social benefits in cash and in kind were separated to check their separate influence on self-employment. Each model included different variables and was modified until I received a good fit with significant estimates and considerable R-squared. I checked all models for heteroscedasticity and autocorrelation and compared fixed and random effects and pooled model. The output of my work is presented below.

1) Social benefits in kind model (Model 7).

After choosing different lags and variables combinations I estimated the best fit for social benefits in kind model:

This specification is corrected to solve heteroscedasticity and autocorrelation by clustering and applying robust standard errors. This specification is better fitted by random effects model according to Hausman test (Test 10).



Table 1

The social benefits in kind regression coefficient is positive, indicating that higher social benefits in kind cause self-employment increase. This result supports my regression hypothesis and opposes the generally excepted view that higher social benefits reduce business activity. I see two possible explanations of such a coefficient. On the one hand, high social benefits in kind act like a safety cushion in case of business failure. Therefore risks of starting up are smaller, as healthcare, education, transportation etc. will still be available to users. Hessels (2008) proposed a similar hypothesis, stating that social security expressed in terms of social benefits supports independent choice to start a business, as the cost of failure is mitigated. This function is similar to social benefits in cash and does not explain the difference in coefficients. On the other hand, social benefits in kind may act as an important support during the business formation, even if it is successful. Most employers in OECD countries, even less developed, provide workers with health insurance. When an employee leaves, he loses the insurance and has to pay for healthcare, which is very expensive and hardly affordable while running a business. Therefore, a potential entrepreneur will be more willing to leave a job for starting a business if he knows that in case of emergency he may still use government support. Similar situation may be applicable to education: a person may be unwilling to leave a job if he finances his children education. A business takes away such an opportunity, government provides high-quality free education in form of social benefits in kind. Therefore the process of running a business in a country with high social benefits in kind is more comfortable and secure. This result has never been achieved before, as no predecessor has separated social benefits in cash and in kind and did not expect different effects.

Higher education perceived opportunities have a significant positive effect on self-employment. It is justified by the previous studies and supported by empirical evidence. Indeed, running a business without a higher education is much more difficult than with: entrepreneurs have a different attitude to life, fewer skills and narrower network. Therefore it was listed as an important factor of business development by many works, including Kerr and Nanda (2009) and Wadhwa, Freeman and Rissing (2008). Perceived opportunities reflect the impression of potential entrepreneurs on the constraints and benefits of running a business. Higher opportunities reflect higher probability of starting up in a country. This relationship was studied by many GEM authors, including Hessels. They concluded that this variable is a significant indicator of business environment and desire of residents to start up.

Overall, this regression supports my hypothesis about positive relationship between social benefits and unemployment and may have a number of explanations that deserve a careful consideration.

2) Social benefits in cash model (Model 8).

The best social benefits in cash model I estimated is:

Hausman test proved that it is a fixed effect model (Test 10). Autocorrelation and heteroscedasticity were corrected by clustering and applying robust standard errors. This model does not have any multicollinearity, as none of the correlations between explanatory variables is more than 17%. All of the variables coefficients are significant and the model has within R squared of 16%, which is a good result for a panel.



Table 2

Intuitively this model supports the previous works results (Parker and Robson (2004), Hessels et al (2007)) that proved negative relationship between social benefits and self-employment. However in my specification I use vastly different variables: low income, perceived opportunities and education and specify social benefits as social benefits in cash rather than overall.

My hypothesis about positive relationship between social benefits in cash and self-employment is rejected and I see a number of reasons for that. First of all, social benefits in cash, expressed in government spending on unemployment benefits, pensions etc., may discourage people from starting a business, because the cost of free time is perceived to be higher than the benefit of starting a business. Indeed, this explanation was provided by Steven Lippman and John McCall (1979), who claimed that the increase in unemployment benefits results in lower opportunity cost of leisure and thus lower employment in the economy. So, people will choose not to leave their work for starting a business, but rather to become unemployed and enjoy free time and high government receivables, maximizing their utility. Therefore social benefits in cash alone decrease the risk of starting a business by less than the increase the opportunity cost of it.

The other variables have an expected effect on self-employment: education positive, low income positive, perceived opportunities positive. Wadhwa, Freeman and Rissing (2008) come to similar conclusions about education: higher education is a driver of business development. Secondary education is not enough to develop a sustainable business, obtain enough funding etc, while over-education reduces the real work experience duration, which is important for starting a business. As discussed in variables description, low income creates a significant incentive to start a business: low wage does not keep workers at work. A different view on this variable is that low wages decrease the cost of starting a business, as labour is cheaper to hire. Finally, perceived opportunities have a direct positive influence on business creation, as if potential businessmen expect fewer constraints in starting up, a larger share of them will.

Overall, social benefits in cash have a significant negative effect on self-employment, which was justified by previous works.

These model specifications partly support the previous studies conclusions and support my hypothesis for 32 countries. However to have a clear picture of the social benefits effect I decided to test the same hypothesis for another country sample. The current sample includes both developing and developed countries and in my opinion there should be difference between the model results for them. Rossouw (1994) Rossouw (1994), “Business Ethics in Developing Countries”, Business Ethics Quarterly Vol. 4, No. 1 (Jan., 1994), page 50 expressed a similar idea, stating that business formation in developing countries is bounded by lack of human rights protection, life instability and pro-active moral concerns. These constraints surely prevent less developed countries in their ability to start and grow a business. That very bound may have a higher weight in potential businessman decision function, than social benefits increase. Indeed, a small change in social benefits is not enough to make people start a business in an environment, restrained by the shortage of democratic institutions.

To isolate those imperfections from my analysis I checked the relationship between social benefits in cash and in kind on developed countries (G7).



3.4 Model: Developed Countries

In this part of the analysis I will isolate country-specific imperfections and test the relationship between social benefits in cash and in cash on developed, rich countries: Japan, United States, United Kingdom, Italy, France and Germany. This sample came from the G7 list, but does not contain Canada as I don't have the full dataset for this country.

1) Social benefits in kind (Model 10)

Same specification shows same coefficients

2) Social benefits in cash (Model 11)

Overall testing G7 countries separately did not show any different results. The coefficients became more stable and R squared increased a lot due to smaller sample size and larger interdependence between this country groups.



4. Analysis

4.1 Analysis: Limitations

Having achieved these innovative conclusions, I have to assess my work critically. I see three possible ways to criticize this work and will try to outline and solve them if possible. The problems may arise from the data specification, data quality and kind and basis for conclusions drawn.

First of all, social benefits may be expressed in many different ways, but historically most of the empirical works tested them as a percent of GDP. Because of that variable specification the authors faced a number of problems. Hessels and authors (2008), for example, stated that one possible explanation of output contradiction their hypothesis is that social benefits are treated in aggregate terms, which may reflect a large number of people receiving benefits instead of the actual size. This variables specification problem is resolved in my study, as I use social benefits as in percentage from GDP per capita, rather than overall GDP.

However I expect another important variable specification problem to arise. Though the positive interdependence between self-employment and social benefits in kind makes intuitive since, this relationship has not been studied deeply before and there may be various explanations, not related to the actual social benefits size. One likely limitation is associated with the form of my testing: I chose to use social benefits as a percentage of GDP per capita, which is very likely to create adverse outcome. GDP is an unstable variable, which may increase or decrease at a various magnitude, depending on the economic cycle. Thus in case of crisis GDP falls, while social benefits expressed as a percentage of GDP per capita stay unchanged or even increase. During crisis business activity is usually less effective and self-employment thus is lower according to Koellinger and Thurik (2009). This may result in negative relationship between social benefits and self-employment and may make my analysis linked to GDP fluctuations effect rather than to social benefits. Having that potential problem in mind I decided to check this hypothesis in the next part of my work.

The second possible drawback arising from my testing and conclusions is associated with data quality. I used 32 countries data for 14 years. This size is moderately large to achieve significant results and relatively up to date. Parker and Robson (2004) analyzed the data for the 20^th century, which is hardly applicable to present situation, while Philipp Koellinger and Maria Minniti (2008) had a problem of missing observations and Hessels (2007) took data for 2 years only, which clearly could not be used to draw any conclusions. Unlike my predecessors, I use fresh data for a rather long time period, which has few missing observations. Nevertheless, the countries in my sample are mostly developed and rich. Thus poor economies with lower democratic values are excluded from the analysis. Moreover the 14 years of observations is not enough to detect non-stationary with certainty, as the stationary variable may act differently at the time period under consideration.

The two potential problems outlined above are not likely to affect my analysis drastically. First of all, testing social benefits increase in developing countries, like Africa will likely create a reverse effect on self-employment due to specifics, discussed by Rossouw (1994). Indeed, I think that testing poor and wealthy countries together deserves a separate piece of research, as the specific difference among them deserves a separate hypothesis. My analysis is thus limited by moderately developed and wealthy countries. As there are not many, I suppose, that the sample I have taken is a good representation of the whole population and the results may be applied to countries with similar characteristics, but not to extreme cases. Therefore my countries sample is enough for drawing a conclusion about developed countries with certainty.

Finally, a possible spot for criticism are the econometrical issues: model and testing. Indeed, it seems odd that social benefits in cash and in kind are significant in different model types (fixed and random effects). However, all the tests (Hausman, F-test) suggest that they are the best specifications. Therefore, I conclude that these specifications are correct and may be used in this model. The other possible problem is the non-stationary process of self-employment, which intuitively is stationary, as the upper bound is 100% GDP. This problem arises from a relatively short time period: self-employment as a percent of GDP is growing by small amounts in time. The Durbin-Watson test reflects a strong non-stationary process for this time period, which is corrected by differencing. I use the difference in my model because this model is aimed to predict the self-employment for a relatively short period of time, which still reflects non-stationary process. To sum up, all the corrections made in my work are done based on econometrical tests and may be justified.

Overall, my conclusions may be challenged from three different sides: variable specification, sample choice and econometrical side. Each of these points is discussed and explained in this section. However my conclusions should be further developed and discussed, as a lot of fine points deserve closer attention. My justification of the result should be challenged and supported by other studies, as there is a lack of research in this field and my results contradict most of the previous studies. To summarize, my conclusions are supported econometrically, intuitively and theoretically and thus should be used in future.

4.2 Analysis: Limitations check

To check the first and most threatening limitation presence - GDP fluctuations effect on self-employment - I checked this model for the absolute social benefits effect in dollars per capita. I used the same figures for other variables, but substituted percentage values for social benefits in cash and in kind to eliminate the GDP fluctuations effect on self-employment.

To eliminate any possible errors in judgment I checked for non-stationary processes, heteroscedasticity and autocorrelation. Social benefits in kind and in cash proved to be stationary; the models however reflected autocorrelation and heteroscedasticity that were eliminated by clustering and taking robust standard errors.

I tested both social benefits in kind (Model 12) and in cash (Model 13). The results for the target variables matched the results in the standard model: social benefits in kind have a positive effect on self-employment, social benefits in cash - negative. This result illustrates independence of the analysis from the economic cycle and proves that the results obtained are stable: they work for G7 countries, for 32 countries in the initial sample and for the explanatory variables presented in dollars and in percent.

Table 3

Table 4

This analysis proved that higher GDP per capita effect, implying lower social benefits as a percentage GDP is not significant enough to outperform the real effect of social benefits on self-employment.

To sum up, the model proposed by my analytical paper is valid for further application.



Conclusion

A lot of research papers analysed the effect of social and unemployment benefits on unemployment, but there were only few, checking the relationship between social benefits and business activity. This topic is new and important for the society, as many governments all over the world are trying to increase the business activity and social benefits may be a way to do so. Most of the previous works in this field concluded that social benefits have a negative effect on self-employment. These results were justified in a similar way to positive effect of social benefits on unemployment: higher social benefits discourage workers from choosing work versus free time. Indeed, with higher social benefits there are fewer incentives to spend more time on work than on leisure and business demands even more time than any job. There was a study by Jolanda Hessels et al (2008), which proved that there is a positive relationship between social benefits and necessity motive for business formation.

No research paper separated social benefits in cash and in kind effect on self-employment and only one hypnotized positive relationship between social benefits and self-employment. My work not only inputs a new segmentation of variables but also provides new insight in the relationship between social benefits and self-employment. I proved that social benefits in cash are in line with consensus view and have a negative effect on self-employment, while social benefits in kind deviate and show a significant positive effect. This result is new and changes the attitude to social benefits policy with respect to business development. I suppose that the effect of social benefits in cash eclipsed the effect of social benefits in kind in the previous works and resulted in negative relationship rather than in positive.

I propose a general explanation, used in many early papers for social benefits in cash effect explanation and propose a new innovative one for the outstanding results about social benefits in kind. Social benefits in cash may be viewed as a factor of worker's utility function as income that he receives no matter whether he works or not. The higher are social benefits in cash the less a worker values money versus leisure and vise versa. Therefore more people will choose to give up their jobs/businesses for free time. This is supported by my analysis results and justified by previous papers. Social benefits in kind, on the contrary serve as a “safety cushion” rather than a factor of utility function: they support a person with essential needs like healthcare and children education while he forms a business and in case of failure. Indeed, when a potential businessman leaves his job, he loses his health insurance and cannot pay for family education, which is a considerable constraint against starting up a firm. When the government provides citizens with better access to such services a person has fewer constraints to starting a business, as he is supported both during the business establishment and in case of business failure. It is important to note that this hypothesis is valid only for countries with democratic regime, low corruption and high human rights support. Otherwise the social benefits effect will be eclipsed by these imperfections and will be insignificant.

To conclude, I have to outline that this paper results stand out and open a new path for further research of the government policy effect on business formation.

One of my hypothesis turned out to be correct: social benefits in kind have a positive effect on self-employment. Moreover I assumed that social benefits in cash will have a smaller positive impact than social benefits in kind. This hypothesis is partially correct, as they turned out to have a negative effect on self-employment, which is smaller than positive - of social benefits in kind.



References

1. Christopher A. Pissarides, 2000, “Equilibrium Unemployment theory”, 2nd edition, Cambridge (MA): MIT Press

2. Dale T. Mortensen, Christopher A. Pissarides, 1994, “Job Creation and Job Destruction in the Theory of Unemployment”, The Review of Economic Studies, Vol. 61, No. 3 (Jul., 1994), pp. 397-415

3. Delmar, F., & Shane, S., (2003), “Does business planning facilitate the development of new ventures?”, Strategic Management Journal, 24, 1165-1185

4. Hey, John D. & Mavromaras, Kostas Gr., 1981. "The effect of unemployment insurance on the riskiness of occupational choice," Journal of Public Economics, Elsevier, vol. 16(3), pages 317-341, December

5. Jolanda Hessels, Andrй van Stel, Peter Brouwer, Sander Wennekers, 2007, “Social security arrangements and early-stage entrepreneurial activity”, Scientific Analysis of Entrepreneurship and SMEs

6. Jolanda Hessels, Marco van Gelderen, Roy Thurik, 2008, “Entrepreneurial aspirations, motivations, and their drivers”, Small Business Economics, Volume 31, Issue 3

7. Marcus Wagner, 2010, “Corporate Social Performance and Innovation with High Social Benefits: A Quantitative Analysis”, Journal of Business Ethics (2010)

8. Morris Altman, 2004, “Why Unemployment Insurance Might Not Only Be Good for the Soul, It Might Also Be Good for the Economy” Review of Social Economy, Vol. 62, No. 4 (December 2004), pp. 517-541

9. Peter Diamond, 1971, “A model of price adjustment”, Journal of Economic Theory, 1971, vol. 3, issue 2, pages 156-168

10. Philipp Koellinger and Maria Minniti, 2008, “Unemployment Benefits Crowd Out Nascent Entrepreneurial Activity”, Erasmus Research Institute of Management

11. Philipp Koellinger and Roy Thurik, 2009, “Entrepreneurship and the Business Cycle”, Tinbergen Institute Discussion Paper No. TI 09-032/3

12. Pollmann-Schult and Felix Buchel, 2005, “Unemployment Benefits, Unemployment Duration and Subsequent Job Quality: Evidence from West Germany”, Acta Sociologica March 2005 48: 21-39

13. Rossouw (1994), “Business Ethics in Developing Countries”, Business Ethics Quarterly Vol. 4, No. 1 (Jan., 1994), pp. 43-51

14. Simon C. Parker, Martin T. Robson, 2004, “Explaining International Variations in Self-Employment: Evidence from a Panel of OECD Countries”, Southern Economics Journal, 71 (2). Pp. 287-301

15. Steven A. Lippman, and John J. McCall. 1979. "Expected Utility Maximizing Job Search," in Studies in the Economics of Search. S. Lippman and J. McCall, eds. Amsterdam: North-Holland, 133-56

16. Vivek Wadhwa, Richard B. Freeman, Ben A. Rissing, 2008, "Education and Tech Entrepreneurship”, Kauffman Foundation of Entrepreneurship

17. William R. Kerr, Ramana Nanda, 2009, “Financing Constraints and Entrepreneurship”, Harvard Business School Entrepreneurial Management Working Paper No. 10-013



Applications: Tests

Test 1.

Test 2



Test 3

Test 4



Test 5

Part 1

Part 2

Part 3

Test 6



Test 7

Test 8

Test 9



Test 10

Test 11

Applications: Corrections

Correction 1



Correction 2

Correction 3

Correction 4



Applications: Models

Model 1

Model 2

Model 3



Model 4

Model 5

Model 6



Model 7

Model 8

Model 9



Model 10

Model 11

Model 12

Model 13



Applications: Correlations

Correlation 1

Correlation 2

Correlation 3

Correlation 4

Correlation 5



Correlation 6

Correlation 7

Correlation 8

Correlation 9

Correlation 10

Correlation 11

Размещено на Allbest.ru

...