ENTREPRENEURSHIP AND UNEMPLOYMENT: RELATIONSHIPS BETWEEN UNEMPLOYMENT AND ENTREPRENEURSHIP IN 37 NATIONS PARTICIPATING IN THE GLOBAL ENTREPRENEURSHIP MONITOR (GEM) 2002
Marc Cowling, Foundation for Entrepreneurial Management, London
William D. Bygrave, Babson College
LITERATURE REVIEW AND HYPOTHESIS DEVELOPMENT
DATA AND METHODOLOGY
This paper studies the relationship between entrepreneurship and unemployment. We focus on Necessity TEA (total entrepreneurial activity for those individuals pushed into entrepreneurship because they have no better alternatives for work). We a priori predict that when unemployment is high, TEA (necessity) will be high as outside alternatives in the labor market diminish. Yet we also predict that this effect will be moderated in nations where unemployment benefits are high. In addition we focus on the composition of the stock of unemployed and how difficult, or easy, it is to start a new business. Both factors have been shown to be important in previous studies (Cowling and Mitchell, 1997; Robson, 1998). Our findings offer some support for our a priori predictions, but show that the unemployment effect is far more complex than previously believed.
Small businesses make an important contribution to the success of a country’s economy. They are major creators of jobs, they innovate, and they spot and exploit new opportunities. Even though many new business start-ups have no explicit growth aspirations, and indeed many cease trading quite soon after start-up, it is still the case that a period of running one’s own business provides an opportunity to learn new skills which are valuable to potential employers. A better understanding of the nature of business start-up by the unemployed, and the forces that drive it, will further the ability of policy-makers to create the conditions under which the unemployed can successfully make the transition into business on their own account. This is of great importance given the direct link between business formation by the unemployed and reductions in the level of unemployment. The great advantage of this study is that we use data for 29 countries across the world, collected as part of the Global Entrepreneurship Monitor (GEM)1 annual surveys. Thus we have a tremendous diversity in terms of economic systems and labor market conditions. The rest of this paper is organized as follows; first we review the theoretical and empirical literature relating to new business formation by the unemployed; then we discuss the data and method to be used in the empirical part of the paper; next we present the sample statistics; this is followed by our multivariate analysis; and we conclude with a discussion of our findings.
LITERATURE REVIEW AND HYPOTHESIS DEVELOPMENT
The individual’s decision to start a new business as a response to unemployment or lack of outside alternatives in the labor market can be explored within the framework of the microeconomic theory of labor supply. The standard model, which is rooted in the theory of consumer choice, predicts that individuals are more likely to participate in the labor-force when:
|The more they like the benefits of working (e.g income, job status) relative to the benefits of leisure.|
|The lower is their income from non-work sources.|
|The lower is their fixed cost of working.|
|The higher is their real wage rate.|
The welfare system determines the trade-off that an unemployed person must make in deciding whether to re-enter the workforce or stay out of work. In a nation with generous welfare benefits, an unemployed person might choose to enjoy 24 hours of daily leisure rather than work. For instance, if the net income earned by working eight hours a day is no bigger than an individual’s unemployment benefit, he has little or no financial incentive to re-enter the workforce. Hence, a low-income individual may choose to stay unemployed until his benefits run out. In contrast, a high-income earner is more likely to re-enter the workforce as soon he can find a job because there is a substantial net income difference between working and being unemployed. Simply put, the lower an individual’s earning power relative to his unemployment benefits the less likely he is re-enter the workforce.
This is important for us, as necessity entrepreneurs are, by definition, unable to find suitable employment in the waged sector of the economy. Thus potential necessity entrepreneurs with low income earning potential, particularly in countries where welfare payments are high, may have a lower income incentive to start their own business. Further, if the fixed costs of working are high these may also reduce the rate of necessity entrepreneurship.
H1: The more generous the welfare system the lower the rate of necessity entrepreneurship.
The final piece of our theoretical development adds a time dimension to the decision to work or consume leisure, although it also impinges on the welfare payment system. This relates to the pension system when individuals become inactive in the labor market due to old age. In general, the more an individual pays into a pension system, the more he receives as unearned income in retirement. This provides an incentive to supply more hours of work thus increasing non-labor income in old age. Yet, in countries where there is little or no pension provision, individuals may have a greater incentive to work and save more to avoid poverty in old age. Thus we might expect that necessity entrepreneurship rates would be higher in countries where larger proportions of the population have no income provision in old age. The provision of state funded pensions can also be seen as reducing the incentive to work by raising total lifetime income from non-work sources.
H2: The higher the proportion of the adult population with pension provisions, the lower the rate of necessity entrepreneurship.
Having developed a simple theory to explain the key factors in the individuals choice to work or consume leisure, and set this in the context of the decision to start your own business as a necessity entrepreneur, we now move on to consider the empirical literature relating to unemployment and business start-up. Table 1 reports the findings of a number of empirical studies investigating the relationship between unemployment and business start-up.
The results show some interesting, and contrasting results, particularly between aggregate, economy-wide studies (from the top the first eight) and individual level, local labor market based studies (the last two). It is also worth noting that while the Abell et al (1995) study initially found an insignificant effect for the unemployment rate, when they ran separate models for entry from employment and unemployment divergences in their results appeared. Their findings suggest that although entry from waged employment are unrelated to the unemployment rate, a positive relationship exists between entry from unemployment and the unemployment rate. This mirrors the local labor market findings of Cowling and Hayward (2000) and Cowling (2003), which analysed records for in excess of 26,000 individuals, the majority of whom were unemployed at the initial stage of investigation. These results, combined with the generally positive aggregate effects found in nearly all non-UK studies, suggest that as the aggregate unemployment rate rises, the probability of securing waged employment may fall even further for the unemployed who typically lack the levels of human capital (both formal and informal) of those in employment. This is given further support from the Cowling and Mitchell (1997, pp.427, 434) study which found that:
“Self-employment is a last resort for certain individuals marginalized in the employed sector and facing lengthy spells of unemployment. . . . Initially the short–term unemployed can compete for waged employment and are re-employed, thus tending to lower the proportion of the workforce in self-employment. But as unemployment spells lengthen these individuals become the long-term unemployed. At the same time the likelihood of obtaining waged employment diminishes and self-employment becomes a last resort option for long-term unemployed people.”
This is close to our definition of the necessity entrepreneur as someone who perceives no suitable employment alternatives as their reason for starting a business. In line with this we propose the following two hypotheses:
H3: When unemployment rates are high, necessity entrepreneurship rates will be high.
H4: When the youth share of the total stock of unemployed is high, the rate of necessity entrepreneurship will be high as they lack the human capital (education, experience and job skills) to secure waged employment.
The majority of studies of new business entry focus on the relative rewards that can be gained in alternative labor market states (see for example Taylor, 1996). The extent of entry barriers can be added to this as logic suggests that the lower the barriers to market entry are, the more probable it is that potential entrants actually enter. The subject of entry barriers is relatively under-researched in this context, with discussion focusing solely on an individuals’ lack of finance or skill. Yet we are concerned with entry barriers in the standard industrial economics sense too as an element of market structure that refers to obstacles in the way of potential newcomers to the market; or those obstacles that operate to discourage entry (e.g. advertising, threats of retaliatory action by incumbent firms, control of essential raw materials, technology, or market outlets, etc). With this in mind we propose one final hypothesis:
H5: Where barriers to market entry by new businesses are high, the rate of necessity entrepreneurship will be lower.
Having generated some testable hypotheses from a survey of the theoretical and empirical literature, we now move on to the empirical part of our paper. The first thing we do is discuss the data to be tested and the methodology.
DATA AND METHODOLOGY
Our study comprises data from 37 nations participating in GEM 2002. Those 37 economies comprise 89 per cent of the GDP and 63 per cent of the population of the world. The GEM assessments are based upon four types of data. The most important are the adult population surveys that examine a representative sample of adults in each of the 37 nations. Local survey research firms are used to collect this information from 1,000 to 16,000 adults in each country. Individuals are interviewed about their participation in, and attitudes towards entrepreneurial activity. From these interviews, data are aggregated to provide aggregate country specific measures of entrepreneurial activity. The key variable as far as this study is concerned is people who are involved in entrepreneurial activity because they have no feasible outside alternatives in the labor market. We refer to these individuals as necessity entrepreneurs. Thus for each country we have a TEA Necessity rate which measures the proportion of the adult population who are involved in necessity entrepreneurship.
The second source of data is the expert interviews. These comprise personal interviews conducted with between 20 and 70 national experts in each GEM country. The experts provide their personal assessments of the unique aspects of their country’s culture and institutional framework in relation to entrepreneurship and entrepreneurial activity. This information is supplemented by a 10 page, standardised, questionnaire filled in by these same experts. The final element of the data collection process is standardised cross-national data. This is drawn from harmonised sources such as IMF (International Monetary Fund), ILO (International Labor Organisation) and the like.
The data analysis for this study is conducted on two levels. First, we explore the basic sample statistics relevant to necessity entrepreneurship and unemployment. The key variables used are drawn from our literature review and hypothesis development section. Then we progress to estimating a series of cross-sectional, econometric models to isolate the key relationships between necessity entrepreneurship rates across countries and unemployment factors. In order to capture the dynamic relationship between the two, we use an appropriate lag structure that incorporates lagged explanatory variables on the right hand side of the models. We also incorporate the lag of our dependent variable in order to capture any short-run persistence in necessity entrepreneurship across nations that may exist. The model can be written thus;
(1) TEA Necessity it = f (unemployment variables it, it-1, it-2, social welfare i, pension provision i, market barriers i, TEA necessity it-1)
where i represents a country and t denotes time.
Thus we are seeking to explain the observed cross-country variation in necessity entrepreneurship in 2002 by a vector of unemployment variables, current and lagged, together with the lag of the dependent variable.
We then estimate a series of first difference models that seek to explain the change in necessity entrepreneurship rates across countries with a vector of unemployment level and first difference variables. The first difference model can be written thus;
(2) TEA Necessity it = f (unemployment variables it, it-1, unemployment variables it, it-1, it-2, social welfare i, pension provision i, market barriers i, TEA necessity it-1)
where i represents a country, t denotes time, and the D represents a first difference.
In both models the variables we use are defined thus:
% of adult population involved in necessity
Unemployment rate: % of labor force unemployed
Youth Unemployment: % of total stock of unemployed accounted for by people under the age of 25.
Social Welfare: Social security payments as % total GDP weighted by the stock of unemployed.
Pension Provision: % of adult population with pension provision.
Low Barriers: Ease of new market entry by new business (scale 1=very difficult to 5=very easy).
In this section we present the sample statistics for the variables to be incorporated in our multivariate analysis, together with a brief discussion of the cross-country variation in each variable. The descriptive statistics are presented in Table 2.
The TEA necessity rates in 2002 are highest in Brazil, Argentina, India and Chile. In these countries the proportion of the adult population involved in necessity entrepreneurship is between 6.5% and 7.5%. This compares to France and many Scandinavian countries where necessity entrepreneurship rates are very low. In France, for example it is only 0.09% of the adult population. In Norway, Denmark and Finland the rate varies between 0.33% and 0.43%. Thus there is considerable dispersion in necessity entrepreneurship rates around the mean of 1.95%. For 2001, India has the highest rate of 7.69% compared to a global mean in our sample of 2.53%. Mexico also has a high rate at 7.11%. This contrasts to countries such as Norway and Denmark, again, and Netherlands who all had very low rates. Once again we observe considerable dispersion around the mean for all countries of 2.53%.
Regarding unemployment rates across countries and over time, we observe substantial variation in both, within and across countries. One notable feature is that the Netherlands has persistent, and low unemployment rates between 2000 and 2002. The mean rates for all countries for these three years are 5.8%, 7.7% and 7.5%. In 2000, Iceland has the lowest rate at 1.4%. This contrasts with 14.1% in Spain, and 10.6% in Italy. For 2001 the Netherlands has the lowest rate at 2.5% and South Africa the highest at 29.5%. In 2002, Iceland has the lowest rate at 2.3% and Croatia the highest at 23.4%. This highlights the volatility of unemployment rates within countries over time, but more dramatically across countries.
Youth unemployment, measured here as the proportion of the total stock of unemployed under the age of 25, is of considerable concern to many governments. It has become more important given the general ageing of populations across many countries and the need for an ever decreasing number of workers to provide the tax revenues to finance the state burden of welfare payments to retired workers (Peters, Cressy, Storey, 1999; Cowling and Greene, 2002; Gruber and Wise,1998). On this, we note that the country average is around 30% of the unemployed. However, in some countries this is much higher. For example, in Brazil, youths account for 51.4% of total unemployment. In Thailand the comparable figure is 43.4%, and in Israel 43.1%. Yet in Germany this figure is only 12.8% and in France 16.7%.
Next we consider the proportion of the adult population with pension provisions. On average 64% of adults have pension entitlements to secure an income in old age. Yet in India this figure is only 7.9%. In Thailand and China this is 17% and in Mexico 31%. Countries where nearly all the adult population have pension provisions are notably Switzerland at 96.8% and Japan at 92.3%.
Social welfare expenditure by governments reflects the relative generosity of the state in supporting the incomes of those out of work. Here we standardise this form of expenditure to take account the actual stock of unemployed people. Thus countries with high expenditure and low levels of unemployment will have more generous welfare payment systems. Using this measure, we note that China, India and South Africa have the least generous welfare systems. Interestingly, Denmark and New Zealand are relatively ungenerous too. This contrasts with the Netherlands and Mexico who both have very generous welfare systems.
Moving on to consider our first difference variables, and focusing on necessity entrepreneurship, we observe that between 2001 and 2002 necessity rates fell, on average, by 26 basis points. Yet in Israel it increased by 128 basis points, in Norway 50 basis points and the Netherlands 32 basis points. The largest falls were recorded in France, -93 basis points, and Japan, -77 basis points.
For changes in unemployment rates we observe that for both periods measured, 2000–2001 and 2001–2002, unemployment grew by 8 basis points. In the earlier period, the US recorded the highest growth at 20 basis points. Hungary and Slovenia reported falls of 14 and 16 basis points respectively. Over the period 2001 to 2002 Iceland saw unemployment growth of 64 basis points, Taiwan of 52 points, Croatia of 44 points and the US of 29. Over the same period Brazilian rates fell by 32 basis points and Korean rates by 21 points.
Having presented the data together with a concurrent discussion, we now move on to the multivariate part of our analysis.
Table 3 reports the results of four models which estimate the determinants of the TEA necessity entrepreneurship rate for 2002. Each of the models is well specified and can explain between 85% and 90% of the variation in TEA necessity rates across the GEM nations. The first point of interest is that there is a degree of short-run persistence in necessity entrepreneurship rates. It says that nations with high (low) rates one year are likely to have high (low) rates in the following year. Thus for a nation such as India, with a 2001 necessity rate of 7.69% compared to the GEM average rate of 2.53%, holding all other factors constant, will have a necessity entrepreneurship rate of the order of 2% higher in 2002. This might imply that there are cultural differences across countries, which means that adults in certain countries will always be more likely to consider starting their own business as a response to unemployment.
Next we focus on the effects of unemployment, here proxied by the unemployment rate over time. The first point of note is that the coefficients for unemployment rates in 2001 and 2002 are robust to alternative model specifications. What the models do show quite clearly is that people’s response to observed unemployment is dependent upon the stage they are at in the business inception process. For example, if unemployment was high in the previous year, this will stimulate people to choose necessity entrepreneurship in the face of declining wage opportunities. Yet if unemployment is high in the current year this will act as a deterrent to people seeking to start their own business. Thus it would appear that demand-side effects, i.e low demand due to high unemployment, dominate when people are further down the path to starting their own business. At earlier stages, labor market effects appear to dominate, i.e not seeing any feasible alternatives in the labor market. Interestingly, the coefficients suggest that the strengths of these two effects are equal, but opposite. Finally, in Model 3 we also observe that unemployment rates lagged two years (i.e in 2000) act to reduce necessity rates in 2002. However, the coefficient is much smaller than those for 2001 and 2002 and is significant at only the five per cent level.
Youth unemployment, defined here as the share of under 25s out of total unemployment, was found to have a small, but positive, effect in Model 1. This suggests that in countries such as Brazil, Thailand and Israel with high shares of youth unemployment necessity entrepreneurship rates will be higher than in countries such as Germany and France with low rates, holding other factors constant. Overall, this suggests that the age composition of the stock of unemployed matters in terms of flows into necessity entrepreneurship.
We also observe, in Model 4, that countries where social welfare payments are more generous have lower necessity entrepreneurship rates. This directly relates to the standard economic model of labor supply, which predicts that individuals with higher incomes from non-work sources will have lower participation rates. It is consistent with the notion that high levels of welfare payments raises the reservation wage (the wage individuals are prepared to work for) of unemployed people thus making it more difficult to earn an income from running your own business that exceeds the welfare payment level.
Finally, we observe that ease of market entry by new businesses raises necessity entrepreneurship rates. The ability of new businesses to gain a foothold in markets appears critical, particularly for those starting from unemployment who are likely to have very scarce resources, both in terms of financial and human capital. If barriers to market entry are high then necessity entrepreneurship rates will be lower. Thus competition policy, which ensures a level playing field, may be critical for the development of new business activity by the unemployed.
Next we turn our attention to estimating first difference models. These are shown in Table 4, which estimates the change in necessity entrepreneurship rates between 2001 and 2002.
We observe that both models are statistically significant and explain between 48% and 65% of the observed cross-country variation in growth rates of necessity entrepreneurship. Next we note that countries with high levels of necessity entrepreneurship in previous time periods will have slower growth rates in the current period. This presumably might occur as the stock of potential necessity entrepreneurs is diminishing. We also observe that high growth in the unemployment rate in the previous period also reduces current growth rates of necessity entrepreneurship. This is likely to be a demand-side effect. Yet the actual rate (the level) for the last period acts in a positive way on current growth of necessity entrepreneurship. Taken together, these results suggest that high, but stable or falling unemployment is the key to higher growth in necessity entrepreneurship rates.
We also find that high levels of youth unemployment act to increase the necessity entrepreneurship growth rate, suggesting that young, unemployed people see little opportunity for waged employment. Finally, we note that pension provision is negatively associated with growth in necessity entrepreneurship, although this effect is not significant at conventional levels (significance 0.11).
Prior to our concluding section we now present a summary table of our initial hypotheses and empirical findings as a means of clarifying exactly what our results show. This is presented in Table 5.
From Table 5, we observe that our predictions are supported more in our first models estimating the actual rate of necessity entrepreneurship than in the growth models. From the levels models we generate 3 correct predictions, one null effect and one, on the unemployment rate effect, inconclusive. By contrast, the growth models generate only one correct prediction, three null effects and, once again, an inconclusive effect on our unemployment variables. This suggests that it is much easier to explain why certain countries have higher rates of necessity entrepreneurship than why some countries are observing higher growth in this form of entrepreneurial activity.
We have explored the relationships between necessity entrepreneurship and unemployment in a large number of countries using empirical data collected as part of the GEM project. From the sample statistics we observe that there is tremendous variation in necessity entrepreneurship rates across countries. We also note that there is a lot of temporal variation within countries. By contrast, unemployment rates exhibit less temporal variation. Yet the youth unemployment share varies dramatically across countries, as does the relative generosity of welfare systems, and the share of the adult population with pension provision.
From our multivariate analysis, there appears to be short-run persistence in necessity entrepreneurship rates. Countries with high rates last year will have high rates this year. Unemployment has varying effects. Current high unemployment will reduce necessity entrepreneurship rates. Previous high unemployment will increase next period necessity rates. The key to high growth in necessity entrepreneurship rates appears to be a high, but falling level of unemployment. Further, the higher the share of total unemployment accounted for by youths (under 25s) the higher the necessity entrepreneurship rate. There is a hint that generous welfare systems reduce flows from unemployment into necessity entrepreneurship. Also there is a glimpse that ease of entry to the market facilitates necessity entrepreneurship.
As to potential explanations for our findings, it appears that in some countries, more than others, there is a culture of the unemployed seeking to create their own employment when waged jobs are scarce. Alternatively, being unemployed is so undesirable, due to low levels of benefits, that starting a business is the only way of earning a living.
There is also an interesting time dimension to the impact of unemployment on necessity entrepreneurship. If unemployment is high when I am deciding whether to start a venture then I am more likely to see no viable outside alternatives (i.e., no waged jobs on offer). But when I am actually at the point of starting for real, if unemployment is high I see low demand for my goods or services, so I don’t actually start. Even with high levels of unemployment, it appears that if people observe that unemployment has peaked, or is falling, then this fact alone appears to be enough for them to start a new venture. This is likely to be an expectations effect of increased demand for goods and services as people become re-employed.
Youths who are unemployed typically lack education and job skills. Thus job search is futile. The only way for them to work is by starting their own business. With scarce resources, as the unemployed typically have, ease of market entry is critical.
CONTACT: Marc Cowling, Foundation for Entrepreneurial Management, London Business School, London NW1 4SA, England; email@example.com
Abell, P, Khalaf, H, Smeaton, D (1995) An Exploration of Entry to and Exit from Self-employment. Centre for Economic Performance, London School of Economics, Discussion Paper No.224.
Acs, Z, Audretsch, D, Evans, D (1994) Why Does the Self-employment Rate Vary Across Countries and Over Time? Centre for Economic Policy Research, London School of Economics, Discussion Paper No.871.
Blanchflower, D (1998) Self-employment in OECD Countries. Paper presented at the Canadian International Labor Network Conference. Burlington, Canada. 24–26 September.
Bogenhold, D, Staber, U (1991) The Decline and Rise of Self-employment. Work, Employment & Society, 5 (2). 223–239.
Cowling, M (2003) Creating Local Opportunity. Department of Work and Pensions, Research Report. www.dwp.gov.uk/asd/
Cowling, M, Greene, F (2002) Evaluation of the PRIME Third Age Self-employment Pilot Scheme. PRIME, London.
Cowling, M, Hayward, R (2000) Out of Unemployment. Research Centre for Industrial Strategy. University of Birmingham Business School. ISBN 0704 421577.
Cowling, M, Mitchell, P (1997) The Evolution of UK Self-employment: A study of government policy and the role of the macroeconomy. The Manchester School, 65 (4). 427–442.
Gruber, J, Wise, D (1998) Social Security and Declining Labor-force Participation: Here and Abroad. American Economic Review, 88 (2).
Meager, N, Bates, P, Cowling, M (2003) Business Start-up Support for Young Adults Delivered by the Prince’s Trust: A comparative study of labor market outcomes. Department of Work and Pensions, Research Report URN 293.
Peters, M, Cressy, R, Storey, D (1999) The Economic Impact of Ageing on Entrepreneurship and SMEs. Forward Studies Unit, European Commission.
Robson, M (1996) Macroeconomic Factors in the Birth and Death of UK Firms. The Manchester School, 64 (2). 170–188.
Robson, M (1998) The Rise in Self-employment Amongst UK Males. Small Business Economics, 10. 199–212.
Robson, M (1998) Self-employment in the UK Regions. Applied Economics, 30. 313–322.
Taylor, M (1996) Earnings, Independence or Unemployment: Why become self-employed? Oxford Bulletin of Economics and Statistics, 58 (2), 253–265.
© 2004 Babson College. All rights reserved. Last updated May 2004.