引言

最近的全球經濟危機其在2008年達到頂峰，導致全球國內生產總值(GDP)下降。它也直接導致金融市場的不穩定性和私營部門需求滯后。其后果尤其是勞動力市場是非常糟糕的。在許多國家里工人失業,工資收入下降,工作時間縮短(分割,2011)是經濟危機的最主要表現形式。在最近的研究之中，Klapper和Love采取了九十三個國家的固定數據從而可以展示很多國家新注冊的公司在經濟危機期間都經歷了一段急劇下滑的經濟周期。女性往往是在經濟沖擊之下最為嚴重的受害者，比如像派內斯,勒納和施瓦茨，我們在這里研究全球性經濟危機對于女性創業的影響，在四十三個國家之內女性參與創業的比率比較底。當女性創業活動可以提供比較有用的見解，但是并沒有在女性創業及經濟危機之中收集到更多具有關聯性的信息，女性創業的喜好對于家庭以及個人生活的影響。

Introduction 

The recent global economic crisis with its peak in 2008 resulted in a decline in global gross domestic product (GDP). It led to unstable financial markets and a lag in the private sector demand (The World Bank, 2010). Its consequences especially for the labor market have been tragic. In many countries workers lost their jobs, wage earnings declined and work hours shortened (LiTS, 2011). In a recent study, Klapper and Love (2011) using panel data for 93 countries show that most countries experienced a sharp drop in new firm registration during the recent global financial crisis. Women are often the hardest victims of an economic shock. Pines, Lerner and Schwartz (2010), examining the impact of the global crisis on female entrepreneurship, find a lower participation rate for women in entrepreneurship in 43 countries. While the cross-country evidences provide useful insights into female entrepreneurial activities, the aggregate findings do not reveal much information on the causal relationship between economic crisis and the propensity of female entrepreneurship at the individual or household level1.

Existing literature suggests that women hit by crisis, face a lower demand for their services and shortage in credit for business (Pines et al, 2010). These factors are likely to dampen the propensity of entrepreneurship for women facing a crisis. However, existing theories on entrepreneurship contend that adverse labor market conditions may force individuals to become necessity entrepreneurs, starting a small business, when alternative livelihood choices are not readily available (Acs, et al, 2005; Allen et al, 2007). Furthermore, Allen et al, (2007) found that necessity entrepreneurship is more prevalent among women, reemphasizing the contention that women engage in self-employment at a higher rate when livelihood choices are limited. This is supported by Arenius and Minniti (2005), who opine that contextual factors along with socio-demographic and perceptual factors influence the complex decision of starting a new business for women.

The use of rigorous impact evaluation in this field has received limited attention (Mckenzie, 2003; 2004; Fallon and Lucas, 2002; Manning, 2000) perhaps due to the unavailability of reliable datasets (Acs et al, 2011; Bruhn and Love, 2012).second round of Life in Transition Survey (2010), we evaluate the impact of the economic crisis on female entrepreneurship across 30 transition countries of Eastern Europe and Central Asia (EECA). We account for the participation rate as well as perception of women towards entrepreneurship. The main motivation of this study comes from the literature on necessity entrepreneurship and the role of contextual factors in entrepreneurship, which we discuss below. Moreover, the transition countries of Eastern Europe and Central Asia were among the hardest hit by the global economic crisis where the crucial transmission channels of the crisis have been reductions in wage earnings and job losses (LiTS report, 2011). This justifies our selection of geographic areas for this study.

Differential economic opportunities and outcomes reflected for instance in lower female entrepreneurship have obvious welfare implications and need to be understood better if the gender parity as part of the Millennium Development Goals is to be met. There is a growing interest in studying the determinants of entrepreneurship especially for women. A variety of factors influence the complex decision of starting a new business for both genders, including socio-demographic, contextual and perceptual factors (Arenius and Minniti, 2005; Kourilsky and Walstad, 1998). Education (Cooper, 1985; Linan, 2004; Rees and Shah, 1986; Robinson and Sexton, 1994; Luber et al. 2000), socialization experience (Hackett and Betz, 1981), exposure to media and expectancies all tend to play a significant role in the setting up of a business for women, more so than they would affect men (Bussey and Bandura, 1999; Ljungren and Kolvereid, 1996; Brown et al., 2006). Personal and family considerations are often more important than economic conditions for women in setting up business (Anna et al, 2000; Orser and Hogarth-Scott, 2005). Contextual factors also play an important role. Verheul et al. (2012) using the theory of planned behavior show that a lower preference for women to become self-employed largely explains their relatively low involvement in self-employment (Minniti et al., 2005; Reynolds et al., 2002).

In this study, we use individual level data from the second round of the Life in Transition Survey (2010). We test the significance of the recent economic crisis as a contextual factor in determining female entrepreneurship. This study considers existing female entrepreneurs, prospective female entrepreneurs and nascent female entrepreneurs separately. This allows us to examine the impact of the crisis at different stages of entrepreneurship. Our baseline identification strategy builds on the direct effect of crisis at the individual level. We use a dummy variable which takes the value of one if respondents were reported to be affected by crisis. In addition, we use a set of victimization indicators as dummy variables identifying various aspects of crisis affecting their livelihood status negatively. It includes direct effects of crisis such as losing jobs, experiencing wage cuts, receiving lower volume of remittances and working for fewer hours, among others. We use standard OLS fixed effect model as our baseline empirical strategy.

Empirical findings indicate that the entrepreneurial propensity of women is higher among households that are the hardest hit by crisis. While the outcomes are significant statistically, it fails to yield robust outcomes across the victimization indicators. We perform a variety of robustness checks. To address the existence of sample selection bias, we employ propensity score matching. We also perform empirical analysis across various subsamples comprising of individuals from countries that experienced crisis at a high level, at a low level; also countries that has high entrepreneurship and low entrepreneurship rate. Additionally, we test for the same relationship between female headed and male headed households. Finally, to check for omitted variable bias, we use a test to identify the relative strength of observables against unobservable in our empirical models following Altonji, Elder, Conley and Taber (2005). The findings from these tests do not alter the main findings. The empirical outcomes show a strong correlation between the propensity of entrepreneurship for women and the direct victimization of crisis, often suggesting a causal relationship. Entrepreneurship has been a part of active household coping strategies for women during and after the crisis. An additional test on this relationship controlling for existing entrepreneurial activity within the household undermines our main findings. We find that women are more like to attempt to setup an enterprise of their own in the awake of a crisis, if someone else in the household already engaged in entrepreneurial activity. However, while this weakens our main hypothesis, it does not negate it; the tendency for the Schumpetarian hypothesis to be prevalent among female members in households without prior entrepreneurial activity can be thought of as the lower bound for such creative destruction. Overall, this study provides empirical evidence supporting that the recent global crisis perhaps worked as contextual factor and contributed to the growth of female entrepreneurs during and after the crisis in Eastern Europe and Central Asia.

The rest of the paper is organized as follows. In section 2, we highlight the key features of female entrepreneurship in transition countries. Section 3 provides a brief description of the LiTS (2010) data and descriptive evidence using the perception of respondents’ on the recent economic crisis and entrepreneurship. Section 4 discusses the main findings of the baseline econometric model. In section 5, we provide empirical outcomes of robustness check tests. This is followed by a concluding section summarizing the main findings and prospects of future research.

Female Entrepreneurship in Transition Countries Traced over time There is worldwide evidence that women are less likely to engage in entrepreneurial activities than men (Minniti et al., 2005; Reynolds et al., 2002). Transition countries are no exception to this. In a recent study, Nikolova et al. (2012) using LiTS (2010) data found that women are less likely to attempt to set up a venture however no less likely to succeed than men once they try. The authors posit that the lower participation of women in entrepreneurship could be due to various factors. Paul and Sattar (2009), using the first round of Life in Transition Survey (2006) data for 27 transition countries, found similar evidence of an overall gender gap of 6.5 percentage points against women in selfemployment.

In retrospect, the beginning of the transition era was characterized by the demise of the centrally planned economic system followed by a process of economic, social and political transformation. One of the major changes was the development of private business ownership in most of the transition countries in central Asia and Eastern Europe. This was particularly important for transition countries in the short run, in terms of an expected increase in the number of jobs and wealth that small private firms offer for individuals, and also in the long run through potential welfare gains for the economy and society as a whole. However, job growth has been limited and labor force participation rates, particularly for women in entrepreneurial activities, have stagnated and even declined in some of these countries.2 Evidence from national studies also confirm lower levels of womens’ entrepreneurial activity in transition countries in Bulgaria (Stoyanowska, 2001), Hungary (Nagy, 1999) and Poland (Lisowska, 2002).

To get a more recent picture, we reflect upon the Global Gender Gap Index 2008 (World Economic Forum) that ranks 130 countries including 25 countries from the transition region comprising of the Eastern European and the Central Asian countries. The global gender gap index is composed of gender gap in four areas: economic participation, education attainment, health and political empowerment. In the global ranking of 130 countries, transition countries are evenly spread out, with Latvia (globally ranked 10th) being the highest compared to Turkey (globally ranked 123rd) being the lowest among the group of 25 transition countries. This indicates that the transition countries as a region are no different compared to the rest of the world when it comes to gender gap in general.

In Appendix A, we show gender gap in entrepreneurship across 26 transition countries. The horizontal bar diagram depicts female entrepreneurs as a percentage of male entrepreneurs using data from the Business Environment and Enterprise Performance Survey (BEEPS) (2008). The evidence is similar to the World Economic Forum ranking we discussed above. The share of female entrepreneurs to male entrepreneurs ranges from as high as .43 (in Latvia) to as low as .14 (in Armenia). Although no clear sub-regional gender gap story emerges from this, most of the new EU member states fall in the gender gap range of .30 - .40; countries form the Balkan region are in the middle, with a gender gap of .20 to .30; the majority of the Commonwealth of Independent States (CIS) fall within the gender gap range of .10 - .20.

Descriptive Evidence on Crisis and Entrepreneurship We use data from the Life in Transition Survey 2010 (LiTS 2010) which comprises of a sample of 39,000 respondents from 34 transition countries and 5 western European countries. The LiTS 2010 survey was conducted jointly by the European Bank for Reconstruction and Development (EBRD) and the World Bank. For our purpose, we employ only 16,807 observations; as our sample comprises only of female respondents who fall within the working age (See Appendix B). We also exclude respondents from the Western European nations that are included in the latest round of the LiTS survey (LiTs 2010), as our goal is to investigate how the crisis affected entrepreneurial efforts among women in the Eastern European and Central Asian Economies (the so called Transition Economies).

The summary statistics (reported in Appendix C), show that a significant proportion (68%) of the respondents had some form of a secondary education, and another 21% of the respondents acquired some form of education higher than the secondary level.

Thestatistics also reveal that 56 per cent of the sample comprise of female headed households, with an average household size of 3.4, where on average 55 per cent of the household comprised of female members.

Using the LiTS (2010) data we graphically present the perceived effect of the financial crisis for the sample of 30 countries we study in Appendix D. We find a heterogeneous effect of the financial crisis across the sub-sample of countries surveyed in LiTS, similar to findings reported in Nikolova et al. (2012). The reported adverse effect of the economic crisis among female respondents was highest in Serbia, where 77% of the respondents were reported as being affected by the crisis; while only 17% of the Uzbek respondents made a similar claim, showing the wide disparity of the perceived effect of the crisis. While most of these countries are likely to have experienced the crisis to an almost similar extent; the perception of the experience is subjective, and sometimes reflects the sociocultural factors that defines perceptions (EBRD, 2011; Nikolova et al., 2012).

Appendix E indicates that the effect of the crisis across male headed and female headed households was almost homogenous. The table indicates that about a half of the respondents from both cohorts identified themselves as being affected by the financial crisis. This is in contrast to findings reported by Pines et al. (2010) who, using GEM (General Entrepreneurship Monitor) 2007 data find that “women (entrepreneurs) are more impacted by the crisis than men”. Our data shows that about a third of the sample (among both male and female headed households) reported having experienced wage reductions during and in the aftermath of the crisis and more significantly 12% of the households reported that the head of the household lost his/her job due to the financial crisis. The table also indicates that about 15.5% of the households and 12.5% of the households experienced delayed/suspended wages and reduced remittance inflows respectively.

Appendix F divulges the self-reported perspective of male and female headed households towards self-employment. Again, the results reported in Appendix F indicate there is no evident distinction in the attitude towards self-employment among male and female headed households across the sub-sample of countries surveyed in LiTS. While only 7 per cent of the male headed households and 6 per cent of the female headed households were at the time of the survey self-employed, about 40 per cent of the sample from both cohorts shared their willingness to be self-employed in the future. Nikolova et al. (2012) highlight access to capital as the biggest barrier for those willing to start their own enterprise; and especially during the financial crisis the severity of this barrier increased adversely.

On the other hand, Figure 3.1 below, maps the perceptive response of the proportion of respondents who wanted to become self-employed for each of the surveyed countries, against the proportion of households that were affected by the financial crisis in each of those respective countries. The graph depicts an inverse relationship with a correlation of .44 between the proportion of affected households and the proportion of respondents who wanted to be self-employed. This implies that being affected by the financial crises subdued the desire to be self-employed, most likely as a result of the economic climate in the worst affected countries such as Serbia, Hungary and Bulgaria. On the other hand, we see a better response in the desire for self-employment in countries such as Belarus, Kyrgyzstan and Uzbekistan where the effect of the crisis was perceived to be lower.

Once we compare the level of self-employment prevalent in countries in 2006 to the levels of self-employment prevalent in the same countries in 2011 (as depicted in Appendix G), it is evident that to a certain degree, that the impact of the financial crisis in fact increased self-employment in the worst affected countries. However, while this did not apply universally across the sub-sample of 30 countries, the evidence is more apparent in certain countries than the rest. For example, Albania recorded the largest increase in the proportion of self-employed from 9 per cent in 2006 to 15 per cent in 2010, at a time when 58 per cent of the Albanian households reported being affected by the financial crisis. In contrast, Uzbekistan which reported the lowest number of crisis affected households of 17 per cent also ironically saw the largest drop in the proportion of self-employed across the same four year time period, from 27 per cent to 10 per cent.
As pointed out earlier, not all countries conformed to this positive relationship between the proportion of crisis affected households and the proportion of self-employed. In Kyrgyzstan and Russia for example, while only 26 per cent and 30 per cent of the households respectively were affected by the financial crisis, the proportion of selfemployed increased from 9 per cent to 13 per cent in Kyrgyztan and more dramatically from 1 per cent to 6 per cent in Russia. Similarly, while two-third of the Latvian households and three-fourths of the Belarusian households reported as being affected by the financial crisis, the proportion of self-employed in fact halved in Latvia and dropped by a third in Belarus.

Appendix H summarizes the matrix of growth incidence in female entrepreneurship to the proportion of crisis affected households with a threshold of 40 per cent. The table does not report 3 countries (Bosnia, Moldova and Turkey) within the sub-sample for whom there is no clear indication of which matrix they fall into due to the low association between the two variables. These results from the table show a lack of clear evidence on the effect of exposure by the financial crisis on female entrepreneurship at the aggregate level, and therefore warrant further investigation at the micro level for better understanding of the transmission mechanisms and the socio-cultural factors that shape the decision for selfemployment.

Baseline Empirical Model and Findings As our baseline econometric strategy, we use the following binary dependent variable model to examine the determinants of the propensity to self-employment for women in 34 transition countries:

We employ a set of dependent variables that distinguish between prospective female entrepreneurs and nascent female entrepreneurs. This allows us to examine the impact of the crisis at different stages of entrepreneurship. In the above equation, ‘y’ represents our dependent binary variables – the first variable, ‘willingness to be self-employed’ takes the value of one if the individual was willing to be self-employed, and zero otherwise, and the second variable ‘attempted self-employment since 2007’ takes the value of one if the individual had attempted to be self-employed since 2007, and zero otherwise. x represents a vector of covariates and is the corresponding coefficient vector. Random factors as well as unobservable factors are captured by the error term also includes a vector of independent variables that capture the channels of crisis victimization for the individual’s household. These include whether the crisis affected the household, whether any member of the household experienced wage reductions, whether the head of the household experienced a job loss, whether a household member experienced delayed or suspended wages, whether the household experienced a reduced flow of remittance, and finally whether a member of the household experienced reduced working hours.

For the baseline model we use a probit analysis, where our dependent and independent variables are as outlined above. Columns (1)-(6) of Table 4.1 reports the marginal effects of each of the ‘ x ’ variables on the willingness for women to be self-employed and their attempts at self-employment since 2007. The marginal effects from the probit analysis show that women from crisis affected households were not significantly different in their willingness to be self-employed in comparison to those women from households who were not affected by the crisis. However, the results show that women’s reluctance to be selfemployed increased if a household member experienced delayed or suspended wages. These results fall short of providing conclusive evidence on, if and how the crisis affected female entrepreneurship in these transitional economies, other than hinting that women who were affected by the crisis in general had a lower preference to be self-employed.

The results reported in columns (7)-(12) of Table 4.1 capture the association between crisis victimization and whether the individuals have been actively trying to setup a business since 2007. The results indicate that women were about 1 to 2 per cent more likely to have attempted to be self-employed if they were affected by the financial crisis or specifically if the head of the household had lost his/her job due to the financial crisis. These results are statistically significant at the 1 per cent level. Given that the average rate of female entrepreneurship across these 30 countries was only 6 per cent, a one or two per cent increase in the probability of setting up an enterprise can translate to significant rates of change in the national female entrepreneurship rate.

The contrast in the signs for the key explanatory variables of interest across columns (1)(6) and (7)-(12) indicate that while women in general disliked being self-employed, they in fact attempted to be self-employed (since 2007), possibly as a measure to ease household finances during the hard times. These results indicate the presence of necessity entrepreneurship rather than opportunity entrepreneurship among respondents across the Transition economies. Necessity entrepreneurship is the starting-up of business as a measure of last resort due to the lack of formal employment – it is a form of forced entrepreneurship (Allen et al, 2007; Nikolova et al. 2012). The latter – opportunity entrepreneurship is a genuine willingness to be an entrepreneur, and reflects a dislike for formal employment. Similar to findings in our study, Nikolova et al. (2012) who used the same dataset, report that necessity entrepreneurship is more likely to be prevalent among countries with lower GDP per capita (as those in the Eastern Europe and Central Asian bloc studied in this paper). Our results also conform to findings by Allen et al (2007) who found that female entrepreneurs are more likely to be necessity entrepreneurs rather than opportunity entrepreneurs and as a result are less likely to start-up their own business compared to men. We consider some additional tests below to evaluate the evidence thus far.

Columns (1) and (7) report the marginal effects of the probit estimation for high crisis countries, whilst columns (2) and (8) report the pseudo R2 for the same. Columns (3) and (9) report the marginal effects of the probit estimation for high self-employment countries, whilst columns (4) and (10) report the pseudo R2 for the same. Columns (5) and (11) report the marginal effects of the probit estimation for high crisis, high self-employment countries; whilst columns (6) and (12) report the pseudo R2 for the same. Only regression coefficients of the main crisis victimization index are shown in the table; *** p<0.01, ** p<0.05, * p<0.1. Robust standard errors are used in the estimation. The regressions include country-specific fixed effects, control for rural and urban areas, control for household specific characteristics such as household size, number of children and adult members, number of female and male members in a household, gender of the head of household and an index of household asset and individual specific characteristic such as age, squared-age, educational attainments, an indicator of risk prone behavior at the individual level.

HIGH CRISIS COUNTRIES Since a diverse groups of countries and individuals are accounted for in the pooled baseline regression (despite country, regional and individual controls), the results may not be so profound. Thus, as an additional measure of analysing the transmission mechanism of the financial crisis towards female entrepreneurship we restricted our sample to high-crisis countries. For this purpose, our sample was restricted to countries that reported at least 50 per cent (half) of the households as being affected by the crisis; what we refer to henceforth as high crisis countries. This includes 20 of the 30 sub-sample of countries (Refer Appendix I).
The estimated outcomes for high crisis countries are reported in columns (1)-(2) and(7)-(8) of Table 5.1. The results are not significantly different from those reported in Table 4.1, where women from all 30 countries were included in the estimation. The size of the coefficient for a family member experiencing a reduction in wages is approximately twice as those reported in Table 4.1, and the head of the household’s job loss no longer seems to significantly affect the probability that the respondents would have attempted to setup a business. In general however, the results reported in Table 5.1 are not very different from the results reported in Table 4.1, and this asserts the findings from our baseline estimation.

Marginal effects from the table also indicate that women were about 2 per cent more likely to have attempted to initiate some form of a self-employment initiative since 2007 if they were affected by the financial crisis and more significantly 3 per cent more likely to have attempted the same, if the head of the household experienced a job loss during the crisis. While the signs and significance levels are in agreement with our baseline findings, the size of the coefficient is about 2-3 times larger than those reported in the baseline model. This again reaffirms our argument that women from high self-employment countries are more necessity based entrepreneurs than women from low self-employment countries.

The ratio is calculated using the estimated coefficients from the full probit models that we ran (in Table 4.1) and a restricted version of the probit model (with only geographic and individual controls). For the first dependent variable, the average of the six ratios is close to 14 whereas for the second dependent variable, the average of the six ratios is close to 18 (Appendix K). Thus, on average the selection of unobservable has to be at least 14 times stronger than the selection of observables to explain away the estimated probit results. Thus, it is less likely that the estimated outcomes will be affected by the selection on unobservable.

Conclusion The female participation rate in economic opportunities and outcomes, especially in entrepreneurship shows a grim picture despite the global initiative undertaken almost a decade ago under the Millennium Development Gender Goals (GEM, 2008). Contributing to a growing body of literature that suggests a variety of factors explaining the lower participation of women in entrepreneurship, this paper examines empirically whether the recent global economic crisis serves as a contextual factor providing an impetus to the female participation rate. Our hypothesis transgresses from the Schumpetarian (1939) process of creative destruction, which is also supported by the necessity entrepreneurship theory that inadequate labor market conditions may force individuals to become necessity entrepreneurs.

Using LiTS (2010) data surveyed in 30 transition countries from the Eastern Europe and the Central Asia, we found a positive correlation between the growth of female entrepreneurs and direct exposure to crisis. Satisfactory outcomes on robustness checks suggest a causal relationship and also support the fact that women are generally necessitybased entrepreneurs.

While theories suggest that necessity entrepreneurship is likely to produce less innovation and thereby play a limited role in transcending economic growth than opportunity based entrepreneurship, the former may still play a major role in times of crisis as long as it creates employment. This way, our findings are in line with other studies such as Allen et al. (2007) and Nikolova et al. (2012). Moreover, Nikolova et al. (2012) point out, even though female entrepreneurs are less likely to start with their own enterprise, once they start they are no less likely to succeed than their male counterparts. This conclusion along with our empirical findings retorts the need for policy frameworks that encourage female entrepreneurship and could supplement the necessity based entrepreneurship. This calls for affirmative actions by the governments especially in transition economies to initiate policy directives that aim to encourage female entrepreneurship. As a multiplier effect, such initiatives are likely to benefit not only the families of the female entrepreneurs but also the country as a whole.

Certain caveats deserve a mention. While our results are robust to alternate specifications, our study might not have controlled for an exhaustive set of control variables determining female entrepreneurship. Also, there could be measurement errors in terms of perceptionbased victimization indicators that could suffer from recall bias. There could also be subjective bias respondents’ attitude towards risks related to the economic crisis. Nevertheless, we hope that our findings can be complemented by studies that are able to explore the household dynamics (at the micro level) and the various socio-cultural factors (both at the micro and macro level) that affect the propensity to entrepreneurship for women at a greater depth. This would improve our understanding of how the various transmission channels affect female entrepreneurial decisions. Also, cross-country studies can provide new insights into regional and structural disparities affecting both types of entrepreneurship: necessity versus opportunity based.