# «Abstract We explore which ﬁnancial constraints matter the most in the choice of becoming an entrepreneur. We consider a randomly assigned welfare ...»

This can be shown by substituting sw = 0 into (2) and (3) and combining (5) and (8).8 7 Our results would hold with less extreme assumptions on savings or borrowing constraints.

8 Notice that no individual who is marginal in the occupational choice sets se = 0 and sw 0. In fact, due to DARA utility, we have u (s + w + C2 ) pu (s + y + C2 ) + (1 − p)u (s + C2 ) when E = 0.

Those for which (11) and (12) hold set se = sw = 0 and we are back to the case presented in equation (10). Also those for whom only (12) holds (and so they set se = 0 and sw 0) are more responsive to future than to current transfers. This can be shown by substituting se = 0 into (2) and (3) and combining (4) and (12).9 As in the previous section, the eﬀect of changing C1 and C2 depends on the fraction of the population who can optimally set its savings or borrowing and the fraction for whom either borrowing or saving constraints bind. Still, we can say that, in this setting, the share of self-employed in period 1 is more responsive to period 2 than to period 1 transfers. The reasons is that, in order to self-insure, households need to have enough wealth in period 2. In case of binding saving constraints, they cannot transfer wealth from period 1 to period 2, which makes them insensitive to C1. At the same time, those with binding borrowing constraints consume all their wealth in period 1 (and still they would prefer consuming more). Hence, increasing C1 does not make them richer in period 2 and so does not aﬀect their willingness to take risk.

** As a summary of the above results, we state the following Proposition.**

Proposition 1 Suppose individuals face constraints in allocating transfers across periods. Then current occupational choices are more responsive to the size of current transfers if liquidity constraints bind, while they are more responsive to the size of future transfers if insurance constraints bind.

4.2 Empirical Strategy In what follows, we restrict our attention to eligible salaried workers residing in treated villages.10 We are then interested in evaluating how the probability to become entrepreneur nei,t depends on the amount of transfers received by household h in the previous six Hence, imposing sw = se = s would imply that this individual would save too much when salaried and so he would always set se ≥ sw.

9 Again, for the same argument developed in footnote 9, the case se 0 and sw = 0 is irrelevant as inconsistent with se ≥ sw.

10 Unemployment seems unlikely to be driven by risk preferences in our setting.

13 months Ch,t and on the transfers they know they will receive in the next six months Ch,t+1.11 As an example of the variation in transfer amounts, suppose we are at the end of the academic year and we consider two households with a 15 years old daughter as a single child. In the ﬁrst household the daughter is enrolled in the eighth grade and so, according to the rules described in Section 2, the household has received 2130 Pesos in the previous six months and will receive 2250 Pesos in the following six months. In the second household, the daughter is enrolled in the ninth grade, and as a result the household has received 2250 Pesos in the previous six months but will only receive 630 Pesos in the following six months (since as mentioned after the ninth grade children are no more eligible for the educational component of the transfer). We then ask in which household adult members are more likely to become entrepreneurs. Of course, this is just one of the several discontinuities in transfer amounts induced by the program’s rules. In what follows, we pool those discontinuities across program eligible children by deﬁning for each household the potential transfers it is entitled to receive.

We use potential transfers and not actual transfers since, beside being possibly measured with error, the latter partly depend on the household’s behavior with respect to children enrollment, and this is likely to be simultaneously determined with occupational choices. We deﬁne potential transfers Ph,t and Ph,t+1 as the amount of transfers a household would be entitled to if its children did not change their pre-program enrollment decisions and, when enrolled, progressed by one grade in each year. These transfers are deterministic functions of children’s characteristics at baseline and by construction they are uncorrelated with any behavioral response to the program.12 We then estimate the following model using alternatively current and future potential

**transfers as explanatory variable:**

nei,t = α1 Ph,t + Childh,t β1 + i,t, (13) nei,t = α2 Ph,t+1 + Childh,t β2 + ui,t, (14) where the vector Childh,t contains age-speciﬁc categorical variables for the number of boys 11 Six months correspond to the shortest time frame we can deﬁne such that future transfers are systematically diﬀerent from current transfers according to the school calendar year. We later consider a one-year time horizon for robustness. Also, since we do not know exactly the date in which individuals have changed occupation between two survey waves, current and future transfers are constructed by taking the month of the interview as the reference. It follows that our future amounts are certainly received after individuals have changed occupation, while part of our current amounts may sometimes still be due at the time in which they switch occupation. If this were the case, our estimates on the diﬀerential eﬀects of future vs. current transfers should be interpreted as a lower bound.

12 The average potential transfers received in the past six months are 1446 Pesos (std. dev. 863) and the average potential transfers to be received in the next six months are 1553 Pesos (std. dev. 964).

14 and girls who are between 6 and 17 years old in each household h and post-treatment period t. This controls for any independent eﬀect of children demographics on occupational choices.

The key identifying assumption for estimation of the α parameters in equations (13) and (14) is that, absent the program, occupational choices respond to children demographics and not to the speciﬁc school grade in which children are enrolled. Therefore, partial variations in potential transfers across households with children of the same age but attending diﬀerent grades should be exogenous. To test this assumption, we look at two alternative placebo samples: program-eligible households living in control villages and non-eligible households living in treated villages. We construct the transfers they would have been entitled to had they been treated, and look at whether entry into selfemployment is directly aﬀected by these transfers. If this were the case, our approach would be invalid since occupational choices would be driven by the exact household characteristics that determine the transfers rather than by the transfers themselves. As shown in Table 7, however, estimates reveal no direct eﬀect of potential transfers on occupational choices in these samples.

Finally, instead of estimating the eﬀects of current and future transfers separately, we directly test for their diﬀerential impact on the probability to become entrepreneur. For

**this purpose, we consider the following alternative speciﬁcation:**

where Dh,t is deﬁned as the diﬀerence between future and current transfers, Ph,t+1 − Ph,t.

4.3 Results In order to ﬁrst provide a visual inspection of our relationships of interest, we estimate equations (13) and (14) non-parametrically. As shown in Figure 2, the shape of the curves suggests that current transfers do not have any eﬀect on the probability to become entrepreneur. On the contrary, this probability seems to depend positively on the amount of transfers that households are entitled to receive in the near future (especially after about 2000 Pesos).

These patterns are conﬁrmed in standard probit estimation of equations (13) and (14). In Table 8, we report the marginal eﬀects of current and future cash transfers on the likelihood to switch from salaried work to self-employment. Columns (1)-(2) display the results for the transfers received in the last six months. There is weak evidence in favor of a positive eﬀect, which however vanishes once control variables are included. This reveals no signiﬁcant eﬀect of current transfers on the probability to become entrepreneur.

15 Columns (3)-(6) report the results for future transfers using either a 6-months or a 1-year horizon. The size of future transfers appears a signiﬁcant determinant of the probability to switch to self-employment. This eﬀect is substantial: a one standard deviation increase in 6-months future transfers increases the average probability to become entrepreneur by 1.2%. This amounts to a 12% increase vis-`-vis the average share of new entrepreneurs in a this sample (9.6%). In relative terms, the corresponding eﬀects for 1-year future transfers are similar: a one standard deviation increase leads to 0.9% more self-employed, which is a 10% increase.

In order directly estimate any diﬀerential impact of current vs. future transfers, and so provide a sharper test of liquidity vs. insurance channels, we now turn to the model in equation (15). As shown in columns (1)-(2) of Table 9, these estimates provide evidence that the probability to become entrepreneur is signiﬁcantly more responsive to the amount of future transfers than to the amount of current transfers. In terms of magnitude, a one standard deviation increase in the diﬀerence between future and current transfers (equal to 0.42) increases the probability to shift to self-employment by 1.2%, which matches our previous estimates in levels. Moreover, in columns (3)-(6), we have included the amount of current transfers and the diﬀerence between future and current transfers in wave 1, respectively, in order to compare similar households in terms of children demographics

**that are facing an upward or downward stream of transfers. Results barely change:**

households facing an increasing stream of transfers are on average more likely to switch occupation and become self-employed.

It is also interesting to notice that the magnitude of these eﬀects is consistent with the reduced-form treatment impacts described in Section 3, in spite of the fact that they arise from two potentially diﬀerent sources of variation. For salaried individuals in treated villages, the treatment increases the probability to become entrepreneur by 1.5% with respect to the control group (see Table 2, column 4), while a standard deviation increase in the diﬀerence between future and current transfers increases such probability by 1.2%.

This suggests that the time proﬁle of the transfers is key for explaining the program eﬀects on occupational choices.

To summarize, these results tend to support the hypothesis that the cash transfers have been eﬀective in promoting micro-entrepreneurship as they have enhanced the willingness to bear risk as opposed to simply relaxing current liquidity constraints.

5 Conclusions We have explored the response of occupational choices to the income shocks induced by the Mexican program Progresa. We have ﬁrst documented that the probability to become 16 entrepreneur increases by about 20% for treated households. We have then shown that current occupational choices are signiﬁcantly more responsive to the amount of transfers expected for the future than to the amount of transfers currently received. Moreover, according to our estimates, the diﬀerential impact of future vs. current transfers is comparable in magnitude to the treated-control diﬀerence, which conﬁrms that the time proﬁle of the transfers is key in explaining the program eﬀects. We have interpreted these results as evidence that in our setting insurance constraints are fundamental determinants of the choice of becoming entrepreneurs.

Our results feature some limitations. For example, little is known on the long run eﬀects of these dynamics. In a related study, Gertler, Martinez and Rubio-Codina [2006] argue that productive investments induced by Progresa had persistent eﬀects on individual welfare. We conjecture that changes in occupational choices are likely to display similar features, but a detailed analysis of this issue is left to further investigation. Moreover, we have not fully addressed the possibility of general equilibrium eﬀects induced by the program. As a ﬁrst step, we have shown that indirect eﬀects on non-eligible households in treated communities are not signiﬁcant. However, we cannot say whether the above described dynamics are only improving the welfare of those who have changed occupation, or they are also altering the functioning of some markets (e.g. in terms of increased labor demand or total production).

Nonetheless, we think our analysis can inform the debate on ﬁnancial constraints and entrepreneurship in developing countries. First, we have shown that it is possible to promote welfare-enhancing entrepreneurship.13 Second, according to our estimates, ﬁnancial barriers to entry into entrepreneurship do not seem insurmountable. Instead, a major barrier may come from the risky prospects self-employment oﬀers. In this view, promoting entrepreneurship requires reducing households’ exposure to risk in other dimensions.

13 Instead, in face of many failed attempts, skeptics question whether policy makers can promote entrepreneurship at all (see e.g. Holtz-Eakin [2000], Acs and Szerb [2007], Parker [2007], Shane [2009] for a discussion).

**Ashraf, Nava, Dean Karlan, and Wesley Yin. 2006. “Tying Odysseus to the Mast:**

Evidence from a Commitment Savings Product in the Philippines.” Quarterly Journal of Economics, 121(2): 635–672.

Banerjee, Abhijit V. 2003. “Contracting Constraints, Credit Markets, and Economic

**Development.” In Advances in Economics and Econometrics: Theory and Applications:**

Eighth World Congress. Cambridge University Press.

Banerjee, Abhijit V., and Esther Duﬂo. 2005. “Growth Theory through the Lens of Development Economics.” In Handbook of Economic Growth. Vol. 1 of Handbook of Economic Growth,, ed. Philippe Aghion and Steven Durlauf, Chapter 7, 473–552.

Elsevier.