An estimated model of entrepreneurial choice under liquidity constraints

Transcription

1 An estimated model of entrepreneurial choice under liquidity constraints Evans and Jovanovic JPE 16/02/2011

2 Motivation Is capitalist function = entrepreneurial function in modern economies? 2 Views: Knight: Capital markets provide little capital to entrepreneurs. Wealth is needed to start-up a business. Schumpeter: Functions of entrepreneurs and capitalist are separate. Entrepreneurs job is to find arbitrage opportunities in the economy. Capitalist will bear the risk for the entrepreneur. Question: Do liquidity constraints hinder people from starting a business? Strategy: Estimate a model of entrepreneurial choice decision in which the tightness of the liquidity constraint is a parameter

3 Model Static. θ is the managerial ability. Individuals observe θ and their wealth and decide whether to be entrepreneur or worker. A worker earns ω = µx γ 1 1 x γ 2 2 ξ where x 1 is working experience, x 2 is education, µ is a constant and ξ is a disturbance. An entrepreneur earns y = θk α ɛ, where k is the amount of capital invested in the business and ɛ is a disturbance. Net income of an entrepreneur is: y + R(z k) where R = 1 + r and z is the wealth. Agents can no default on their debt. Borrowing constraint: 0 k λz, where λ 1

4 Decision An entrepreneur maximizes (1) max k [0,λz] θkα + R(z k) In an interior solution: k = ( θα R ) 1/(1 α) An entrepreneur is unconstrained if θ (λz) 1 α R α The agent will choose entrepreneurship iff: (2) max[θk α + R(z k)] µx γ 1 1 x γ 2 2 Proposition 1: A person with characteristics (θ, z, x) chooses entrepreneurship if θ (f (z, x), ) where f (z, x) is the value that solves (2) as an exact equality.

5

6 Joint distribution of θ and z They test the correlation between θ and z They assume θ is independent of everything except z The regression they run is: ln(θ) = δ 0 + δ 1 ln(z) + η θ N(δ 0 + σ 1 ln(z), ση 2 ) δ 1 could reflect: Greater past savings of those with high θ Lower absolute risk aversion of wealthy people. They find that δ 1 is negative. High asset people are poor entrepreneurs. Model is mispecified

7 Data They use the National Longitudinal Survey of Young Men (NLS) This include mens of age 14 to 24 in They were queried periodically between 1966 and Look at males who were worker in 1976 and keep as a worker or become self-employed in Do not take into account workers who become unemployed. The proportion of self-employed increases with ages from 2.5 % for males 21 to 25 years old to 11.2 % for males of 36 years old or older The average self-employed man earned less than the average wage earners. Also the increase of earnings was smaller for those who switch to self-employment compare to those who stayed as wage workers.

8 Probit estimations If one assumes zero correlation between assets and entrepreneurial ability the model has 2 implications: A positive correlation between the probability of starting a business and assets will exist iff there are liquidity constraints. (table 2) There is a positive correlation between entrepreneurial earnings and initial assets since wealthier people have started businesses at a more efficient capital levels. (table 3) Finally as you can see comparing column 1 and 2 in table 3 assets is not a good proxy for entrepreneurial ability since in the 1981 regressions is not significant anymore.

9

10

11 Maximum Likelihood Estimates From table 3 you can see that education and experience are not statistically significant for entrepreneurial earnings. So Entrepreneurs earnings are assumed to depend only on assets. Wage earnings are assumed to depend on education and experience. The estimated δ 1 which measures the correlation between entrepreneurial ability and assets is negative and statistically significant. Another proof that asset is not a good proxy for entrepreneurial ability. In column 2 they set δ 1 = 0 and find a λ = 1.44 which supports the view of Knight (λ = 1) over Schumpeter (λ = ). The estimate of σ ɛ σ θ. This tells us that entrepreneurs face considerable risk. Surprisingly σ ɛ σ ξ

12

13 Impact of Liquidity Constraints (LC) They find that only high ability/low-asset people are affected by the wealth constraint. But is precisely these people who are most likely to switch to self-employment. They estimate the average probability of being constrained or unconstrained entrepreneur. They find that the first is 3.75 % and the second is 0.06%. So most entrepreneurs are constrained. LC reduces de capital flowing to entrepreneurship in 2 ways: It prevents some people trying entrepreneurship. If we remove it the average probability of becoming entrepreneur would increase to 5.11 % instead of the actual 3.81 % Entrepreneurs now use less capital than the optimal amount. They estimate that the total lost because of LC was $2.7 billon in 1976 dollars (about 2% of gross domestic investment in 1976).

14 Conclusions They develop and estimates a model of entrepreneurial choice under liquidity constraints. They find that liquidity constraints bind for people entering entrepreneurship. The latter let them reject the Schumpeter view that entreprenurial and capitalist functions are separate and accept the Knight view.