The Right Stuff

Who has ‘The Right Stuff’?Human Capital, Entrepreneurship and

Institutional Change in China

Charles E. [email protected]

Stanford UniversityDepartment of Management Science & Engineering

Washington UniversityOct. 7th, 2009

(with support of a Kauffman Foundation Dissertation Fellowship, the Tsinghua Univ. Alumni Association, and the MIT Entrepreneurship Center)

mailto:[email protected]

Strategy / Economics of Innovation and Technological Change

Current Research

Charles Eesley

Motivation19892004 China 29% growth vs. US 1% (State Statistics Bureau)

(Chinese State Statistics Bureau)

• Critical for their role in creating new markets, technologies and products, … process of “creative destruction” (King and Levine, 1993a; 1993b; Djankov et al. 2002; Klapper et al., 2007)

• Acs, Z. J., and D. B. Audretsch, 1988; Christensen; Henderson, Utterback, etc.

• How might we encourage highly talent individuals to found firms?

Charles Eesley - The Right Stuff

“High-impact or high-growth entrepreneurs … in particular, help drive growth in productivity and living standards and, thus, are of special interest and importance.” – Robert Litan, Vice President, Research and Policy Ewing Marion Kauffman Foundation

19921993

19941995

19961997

19981999

20002001

20022003

20042005

20062007

0.00

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0.09

0.10

Proportion Becoming Entrepreneurs by Year (Unobserved Human Capital)

Low human capitalHigh human capital

Founding Year

Prop

ortio

n


Results

Overview

1.Theory2.Hypotheses3.Two reforms4.Empirical context – novel survey

data5.Results6.Conclusion and Implications7.Robustness checks (US data)

Institutional Level Theories• Linked with long-run economic growth

(Acemoglu, 2002; 2005; Johnson, McMillan & Woodruff; Porta, 1998)

• Transition, institutional reform, property rights, financial liberalization, liquidity constraints (Walder, 2003; Gans, Hsu & Stern 2002; Johnson, McMillan & Woodruff; Baumol, 1990; Thornton, 1999; Nee, 1996; Katila & Chen, 2009; Katila & Shane, 2005)

• Neo-institutionalism (Meyer & Rowan, 1977; DiMaggio & Powell, 1983)


Individual Level Theories• Training and prior experience

(Hsu, 2007; Sorensen, 2007; Baumol, 2004; Groysberg, et al. 2007; Burton et al. 2002; Phillips, 2002; Dobrev & Barnett, 2005, Gompers et al, 2005; Simons & Roberts, 2007; Baron et al. 1996; Klepper & Sleeper, 2002; Eisenhardt & Schoonhoven, 1990)

• Networks (Stuart and Ding, 2006; Katila et al., 2008)

• Human capital, preferences(Zucker, Darby & Brewer, 1998; Lazear 2004; Irigoyen 2002, Nanda 2008; Amit et al. 1995; Amit et al. 1990; Knight 1921; Lucas 1978; Jovanovic 1982; Holmes and Schmitz 1990; Jovanovic and Nyarko 1996; Sorensen 2004)

“The only idea common to all usages of the term ‘institution’ is that of some sort of establishment of relative permanence of a

distinctly social sort.” (Hughes, 1936: 180)


Institutional Level

Individual Level

Very little on ability/human capital (Banerjee and Munshi, 2004)


Focus on Human Capital as a Driver of Selection • Talent and labor skills- Roy Model (1951) Selection on Potential

Earnings• Identification requires exclusion restrictions, exogenous shocks, or instruments

affecting returns or skill in one sector only• Jack-of-all trades

• Liquidity Constraints – Evans and Jovanovic (1989)• Assets are endogenous, static models, typically poorly measured• Buera (2008)

• Psychological Traits – Kihlstrom and Laffont (1979), Dunn and Holtz-Eakin (1996), Stajkovic et al (2000)

• Evidence on risk-aversion or tastes for independence is limited

• Network factors (Stuart and Ding, 2006, Ruef, Aldrich, & Carter, 2003, Nicolaou and Birley, 2003)• Demographic factors (McClelland, 1961, Blau & Duncan, 1967Dunn & Holtz-Eakin, 2000; Roberts, 1991)

ModelRoy Model as extended by Borjas (1987)Labor markets 0 and 1, respectively. Log earnings in the wage sector:

w0 = μ0 + ε0

ε0 ~ N (0, σ02)

The wage sector earnings would be the following:w1 = μ1 + ε1

ε1 ~ N (0, σ12)

Assume that the cost of becoming an entrepreneur is C π = C/w0

The correlation between entrepreneur and wage worker earnings is σ01 = cov(σ0, σ1). A worker will choose entrepreneurship if


ModelDefine the indicator variable I, equal to 1 if this selection condition is satisfied, 0

otherwise.Define ν = ε1 − ε0. The probability that a randomly chosen worker from the wage

sector chooses to entrepreneurship is equal to:

Φ (·) is the CDF of the standard normal and z = (μ0 − μ1 + π) /σν

Calculate expectation of earnings in wage and entrep. sectors for those who chose the other sector.

Selection conditions:THREE CASES:Define Q0 = E (ε0|I = 1), Q1 = E (ε1|I = 1)


ModelCase 1: Positive hierarchical sorting:• Entrepreneurs are positively selected from the wage sector distribution and

are also above the mean of the entrepreneurship distribution: Q0 > 0, Q1 > 0. This will be true iff

• First, σ1/ σ0 > 1 implies that entrepreneurship has a higher ‘return to skill’ than the wage sector. Second, ρ > σ0 /σ1, implies that the correlation between the skills valued in the wage sector and in entrepreneurship is sufficiently high

Case 2: Negative hierarchical sorting• In this case, entrepreneurs are negatively selected from the wage sector

distribution and are also below the average of the entrepreneur distribution: Q0 < 0, Q1 < 0. This will be true iff


ModelCase 3: ‘Reverse’ sortingWhere Q0 < 0, Q1 > 0, that is, entrepreneurs are selected from the lower tail of

the wage sector distribution but arrive in the upper tail of the entrepreneurship distribution. This can only occur if

A fourth case?Note that there is not a fourth case where Q0 > 0, Q1 < 0. This would only

happen if an individual from the top of the wage sector distribution joined the bottom tail of the entrepreneurship distribution.


Higher Human Capital

Initial institutional environment

Institutional environment #2

Cost to Start a Business

Increase is among those of relatively lower ability (Nanda, 2008)

Net Returns to Entrepreneurship

Conventional View


Entrepreneurial Returns(net of entry $)



Wage empl. income

Much Less Focus on Returns

Selection on Talent and Expected Earnings


Payoff in low variance sector

Payoff in high variance sector

(Roy, 1951)Individuals form expectations on talent / earnings in each sector

2. Hypotheses

H1: An institutional change reducing barriers to growth will increase entrepreneurship among individuals located relatively higher in the talent distribution.

H2: Individuals who show evidence of higher talent in their wage employment careers will experience higher returns to talent (in entrepreneurship) after an institutional change lowering barriers to growth.

H3: Individuals who show evidence of higher talent in their wage employment careers will start firms that have higher performance.



3. Ideal Experiment

Returns to talent in entrepreneurship relative to wage employment

t

Exogenous increase (decrease) in σ1

Roy Model (Appendix A)

t

1999 – barriers to growth

1988 – barriers to entry

Ideal data is on the relative variance of wage and entrepreneurial income

18

Reform Easing Constraints on Growth

March 1999 An amendment of Article 11 of the Constitution

- Officially ending discriminatory practices against private firms

…places private businesses on an equal footing with the public sector by changing the original clause "the private economy is a supplement to public ownership" to "the non-public sector, including individual and private businesses, as an important component of the socialist market economy" (China Daily, March 16, 1999).

Kellee S. Tsai, “Capitalists without a Class: Political Diversity Among Private Entrepreneurs in China, Comparative Political Studies, 38:9, 2005.

- Private property

- R&D policy (expansion of tax incentives, incubators, science parks)

ReformYingyi Qian, current SEM Dean at Tsinghua University, former

economics faculty at Stanford and UC Berkeley

• …places private businesses on an equal footing with the public sector

• The Jiangsu Provincial Government adopted a new policy to give private enterprises equal treatment as state-owned and collective enterprises… (People's Daily, April 9, 1999).

Qian, Yingyi. "The Process of China's Market Transition (1978-1998): The Evolutionary, Historical, and Comparative Perspectives." Journal of Institutional and Theoretical Economics, March 2000, 156(1), pp. 151-171.


ReformYingqiu Liu, a Senior Research Fellow and Professor of Economics at the

Chinese Academy of Social Sciences (CASS)

Since 1999:• Zeng Peiyan, minister at the State Development Planning Commission:

“[We will] eliminate all restrictive and discriminatory regulations that are not friendly towards private investment and private economic development in taxes, land use, business start-ups, and import and export. In the area of stock listings, private enterprise should enjoy equal opportunity which was enjoyed by the state-owned enterprises.”

• A large number of provincial governments that have issued documents that support and promote private enterprises in order to help them develop and grow rapidly.

Liu, Y. Development of Private Entrepreneurship in China: Process, Problems and Countermeasures. Working paper.

• One pair of founders had very high human capital with one being a lawyer and an MBA and the other having a Ph.D. Founded in 2003 and said that:

I then spent 20 years in the Bay area in life sciences companies. In the mid-1990’s I came back to China to survey biotech companies in China and found that the environment was not ready yet.

…spent an entire year just looking for the right office space…each product must be registered and approved by the government. It’s an expensive and time consuming procedure. She eventually found space for the company’s first store in a children’s museum which was perfect since they were selling toy bears aimed at children. This also allowed them to “hide” from government inspectors.


Entrepreneur’s Perspective

22

4. ContextAlumni survey Tsinghua University

• 30,000 mailed• 3,000 surveys• 10% r.r.• Growing tradition: Stanford GSB, Chicago, HBS• Disadvantages: Biased towards tech., other

response bias?• Advantages: Defined‘at risk’ set, first abroad,

detailed work history and founding data, less biased by govt. concerns

Hsu, D.H., Roberts, E.B., Eesley, Charles. 2007. Entrepreneurs from Technology-Based Universities: Evidence from MIT. Research Policy, 36: 768–788.

“Additionally, there is a strong need to develop data sets to study how economic and political factors affect entrepreneurship.” -(Klapper, Amit, Guillén, & Quesada, 2008)

Survey Questions

8. How many companies have you founded (not including State-Owned Enterprises (SOEs))? ____

9. How many companies have you privatized or bought? ____

Please answer the remaining following questions about your first company (Company A)

2. Will spending money on research and development be a major priority for this new business? Yes No

3. Were the products and services to be provided by your new business available on the market 3 years ago? No Yes, but only a few Yes, but not too many Yes, almost everywhere



1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

-2

-1.5

-1

-0.5

0

0.5

1

Coefficient on Year Fixed Effect

Dep. Var. = income from start-upDep. Var.= employees

Coeffi

cien

t


***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively.

Reform

Indep. Vars. log(revenue) log(employees)

POST-1999 1.262** (0.744) 0.510** (0.309)Master’s degree 0.370 (0.348) 0.083 (0.164)Doctorate degree -0.342 (0.631) 0.200 (0.305)Privatized 1.407** (0.690) 1.409*** (0.315)Bought -1.212 (0.751) 0.087 (0.349)Firm Age 0.460*** (0.085) 0.278*** (0.034)Communist Party 0.128 (0.346) 0.039 (0.163)Overseas 0.710 (0.449) 0.328* (0.198)Family Economic Status -0.078 (0.179) -0.018 (0.083)Constant 2.544** (1.208) 1.420** (0.518)Obs. 195 267


Distribution of Observable Human Capital

First (Highest) Quartile Second Quartile Third Quartile Fourth (Lowest) Quartile0

5

10

15

20

25

30

35

Histogram of GPA Rank

Bachelor's Master's Doctorate0

10

20

30

40

50

60

Highest Degree Earned

Perc

ent

28

5. MethodsDifferences-in-differences estimation

Cox Hazard Rate Model

(robust to logit)Prob (founding a firm = 1) = F(α + 1’θi + 2’ θi*POST + 3’POSTi + ’Xi

+ τt+ ηj + φa+ εi)

• Dependent variable: First start-up founded• Subjects start being “at risk” of founding a firm at the

time of their graduation • Θi = human capital measures• POST = 1 if individual was at risk for founding between

2000 and 2007.• Xi = Set of controls academic dept., region, education,

work history, job type, Communist party, overseas educ. or work, family economic status.

• Include (τ + η + φ) grad. year, region and Bachelor’s academic dept. fixed effects

• (Acemoglu & Finkelstein, JPE 2008)

Proportional Hazards Test


19961997

19981999

20002002

20032004

0.50.70.91.11.31.51.7

Hazard Rate Coefficients for Year-by-Year Interac-tions

Unobs. Talent (Income Residual)

Years Educa-tion

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

00.10.20.30.40.50.60.70.80.9

1

Coefficients on Year Fixed Effects (controls for region, education, department)

Year fixed effects

-0.45

-0.35

-0.25

-0.15

-0.05

0.05

0.15

0.25

**

**

**

0.8

0.10.2

0.2

0.1

Coefficients on Educ. Interaction with Year Fixed Effects

Master's Degree

Reduced Barriers to Growth: Observable Measures

Master’s degree

0.444***(0.121)

Master’s x POST

1.771*(0.581)

PhD degree 1.131(0.630)

PhD x POST 0.889(0.549)

Student leader

0.718**

(0.097)

Leader x POST 1.336*(0.209

)Log(yrs. work exp)

0.832***

(0.022)

Log(yr. work exp.) 0.999

x POST(0.027

)

Promoted0.216

***(0.11

2)Promoted x POST

3.361**

(1.953)

Years 2000-07 0.000***0.033*

**0.060*

**0.012

***

(POST) (0.000)(0.001

)(0.019

)(0.00

7)

N=1,821; 308 foundings; 44,248 total years at risk; ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively.

Dependent Variable = Start-up founded (subjects start being at risk upon graduation) Note: reported coefficients are hazard ratios

Highest salary

0.771**

(pre-founding)

(0.079)

Salary x POST 1.225*

(0.147)

Parents’ educ.

0.240***

(above median) (0.073)Parent edu. x POST

4.575***

(1.588)

High GPA0.350*

**(0.111

)GPA x POST 1.811*

(0.651)

Years 2000-07

0.007***

0.027***

0.020***

(POST)(0.004

)(0.007

)(0.006

)

Drawbacks to Observable Measures

• Could be correlated with family wealth

• Consistent with an opportunity costs story

• Subject to shifts in who gets a graduate degree, gets promoted, etc.

• Difficult to test changes in the shape of the distribution

Ideally want some continuous underlying measure of talent


Unobserved Talent: Income Regression

Specification for the ordered logit regression is as follows:

Yi = Φ(α + γXi + τt+ ηj + φa + εi)

Individual i ; 2:1 match on grad. yr. and salary year

Xit = controls including job type (academia, business,

government), tenure, Communist party, overseas, family wealth, and education

Dependent variable: salary (6 categories)τt+ ηj + φa = year + region + Bachelor’s department

Use εi in the entrepreneurship regressions


Andersson, F., M. Freedman, J.C. Haltiwanger, J. Lane, K.L. Shaw. 2006. Reaching for the stars: Who pays for talent in innovative industries? NBER Working Paper No. 12435

Gibbons, R. et al., 2005 JLE

33

5. Quantile Regression Model

• Many of the attractive properties of OLS or mean regression

• Advantage of allowing changes in the shape of the entire conditional distribution to be examined (Koenker & Basset, 1982; Koenker, Hallock 2001)

• Dependent variable: income regression residual

• Bootstrap method (with 100 repetitions) is used to generate standard errors (Horowitz 2001, Rogers 1992)

34

Quantile Regression

Controls for firm age, registered capital, privatized, bought, and salary.Dependent variable is the residual from the income regression in Table 6. Bootstrapped standard errors (100 repetitions); Standard errors are in parentheses. * p < 0.10, ** p < 0.05, *** p < 0.01

Panel A Dependent variable = income residuals (N=595)

Percentiles 10 25 50 75 80Founded in 1978-89 1.558*** 0.776*** 0.125** -0.494*** -0.680***

(0.506) (0.271) (0.065) (0.187) (0.246)Founded in 1990-99 0.769*** 0.496*** 0.009 -0.210 -0.233

(0.196) (0.161) (0.072) (0.259) (0.381)Founded in 2000-07 0.501*** 0.519*** 0.382*** 0.722*** 0.688***

(0.215) (0.174) (0.136) (0.194) (0.151)

Panel B Entrepreneurs only (N=132)

Percentiles 10 25 50 75 90

Ln(profit) 0.215 0.093 0.195 0.242** 0.246**

(0.295) (0.234) (0.155) (0.123) (0.126)

Observations 132 132 132 132 132

Pseudo R-squared 0.461 0.360 0.293 0.418 0.601


Indiv. F.E. 90th quartile 0.651

(0.203)Indiv. F.E. 90th * POST 1.293*

(0.206)Indiv. F.E. 10th quartile 0.976

(0.479)Indiv. F.E. 10th * POST 1.130

(0.810)Indiv. F.E. > median 0.624**

(0.144)Indiv. F.E. median*POST 7.329***

(2.165)

N=1800; Controls not shown; ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively.


Individual Fixed Effects

HypothesesH1: An institutional change reducing barriers to growth will increase entrepreneurship among individuals located relatively higher in the talent distribution.




37

Shift in Returns to Talent in Entrep.Independent Variables

Log(income from start-up)Log(profit margin)

(7-5) (7-6) (7-7) (7-8) (7-1) (7-4)Master’s degree -1.037 0.090 -0.449 -0.269

(0.905) (0.928) (0.276) (0.340)POST x Master’s degree 1.783* 0.163 0.236 -0.111

(1.039) (1.054) (0.345) (0.412)High GPA -1.191 -1.417 -0.523

(0.737) (0.903) (0.589)

POST x High GPA2.348**

* 2.376** 0.866(0.887) (1.042) (0.595)

Income residual

-3.239**

* -- --(0.901)

POST x income residual 2.624** -- --

(0.987)Promoted -1.123 -0.900

(1.182) (0.611)

POST x Promoted5.803**

* 1.751**(1.678) (0.768)

POST-1999 founding year -1.329 -0.139 1.027

-6.443**

* -0.301-

2.200**(0.977) (0.686) (0.819) (1.861) (0.455) (0.937)

Standard errors are robust. The results are Tobit models, but are robust to a Poisson specification (as well as not taking the natural log). ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively. All models include controls for year and region fixed effects as well as revenues, capital, employees, firm age, overseas returnee, and bach. graduation year.

Robustness checks1. Returns to talent in wage employment

2. MIT Data – 1998-2000 dotcom boom

3. Increased legitimacy

4. Alternative talent measures

5. Universities and research institutes began to invest and had better information on underlying ability levels

6. Demography of the funders/lead investor community itself might have changed and preferences towards higher ability individuals or technically skilled entrepreneurs (dotcom)

7. The nature of the economic opportunities or competition changed as China liberalized and the available opportunities required those with higher ability levels. Or payoff to skills, demand for innovation/R&D in the economy increased

8. Labor market, Increased college student enrollment, flood of graduate students with limited wage job opportunities

9. Specification

Dotcom Boom (MIT Alumni Data)

Note: Grad. years 1980 and after; 52 failures; 44,525 total years at risk; ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively. All models include controls for Bachelor’s graduation year (age), Bachelor’s Major (academic department).


Independent vars.

Dependent Variable = Start-up founded (subjects start being at

risk upon graduation)Note: reported coefficients are

hazard ratiosSoftware firms

only, only EE&CS grads

Software firms only, all grads

All Grads, All firms

Master’s degree 0.281*** 0.767 1.226(0.120) (0.204) (0.098)

Doctorate Degree 0.249** 0.733 1.261(0.153) (0.333) (0.127)

Master’s x Years 98-00 7.466** 1.936* 0.936

(5.844) (0.741) (0.152)Doctorate x Years 98-00 5.560* 1.446 0.905

(5.683) (0.849) (0.185)Non-U.S. citizen 1.866* 1.172 0.825

(0.634) (0.292) (0.078)Gender (male=1) 5.814* 3.305*** 1.495

(5.934) (1.302) (0.169)Years 1998-2000 0.001*** 0.002*** 0.011

(0.001) (0.001) (0.002)Obs. 3,266 18,896 19,188

40

6. Conclusion and ImplicationsHypothesis: Supported?

H1 Decr. barriers to growth = increases entrep. by those higher in talent distribution YES

H2 Talent = higher performance in entrepreneurship YES

•Institutional Level• Second margin at which the institutional environment affects

entrepreneurship• Different market failures affecting diff. individuals and start-

ups, at different stages in market development

•Individual Level• Suggestive of policies to encourage high ability

entrepreneurs (need more analysis here)• Human capital characteristics associated with

performance/innovation

•Strategy• Large firm new ventures• Understanding the drivers of high growth entrepreneurship• Institutional environment can shift competition and possibly

the types of start-ups being created

Social Welfare AnalysisΔSW = ΔCS + ΔPSΔ CS = Δp(Innovation) + CompetitionΔ PS = I(profith-profitl) – C(prev. organization)Induce greater investments in certain types of human capital

Cofounder / Investor Choice AnalysisIncreased probability of survival or growthExtra equity or salary to obtain cofounder (amt. required affected by inst. change)

Strategy / Economics of Innovation and Technological Change

Current Research

Charles Eesley

•Interact technical skills and science and engineering / industry context•Regional level policy experiments•Comparison with US data

Research Trajectory


• Eesley, Charles. 2009. Who has ‘The Right Stuff’? Human Capital, Entrepreneurship and Institutional Change in China. Winner SASE Student Paper Award, June 2009.

• Eesley, Charles; Roberts, E.B. 2009. Cutting Your Teeth: Learning from (One or More) Rare Experiences. Under review.

• Roberts, E.B.; Eesley, Charles. Entrepreneurial Impact: The Role of MIT. Kauffman Foundation, Feb. 2009.

• Hsu, D.H., Roberts, E.B., Eesley, Charles. 2009. Entrepreneurial Ventures from Technology-Based Universities: Evidence from MIT. Work in progress.

• Hsu, D.H., Roberts, E.B., Eesley, Charles. 2007. Entrepreneurs from Technology-Based Universities. Research Policy, 36: 768–788.

• Eesley, Charles, Lenox, Michael. 2009. Secondary Stakeholder Actions and the Selection of Firm Targets. Working paper.

• Rockart, Scott & Eesley, Charles. 2009. Prestige and Collaboration Patterns. Work in progress.

• Lenox, M. and Eesley, Charles. 2009. Private Environmental Activism and the Selection and Response of Firm Targets. Journal of Economics Management and Strategy, 18(1), Jan. issue.

• Eesley, Charles; Lenox, Michael. 2006. Firm Responses to Secondary Stakeholder Action. Strategic Management Journal, 27(8):765-781.

• Sloan, Frank A.; Eesley, Charles. 2007. Implementing a Public Subsidy for Vaccines. in Pharmaceutical Innovation: Incentives, Competition, and Cost-Benefit Analysis in International Perspective. edited by Frank A. Sloan and Chee-Ruey Hsieh. New York: Cambridge University Press.

Directions for Future WorkFocus on research designs disentangling causal mechanisms in commercializing innovation

Natural Experiments - Work well for institutional and policy questions

Instrumental Variables- Difficult to find instruments in strategic contexts

Thank you!

Chuck [email protected]

Performance


N=150; Standard errors are in parentheses. * p < 0.10, ** p < 0.05, *** p < 0.01

Independent Variables

Log(profit margin)

(5-1)

Log(employees)

(5-3)

Pr(IPR important)

(5-5)

Promoted -0.118(0.22

8) 0.380*(0.21

1) 0.042(0.96

6)

Log(work exp.) 0.986(1.23

0)

-2.447*

*(1.09

4) 0.668(4.72

3)Years of Education 0.004

(0.086) 0.152*

(0.082)

0.850**

(0.407)

Talent (income residual)

0.410**

(0.180) -0.12

(0.159) 1.535*

(0.837)

Prior salary

-0.442*

*(0.17

0) 0.173(0.15

0) -0.231(0.72

7)

Overseas0.731*

*(0.35

9) 0.128(0.30

5) 2.120(1.67

2)

High GPA 0.308(0.23

2) -0.321(0.21

3) -0.848(0.91

4)

Worked in R&D - - 1.639*(0.92

5)

Index of Backup Slides• Future Work• Identification• Theory contribution• Constructs/Measures• Response Bias• Boundary Conditions• Opportunity Costs• Descriptive statistics

– # jobs, industries• Implications• China policy reforms• Tsinghua history• Roy Model

• Quotes• Types of Talent• Web survey• Four Levels of Social A

nalysis• Robustness checks

– Returns in wage emp.

– U.S. data– Placebo regression– Legitimacy– Parallel Trends– Logit– Macroeconomic data– Proportional Hazard

s


Theoretical contribution• Economics of technological innovation and

entrepreneurship• Rational expectations, profit max. – not

necessarily• Institutional theory• Talent theory of bubbles or technology progress• Theory of the direction of innovation (non-SCOT)

– artifact – larger scale story

• Large vs. small firms tech waves of creative destruction

• Industry differences– more room for strategically influencing perceptions of

payoffs– VC herding behavior

• Strategic influencing directions (e.g. Eyeonics, YouTube, vaccines - less competition)


Reform Easing Constraints on Entry1988

Reform officially recognizing private businesses with more than 8 employees (Xu and Zhao, 2008)

Lower barriers to entry Lower quality entrepreneurs

• Nanda, 2008• Xu and Zhao, 2008

In the years following the 1988 reduction in barriers to entry individuals located relatively lower in the talent distribution were more likely to found firms.


MethodsCox Hazard Rate Model

(robust to logit)Prob (founding a firm = 1) = F(α + 1’θi + ’Xi + τt+ ηj + φa+ εi)

• Dependent variable: First start-up founded• Subjects start being “at risk” of founding a firm at the

time of their graduation • Θi = human capital measures• Xi = Set of controls academic dept., region, education,

work history, job type, comm. party, overseas educ. or work, family economic status.

• Include (τ + η + φ) grad. year, region and Bachelor’s academic dept. fixed effects


51


Dependent Variable = Start-up founded (subjects start being at risk upon graduation) Note: reported coefficients are hazard ratios; coefficients below 1.0 represent a decreased likelihood of entrepreneurship (N=1,540)

Master’s degree 0.675* 0.562**(0.158) (0.152)

Doctorate degree0.344*

* 0.641(0.166) (0.323)

Low work exper. (0-10 yrs.) 1.333 1.500

(0.351) (0.442)High work exper. (>30 yrs.)

0.060*** -

(0.039) -Promoted 0.689 0.686

(0.251) (0.270)High GPA (above median) 0.685* 1.288

(0.150) (0.325)Last Salary (Pre-founding) 0.667*** 0.694***

(0.065) (0.071) Communist Party member 0.760 0.744 0.807 0.77 0.774 0.776

(0.167) (0.166) (0.176) (0.169) (0.173) (0.185)

Note: 102 foundings; 30,716 total years at risk; ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively. All models include controls for Bachelor’s graduation year (age), Bachelor’s Major (academic department), and region fixed effects.

Reduced Barriers to Entry: 1988 - 1999

Controls for academic dept., region, education, work history, job type, comm. party, overseas educ. or work, family economic status, graduation year, major, and region fixed effects.

52

Income RegressionOrdered Logit

Independent VariablesDependent variable = salary category (1-6)

Master’s degree 0.466***(0.051)

Doctorate degree 0.905***(0.051)

Work exper. (0-10 yrs.) 0.938***(0.054)

Work exper. (10-30 yrs.) 1.065***(0.047)

Work exper. (>30 yrs.) 0.500***(0.055)

Gender (male=1) 0.498***(0.058)

GPA quartile (1st = top) -0.199***(0.046)

GPA quartile (3rd) 0.514***(0.046)

GPA quartile (4th = bottom) 0.047(0.045)

Overseas exper. 0.407***(0.043)

Academia -0.943***(0.051)

Business 0.335***(0.052)

Government -0.326***(0.044)

Pseudo R2 0.148Number of observations 561

Bach. Dept. and Year Effects included.Standard errors are in parentheses. * p < 0.10, ** p < 0.05, *** p < 0.01

1982

1984

1986

1987

1988

1989

1990

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

0

5

10

15

20

25

30

35

First Foundings by Year

First Founding Year

Nu

mb

er o

f S

tart

-up

s

INDUSTRYNUMBER OF

FIRMS %AEROSPACE 3 0.90ARCHITECTURE 13 3.88BIOTECH AND DRUGS 7 1.09CHEMICALS 8 2.39CONSUMER PRODUCTS 17 5.07ELECTRIC 12 3.58ELECTRONICS 69 20.60ENERGY 14 4.18FINANCE 10 2.99INTERNET 33 9.85LAW, ACCOUNTING 22 6.57MACHINERY 19 5.67MANAGEMENT 21 6.27MATERIALS 13 3.88MED DEVICES 4 1.19OTHER MFG 16 4.78PUBLISHING 11 3.28SOFTWARE 34 10.15TELECOM 9 2.69TOTAL 335 100

Industry Breakdown for First Firms


Income Regression Residuals

Founders only

Matched sample

Acknowledgements• Funding - Ewing Marion Kauffman Foundation

Dissertation Fellowship, Charles Zhang, founder and CEO of Sohu.com, and the MIT Entrepreneurship Center

• MIT – Professors Diane Burton, Fiona Murray, Edward B. Roberts– Yanbo Wang (PhD student)– Profs. Yasheng Huang and Elena Obukhova– Undergraduate Research Assistants

• Stephanie Yaung, Jennifer Sim, Celia Chen• Tsinghua University

– Prof. Delin Yang– Li Zhihua (Tsinghua Alumni Association)


1987198819901992199319941995199619971998199920002001200220032004200520062007

0

0.5

1

1.5

2

2.5

3

3.5

4

Percentage of "At-Risk" Individuals Becoming En-trepreneurs (By Education Level)

Graduate DegreeBachelor's Only

Per

cent

age

(%)


Ordered Logit Dep. Var = salary (6 bands)Master’s degree 0.693***

(0.137)Master’s x POST -0.228

(0.157)Doctorate degree 1.211***

(0.237)Ph.D. x POST -0.782***

(0.243)High GPA 0.384**

(0.170)High GPA x POST -0.292

(0.193) Ln(yrs. tenure) -0.028*** -0.033***

(0.005) (0.008)Business 0.179 0.24

(0.149) (0.238)Government -0.883*** -1.054***

(0.160) (0.244)Academia -0.646*** -0.659***

(0.165) (0.248)Years 2000-07 (POST) 1.760*** 1.348***

(0.157) (0.177)Charles Eesley - The Right Stuff

Returns to Talent in Wage Empl.Non-entrepreneurs only

Robust to high salary and last salary as well as negative binomial specifications.

Controls for bachelor’s graduation year, age, gender, overseas, comm. Party, and family economic status.

N=3,276 job spells; Robust standard errors (clustered at individual level) are in parentheses. * p < 0.10, ** p < 0.05, *** p < 0.01

Robustness checks1. Returns to Talent in wage employment

2. MIT Data – 1998-2000 dotcom boom

3. Increased legitimacy

4. Alternative talent measures

5. Universities and research institutes began to invest and had better information on underlying ability levels

6. Demography of the funders/lead investor community itself might have changed and preferences towards higher ability individuals or technically skilled entrepreneurs (dotcom)

7. The nature of the economic opportunities or competition changed as China liberalized and the available opportunities required those with higher ability levels. Or payoff to skills, demand for innovation/R&D in the economy increased

8. Labor market, Increased college student enrollment, flood of graduate students with limited wage job opportunities

9. Specification

Robustness – legitimacyhigh status = Ever job gov., Comm. Party, Ph.D.


Dependent Variable = Year start-up founded (subjects start being at risk

upon graduation) Note: reported coefficients are hazard ratios (N =

1,910)(1) (2)

Years Education - 0.537*** (0.072)Education x Post-1999 - 1.737*** (0.259)High Status (Gov., Ph.D. and Comm. Party) 1.442 (0.401) 1.269 (0.341)Status x Post-1999 0.590* (0.187) 0.725 (0.224)

Post-1999 dummy 0.055*** (0.014) 0.000***(0.000)

1991-1999 dummy 0.071*** (0.011) 0.064*** (0.010)

N= 626; Standard errors are in parentheses. * p < 0.10, ** p < 0.05, *** p < 0.01

pair-wise correlations High StatusYears of Education 0.101High GPA 0.080Promoted 0.040Ln(work exp.) 0.066Parents’ Education 0.046

Back


Placebo Regression

Years of Education

0.630***

(0.086)

Educ. x POST 1.082(0.20

5)

GPA High 0.726(0.23

1)

GPA x POST 0.716(0.31

7)

Promoted 0.218**(0.13

1)Promoted x POST 7.714**

(7.210)

Ln(workexp)0.454*

**(0.11

6)Work exp. x POST

0.367***

(0.119)

Income residual 0.717

(0.179)

Residual x POST 1.333

(0.450)

Years 1997-1999(POST) 0.160

(0.521) 0.690

(0.256)

0.083***

(0.079) 4.362*

(3.346) 0.887

(0.349)

Years 1990-1996

0.206***

(0.065)

0.307***

(0.098)

0.216***

(0.070)

0.258***

(0.097)

0.479*

(0.182)

Controls for academic dept., region, education, work history, job type, comm. party, overseas educ. or work, family economic status, graduation year, major, and region fixed effects.

N=1,821; Note: 119 failures; 20,541 total years at risk; ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively.


Identification

Concerns:• 1999 reform lowered entry

barriers• Reform had broader effects• More continuous changes• Reform did not increase

returns to entrepreneurship• Simultaneous changes with

similar results on growth barriers

Tests:• MIT Data• Placebo regression• Direct test of returns to

talent in entrep.• Incr. in foundings controlling

for macro-economic vars.• Returns in wage empl.• Qualitative data


Boundary Conditions1) Types of skills and talent necessary to overcome entry barriers

are not largely orthogonal to those useful for firm growth

2) Feedback loops are weak: that is, increases in the number of entrepreneurial firms do not strongly increase competition or create significantly better wage employment opportunities

3) Initial relationship between talent and returns to entrepreneurship is not one where primarily those at the top of the talent distribution become entrepreneurs

4) Sufficient variation in the distribution of talent (by measures relevant for wage and entrepreneurial payoffs) in the sample at risk for entrepreneurship


Back

Higher Non-Market Ability (Gov. connections, bureaucracy nav., wealth)

Mar

ket A

bilit

y /

Hum

an C

apita

l



The increase is among those of relatively higher ability.

Back

Kaplan-Meier curves


Kaplan-Meier curves


Kaplan-Meier curves

Charles Eesley - The Right Stuff Back Backup

Tests and graphs based on the Schoenfeld residuals


0 1 2 3 4 50

10

20

30

40

50

60

Histogram of Number of Jobs

Number of Jobs Reported

Perc

ent


Proportion of Foundings with a Direct Match to the Industry and Founder's Bachelor's Degree

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

Before 1999 After 1999

Prop

ortio

n

* Differences are statistically significant (p<0.05; t = 1.759)


Human Capital and Opp. Costs

wages = a+ *schoolinghigh-HC people can either have high a

or high

wages=a+1*HC+(2-c)*entrep+3*HC*entrep


Human Capital and Opportunity Costs



Entrepreneurial Returns

Wage empl. income

$

(incr. opportunity costs, decreased entrepreneurship)

Human Capital and Opportunity Costs




Wage empl. income

$

(Incr. opportunity costs, decreased entrepreneurship)

Human Capital and Opportunity Costs(incr. entrep. returns to talent)




Wage empl. income

$

Robustness: Logit

Indep. variablesDep. Var = 1 if founded a firm

Graduation Years < 2002Talent (income resid.) -0.400 -0.643

(0.460) (0.521)

Talent x POST 1.099**1.454**

*

(0.475) (0.544)Years of Education -0.402 -0.974

(0.311) (0.723)Educ. x POST 0.320 0.424

(0.319) (0.739)Promoted -0.398 0.678

(0.900) (1.718)Promoted x POST 0.294 -0.873

(0.957) (1.796)

POST

-4.722**

* -9.964*

-4.742**

* -11.388

(0.943) (5.544) (0.915) (13.029)

Observations 451 1277 1277 451

Controls for academic dept., region, education, work history, job type, comm. party, overseas educ. or work, family economic status, graduation year, major, and region fixed effects.***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively.

• Human Capital/Talent – education, parents’ education, GPA, work experience, income, income residuals, student leader

• Variance of returns – categories of wage income, income from start-up, profit

• Barriers to Growth

• Status/Legitimacy – PhD, government


Constructs and Measures

Variable Responded before Aug. 2007(N=2,667)

Responded during/after Aug. 2007 (N=299)

t-stat for equal means

Age 49.3 54.1 -4.216**Age (founders only) 38.4 37.4 0.602Bachelor’s Graduation Yr 1980.9 1977.4 3.777**Bach. Grad yr (founders only) 1991.6 1993.2 0.941Years of Education 17.2 17.0 2.381**Entrepreneur parents 0.09 0.12 -0.713EntrepreneurPrivatizedFirst start-up founded

0.290.10

2000.3

0.400.05

2001.1

-2.168**1.392-0.661

Tech only 0.28 0.29 0.757Business only 0.10 0.09 0.235Gender 0.88 0.90 0.901Family economic status 3.75 3.85 -1.871*High Salary 3.21 2.93 3.351**Avg. Tenure 6.94 8.01 -2.045*Overseas work exp. 0.26 0.26 -0.126Number of positions 2.39 2.26 -2.012*High governmentLow government

0.030.18

0.030.17

-0.2390.617

Last job academia 0.19 0.19 -0.051Ever job academia 0.32 0.27 2.323**Last job business 0.62 0.61 0.348Student Leader 0.61 0.57 0.874GPA Rank 2.28 2.58 -2.661**Bach. Grad Yr. 10th percentile 1954 1953 --Bach. Grad Yr. 25th percentile 1965 1961 --Bach. Grad Yr. 50th percentile 1986 1979 --Bach. Grad Yr. 75th percentile 1996 1993 --Bach. Grad Yr. 90th percentile 2001 2001 --

77

Appendix AComparison of Key Demographic Characteristics by Survey Wave

***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively.Back


Categories Tsinghua CHNS NBS HH survey

NBS HH survey

Sample Urban Rural and Urban

Urban – self-employed

Urban – non. Entrep.

Male 0.89 0.53 0.56 0.50

Age 50.13 41.45 36.2 37.2

Married 0.88 0.98 83.4 84.2

Years of Education

17.1 9.1 9.2 9.4

Household Size

3.40 3.9 -- --

Self-employed 0.26(0.8% in

1999)

0.14 (4% in 1999) --

Experienced a layoff

0.13 -- 0.26 0.19

Father’s Educ. 4.11 -- 5.4 5.2

Mother’s Educ.

4.89 -- 6.0 5.9

Parent Self-Empl.

0.08 -- 0.06 0.05

Comm. Party 0.62 -- 0.05 0.18

Benchmarking Tsinghua Data

Work History AttributesNon-

Founders (n=2152)

Founders (n=670)

t-stat

Variable MeanAge 52.8 42.9 12.388***Entrepreneur Parents 0.036 0.031 0.614Privatized 0 0.260 -Male 0.879 0.933 -3.931***Family Econ. Status

(category) 3.834 3.639 4.348***Recent Salary

(category) 2.317 2.045 3.686***Avg. Tenure 8.074 4.813 7.554***Overseas 0.126 0.212 -5.525***Number of Positions 2.115 3.109 -18.605Ever Job High

Government 0.036 0.042 -0.713Ever Job Low

Government 0.242 0.176 3.574***Last Job Academia 0.166 0.081 5.475***Ever Job Academia 0.233 0.207 1.402*

Last Job Business 0.399 0.687 -

13.451***Master’s 0.402 0.550 7.187***Ph.D. 0.101 0.115 1.066Student leader 0.674 0.903 -5.314***

1982

1984

1986

1987

1988

1989

1990

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

0

0.2

0.4

0.6

0.8

1

1.2

1.4

Percentage of "At-Risk" Alumni Body Becoming First-Time Entrepreneurs By Year

Per

cent

age

(%)


Four Levels of Social AnalysisWilliamson (2000)• Social Embeddedness

– Culture, Religion (Economic Historians, Anthropologists, Sociologists)

• Institutional Environment– “Rules of the Game”, Constitutions, judiciary, politics (Poli. Sci.,

Inst. Economics)

• Governance– Contracts, Theory of the firm (Contract/Info Econ, TCE)

• Resource Allocation– Incentive aligment, Quantities and Prices (Neoclassical

Economists)

83

Independent Variables Dependent variable = number of firm foundings (1959-2007)

R&D to GDP ratio (t-1) -0.143

GDP per capita (in RMB, t-1) (0.723)Shanghai Stock Exchange Market Cap (t-1)

3.43E-04***

3.30E-04***

Domestic Patents Issued (t-1)(8.59E-

05) (8.63E-05)

Post-1999 dummy -2.52E-

05**-2.36e-

05**

Post-1988 dummy(1.16E-

05) (1.20E-05)

Constant5.01E-

06-1.01e-

05** -9.48E-06*

Log likelihood(3.60E-

06)(4.82E-

06) (5.04E-06)

Num. obs.4.326**

* 3.856*** 2.901***Pseudo R2 (0.652) (0.530) (0.590)

Negative Binomial Regressions on Macroeconomic Data

Data merged from the State Statistics Bureau, Chinese Statistical Yearbooks***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively.

Lower Barriers to Entry

Lower Barriers to Growth

Market/Commercializatio

n Talent is Orthogonal to ‘Bureaucratic’

Talent

Increase in high and low ability entrepreneurs

No prediction on the type of firms founded

Possibly greatest increase in marginal firms which were not profitable with the previously high cost of entry

Increased competition (possibly lower profit margins?)

Unclear predictions on whether increase is among high or low ability

Easier to overcome the opportunity costs for entrepreneurship

Increase in high growth firms

More high-growth entrepreneurial opportunities supports venture capital

Market/Commercialization Talent is NOT Orthogonal to ‘Bureaucratic’

Talent(a large component is common between

the two)

Relative increase in low ability entrepreneurs

Those who cannot maintain wage employment can overcome the barriers to entry

Relative increase in low growth firms

Increased competition, smaller, less profitable firms

Increase in high ability entrepreneurs

Easier to overcome the opportunity costs for entrepreneurship

Easier to recruit talented co-founders

More high-growth entrepreneurial opportunities supports venture capital

Increase in high-growth firms

Increase in innovative firms (higher returns can support higher risk of an innovation strategy)

Strategic and Competitive Implications of Different Institutional Shifts


China’s S&T Policy Reform1978-85 1986-97 1998-2006

• Deng Xiao Ping reform report (1975)

• Open door policy (1978)

• National Key Tech. R&D Program (1984)

• Univ. reform (1985)

• SE Zones created (1980)

• 1992 Tour• Bankruptcy law

for SOEs (1986)

• National Natural Science Fndtn. (1986)

• >7 emp. permittedTorch Program (1988)

• Stock Exchange (1990)

• 1997

• Private ownership (1999)

• Adoption of medium and Long Term S&T Strategic Plan (2006)

• Promotion of VC/PE (1998)

• CAS Knowledge Innov. Prog. (1998)

• Innov. Fund for Tech. SMEs (1999)

• Join WTO (2001)


Tsinghua Univ.

86

• Established in Beijing in 1911

• 1952 reorganized Soviet style

• 1966-1976 Battlefield during Cultural Revolution

• 1978 restored departments in sciences, economics and management, and humanities

• 1984 – First Graduate school in China created at Tsinghua

• 1998 – Tsinghua Science Park established

ModelRoy Model as extended by Borjas (1987)Labor markets 0 and 1, respectively. Log earnings in the wage sector:

w0 = μ0 + ε0

ε0 ~ N (0, σ02)

The wage sector earnings would be the following if everyone from the wage sector were to migrate to entrepreneurship (ignoring general equilibrium effects):

w1 = μ1 + ε1

ε1 ~ N (0, σ12) .

Assume that the cost of becoming an entrepreneur is C π = C/w0

π is constant, meaning that C is directly proportional to w0

The correlation between entrepreneur and wage worker earnings is σ01 = cov(σ0, σ1). A worker will choose entrepreneurship if


• One pair of founders had very high human capital with one being a lawyer and an MBA and the other having a Ph.D. Founded in 2003 and said that:

I then spent 20 years in the Bay area in life sciences companies. In the mid-1990’s I came back to China to survey biotech companies in China and found that the environment was not ready yet.



Entrepreneur’s Perspective

• Environment was costly, not prohibitive for entrepreneurship. One entrepreneur reported that she:


• Individuals managed to find ways to raise funding

At that time there was not much VC investment in China (1997) so they raised the first round of money among their family and divided up the shares according to how much each of them could raise from family. – JX


Entrepreneur’s Perspective on Pre-period (4/5)

90

In the 1990s … government was giving them less support. Many universities created “university –run” enterprises and were basically selling off the periphery of campus. Even high schools had school—run enterprises that were considered acceptable. There is a Tsinghua name to many of them. - GC – 2nd generation investor


Interview Quotes


Strategy-making is different in China than in the US. In China you pick up a person who has resources, then you choose the appropriate field to enter then you come up with the strategic goals, then the vision. In the US he feels that it is the opposite. - WW – Sea Corp.

Interview Quotes, cont.

Percentage of Businessmen/Businesswomen Becoming Entrepreneurs

0.00%

0.50%

1.00%

1.50%

2.00%

2.50%

3.00%

Founding Year

Perc

en

tag

e

Last Job was inBusiness

Ever had a job inBusiness


Percentage of Government Officials Becoming Entrepreneurs

0.00%

2.00%

4.00%

6.00%

8.00%

10.00%

12.00%

Founding Year

Per

cen

tag

e

High levelgovernmentofficial positionLow levelgovernmentofficial


Why China?• Tech-based entrepreneurship in developing countries

rarely appears in academic literature (Lu 1997, 2000; Puga and Trefler, 2005)

• Vernon’s (1966) product-cycle model

• 19892004 China 29% vs. US 1% (State Statistics Bureau)19782004 # employed in private business up 300X

• Policies and institutions changing rapidly (Cull & Xu, 2006; Nee, 1998; 1992; 1996; Peng & Heath, 1996; Steinfeld, 2007)


Context


Entrepreneurship and Institutions

(Gollin, JPE 2002)

Research Trajectory


Future Work


• Eesley, Charles, Lenox, Michael. 2008. Secondary Stakeholder Actions and the Selection of Firm Targets. Working paper.

• Lenox, M. and Eesley, Charles. 2009. Private Environmental Activism and the Selection and Response of Firm Targets. Journal of Economics Management and Strategy, 18(1), Jan. issue.

• Eesley, Charles; Lenox, Michael. 2006. Firm Responses to Secondary Stakeholder Action. Strategic Management Journal, 27(8):765-781.

• Eesley, Charles. 2008. Who has „The Right Stuff‟? Human Capital, Entrepreneurship and Institutional Change in China. Winner SASE Student Paper Award, June 2008.

• Eesley, Charles; Roberts, E.B. 2008. Cutting Your Teeth: Learning from (One or More) Rare Experiences. Under review.

• Hsu, D.H., Roberts, E.B., Eesley, Charles. 2007. Entrepreneurs from Technology-Based Universities. Research Policy, 36: 768–788.

HypothesesH1: An institutional change reducing barriers to growth will increase entrepreneurship among individuals located relatively higher in the talent distribution.


