Is pension insurance a barrier to entrepreneurship? New evidence from China

Figures & data

Table 1. Description of variables.

Variable	Definition
entrep	If you have set up a business (1 = yes, 0 = no)
basic_insurance	If you are participating in basic pension insurance (1 = Insured, 0 = uninsured)
busi_retire	If you are participating in commercial pension insurance (1 = Insured, 0 = uninsured)
age	Age in years
age^2	Age squared
edu	Educational level (1–6, 1: illiterate; 2: primary school and below; 3: junior secondary school; 4: senior secondary school and its equivalent; 5: college diploma; 6: bachelor’s degree and above)
logincome	Logarithmic total annual income
social	How often you are social (1 ‘rarely’ 2 ‘sometimes’ 3 ‘often’)
sp_work	If your spouse is employed (1= employed, 0 = unemployed)
marriage	Marital Status (1 = married , 0= unmarried)
urban	1 = urban, 0= rural
work_year	Number of working years
party	If you are a CPC member (1 = yes, 0 = no)
gender	1 = male, 0= female

Data source: Chinese General Social Survey (2013) calculated by stata software.

Table 2. Descriptive statistics.

Variable	N	Mean	p50	SD	min	max
entrepr	4,597	0.2173	0.0000	0.4125	0.0000	1.0000
basic insurance	11,218	0.6807	1.0000	0.4662	0.0000	1.0000
busi_retire	11,389	0.0605	0.0000	0.2384	0.0000	1.0000
Age	11,437	48.5974	48.0000	16.3882	17.0000	97.0000
age2	11,437	2,630.2520	2,304.0000	1,655.0740	289.0000	9,409.0000
Edu	11,432	2.0226	2.0000	1.2687	0.0000	5.0000
logincome	9,072	9.6156	9.9035	1.1884	4.3820	13.8155
social	11,434	1.8874	2.0000	0.8059	1.0000	3.0000
urban	11,424	0.4456	0.0000	0.4971	0.0000	1.0000
sp_work	9,017	0.6299	1.0000	0.4829	0.0000	1.0000
marri	11,415	0.8901	1.0000	0.3127	0.0000	1.0000
wor_year	4,527	15.1401	13.0000	10.8256	1.0000	69.0000
party	11,372	0.1021	0.0000	0.3028	0.0000	1.0000
gender	11,438	0.5032	1.0000	0.5000	0.0000	1.0000
sma_busi	11,426	0.0712	0.0000	0.2571	0.0000	1.0000
Inno_driven	11,438	0.9289	1.0000	0.2570	0.0000	1.0000

Note: Chinese General Social Survey (2013), all variables used in this article are displayed.

Source: calculated by stata software.

Table 3. Probit regression of two types of pension insurance on entrepreneurship, Marginal Effects.

	(1)	(2)	(3)	(4)
	Entrepr	entrepr	entrepr	entrepr
basic_insur	−0.3657***	−0.2492***
	(−5.1847)	(−3.5215)
busi_retire			−0.2456*	−0.2699*
			(−1.7257)	(−1.8224)
gender	0.0258	−0.0702	−0.0681	−0.0431
	(0.5081)	(−1.6045)	(−1.6220)	(−1.1057)
age	0.0465***	−0.0087	−0.0115	−0.0144
	(2.8966)	(−0.3893)	(−0.5355)	(−0.6272)
age2	−0.0005**	0.0002	0.0001	0.0002
	(−2.5041)	(0.5914)	(0.5458)	(0.6935)
edu1	−0.1805***	−0.1723***	−0.2867***	−0.2489***
	(−5.2104)	(−3.9327)	(−7.2919)	(−6.1027)
logincome		0.2468***	0.2395***	0.2507***
		(4.6251)	(4.4560)	(4.5537)
social		0.0581*		0.0612*
		(1.7879)		(1.8513)
sp_work		0.1117	0.1079	0.1129
		(1.0417)	(1.0455)	(1.0572)
marri		0.7137***		0.6933***
		(3.1255)		(2.9719)
wor_year		0.0001	−0.0042	−0.0031
		(0.0110)	(−1.1066)	(−0.7745)
party		−0.4245***		−0.4444***
		(−3.3646)		(−3.7241)
urban		−0.2688***
		(−4.0401)
_cons	−1.1783***	−3.2542***	−2.1965***	−3.1549***
	(−2.8890)	(−4.8994)	(−3.3504)	(−4.6260)
N	4521	3345	3410	3386
pseudo R^2	0.0453	0.0620	0.0430	0.0531

Note: Standard errors are clustered at province level and reported in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. Controls include age, age², educational level, logarithmic total annual income, social frequency, number of working years and along with dummies for marital status, sex, party, the employment status of the spouse, and household registration. All data are from the Chinese General Social Survey (2013). Coefficients reported for probit regressions are the average marginal effects. Regression models are sample weighted using population.

Source: calculated by stata software.

Table 4. Two stage regression results for instrumental variable approach.

	OLS (1)	OLS (2)	First stage (3)	Second stage (4)	First stage (5)	Second stage (6)
	residual term	entrepr	basic_insur	entrepr	busi_retire	entrepr
$\mathit{other}\_\mathit{basi}{c}_i$	0.3043 (1.02)	0.3043 (1.02)	0.82e -06** (2.49)		5.71e-02* (1.92)
basic_insur		−60.169* (−1.99)		−61.102* (−1.97)		−100.2383* (−1.69)
Control variables	Yes	Yes	Yes	Yes	Yes	Yes
Observations	1912	1912	1912	1912	1933	1933
F-test			19.51		12.61

Note: The robust t statistics are in parentheses. ***, **, * representing significance at 1%, 5% and 10%, respectively. The dependent variable in column (1) is the residual term. The dependent variable in Column (2) is entrepr. Column (3) and Column (5) gives the first stage regression results, the regression results of $\mathit{other}\_\mathit{basi}{c}_i$ on individual basic pension insurance and business pension insurance respectively. Other variables are all controlled in the regression.

Source: calculated by stata software.

Table 5. Estimated results by region, marginal effects.

	East	Northwest	Southwest	Central	Northeast
	(1)	(2)	(3)	(4)	(5)
	entrepr	entrepr	Entrepr	entrepr	entrepr
basic_insur	−0.2060**	−0.3080	−0.3828***	−0.1879	−0.5029**
	(−2.1879)	(−1.0857)	(−4.7392)	(−1.1051)	(−2.4332)
Controls	Yes	Yes	Yes	Yes	Yes
_cons	−4.4636***	−0.4615	−0.3534	−7.3233***	0.1238
	(−3.4374)	(−0.4579)	(−0.2263)	(−4.0399)	(0.0773)
N	1568	245	497	687	517
pseudo R^2	0.0623	0.0570	0.0974	0.1109	0.0854

Note: Coefficients reported are the marginal effects of probit regressions. Controls include age, age², educational level, logarithmic total annual income, social frequency, number of working years and along with dummies for marital status, sex, party, the employment status of the spouse and household registration. Standard errors are clustered at province level and reported in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. Regression models are sample weighted using population. All data are from the Chinese General Social Survey (2013).

Source: calculated by stata software.

Table 6. Effect of two types of pension insurance on entrepreneurship by education, Marginal Effects.

	College graduate		High school graduate		No degree
	(1)	(2)	(3)	(4)	(5)	(6)
	entrepr	entrepr	entrepr	entrepr	entrepr	entrepr
basic_insur	−0.4738*		−0.4851***		0.2220
	(−1.6841)		(−3.6800)		(0.6980)
busi_retire		−0.3410		−0.4607***		1.4182
		(−1.2529)		(−2.8682)		(1.0969)
Control	Yes	Yes	Yes	Yes	Yes	Yes
_cons	−6.4856***	−6.9005***	−2.9915***	−3.1022***	−3.8523*	−3.8067*
	(−6.6767)	(−6.6734)	(−3.3987)	(−3.5043)	(−1.8126)	(−1.7193)
N	1152	1178	902	906	84	84
pseudo R^2	0.0647	0.0620	0.0839	0.0753	0.0956	0.0980

Note: Standard errors are clustered at province level and reported in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. Coefficients reported are the marginal effects of probit regressions. The group by ‘College graduate,’ ‘High school graduate’ and ‘No degree” are based on educational level of the individual. Regression models are sample weighted using population. All data are from Chinese General Social Survey (2013).

Source: calculated by stata software.

Table 7. Effect of pension insurance on men and women, marginal effects.

	Men	Women	Men	Women
	(1)	(2)	(3)	(4)
	entrepr	entrepr	entrepr	entrepr
basic_insur	−0.2212***	−0.2584***
	(−3.0845)	(−2.8393)
busi_medi			−0.2872**	−0.2310**
			(−2.5037)	(−2.0007)
controls	Yes	Yes	Yes	Yes
_cons	−4.3129***	−2.4549***	−3.7955***	−0.7617***
	(−5.7818)	(−2.6312)	(−5.1053)	(−22.0217)
N	2055	1290	2090	1814
pseudo R^2	0.0654	0.0894	0.0370	0.0022

Note: Standard errors are clustered at province level and reported in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. Coefficients reported are the marginal effects of probit regressions. Controls include age, age2, educational level, logarithmic total annual income, social frequency, number of working years, along with dummies for marital status, sex, party, the employment status of the spouse, and household registration. All data are from the Chinese General Social Survey (2013).

Source: calculated by stata software.

Table 8. Effect of pension insurance on two social network groups, marginal effects.

	Low social	High social	Low social	High social
	(1)	(2)	(3)	(4)
	entrepr	entrepr	entrepr	entrepr
basic_insur	−0.2044***	−0.2965**
	(−2.9583)	(−2.3675)
busi_retire			−0.3293*	−0.3118
			(−1.8282)	(−1.5277)
Controls	Yes	Yes	Yes	Yes
_cons	−3.9532***	−2.0349**	−3.6135***	−0.7838***
	(−4.5486)	(−2.0222)	(−4.0964)	(−11.8799)
N	2055	1290	2090	1814
pseudo R^2	0.0644	0.0788	0.0378	0.0039

Note: Standard errors are clustered at province level and reported in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. Coefficients reported are the marginal effects of probit regressions. Group is based on the mean of the variable social. If the value of the variable ‘social’ is below the median (1.8874), we classify it as ‘low social network group,’ and when the value of the variable ‘social’ is above the median (1.8874), we classify it as ‘high social network group.’ Controls include age, age², educational level, logarithmic total annual income, social frequency, the number of working years, and dummies for marital status, sex, party, employment status of the spouse, and household registration. Regression models are sample weighted using population. All data are from the Chinese General Social Survey (2013).

Source: calculated by stata software.

Table 9. Effect of pension insurance on two income groups, marginal effects.

	Low income	High income	Low income	High income
	(1)	(2)	(3)	(4)
	entrepr	entrepr	Entrepr	entrepr
basic_insur	−0.1903*	−0.2320***
	(−1.7898)	(−3.6039)
busi_retire			0.0183	−0.2220**
			(0.0712)	(−2.2812)
Controls	Yes	Yes	Yes	Yes
_cons	−0.2041	−6.0992***	0.9737	−6.2982***
	(−0.2153)	(−8.6539)	(0.7354)	(−8.8914)
N	759	2754	592	2791
pseudo R^2	0.0287	0.0941	0.0300	0.0933

Note: Robust standard errors are reported in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. Coefficients reported are the marginal effects of probit regressions. Group is based on the mean of the variable logincome. If the value of the variable logincome is below the median (9.6155), we classify it as ‘low income group,’ and when the value of the variable logincome is above the median (9.6155), we classify it as ‘high income group.’ Controls include age, age2, educational level, logarithmic total annual income, social frequency, number of working years, along with dummies for marital status, sex, party, employment status of the spouse, and household registration. All data are from the Chinese General Social Survey (2013).

Source: calculated by stata software.

Table 10. Effect of two types of pension insurance on entrepreneurship by marital status.

	Married	Not married	Married	Not married
	(1)	(2)	(3)	(4)
	entrepr	entrepr	entrepr	entrepr
basic_insur	−0.2345***	−0.7816***
	(−3.9591)	(−3.7403)
busi_retire			−0.1476**	−0.4684*
			(−1.9697)	(−1.6895)
Controls	Yes	Yes	Yes	Yes
_cons	−2.8637***	−4.4559***	−0.7011***	−1.6268***
	(−4.4444)	(−3.8074)	(−16.9165)	(−3.4349)
N	3308	549	3918	641
pseudo R^2	0.0633	0.1600	0.0009	0.0444

Note: Standard errors are clustered at province level and reported in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. Coefficients reported are the marginal effects of probit regressions. The group by ‘married’ and ‘not married’ are based on the marital status of the individual. Married = 1, unmarried = 0. Data are from Chinese General Social Survey (2013).

Source: calculated by stata software.

Table 11. Effect of pension insurance on the two types of entrepreneurs, marginal effects.

	(1)	(2)	(3)	(4)
	sma_busi	inno_driven	sma_busi	inno_driven
basic_insur	−0.2417***	0.2383***
	(−4.2363)	(4.1833)
busi_retire			−0.2490***	0.2494***
			(−2.6003)	(2.6047)
Controls	Yes	Yes	Yes	Yes
_cons	−2.5284***	2.5001***	−2.5502***	2.5276***
	(−4.1706)	(4.1262)	(−4.1703)	(4.1351)
N	3342	3352	3382	3392
pseudo R^2	0.0616	0.0614	0.0595	0.0595

Note: Robust standard errors are reported in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. Coefficients reported are the marginal effects of probit regressions. If the individual hired fewer than 7 employees, the entrepreneur is classified as a small business. However, if the individual hired more than 7 employees, they are classified as an, innovation-driven entrepreneur. Controls include age, age², educational level, logarithmic total annual income, social frequency, number of working years, along with dummies for marital status, sex, party, the employment status of the spouse, and household registration. All data are from the Chinese General Social Survey (2013).

Source: calculated by stata software.

Table 12. The transmission channels behind the negative effect of the two types of pension insurance on entrepreneurship.

	(1)	(2)	(3)	(4)	(5)	(6)
	l_safe	l_safe	total_income	total_income	Risk_share	Risk_share
basic_insur	−0.0773**		−0.0450*		0.0907**
	(−2.5256)		(−1.7696)		(2.0120)
busi_retire		0.0290		0.0563		−0.0045
		(0.4692)		(0.6077)		(−0.0700)
Control	Yes	Yes	Yes	Yes	Yes	Yes
_cons			0.1298***	−14.1968***	−1.2568***	−1.3072***
			(6.1803)	(−12.1347)	(−5.1231)	(−5.3805)
N	5507	5586	11218	3392	4409	4472
pseudo R^2	0.0006	0.0002	0.0002	0.3932	0.0225	0.0214

Note: Robust standard errors are reported in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. Family total income has a value of 1 when the economic condition of the family is above the average and a value of 0 otherwise. The value of lack of security ranges from 1 to 5, and a larger value means a greater sense of insecurity. Risk sharing is valued 1 when ‘Government, children and elderly equally sharing responsibility’ was selected and 0 otherwise.

Source: calculated by stata software.