Political Background and Household Financial Asset Allocation in China

ABSTRACT

Political background is an important factor in determining the household economic behavior. Using 2014–2018 households panel data from the China Family Panel Studies (CFPS), we investigate the effects of political background on China’s household asset allocation behavior. We find that political background has a significant positive impact on the financial market participation. Mediation analysis indicates that political background leads to higher household wealth, better social capital, and fewer credit constraints, thus promotes households investments. Further analysis shows that the marginal impact of political background on household investment behavior is more significant in Eastern and urban areas. Our results contribute to the existing literature on the relationship between the political background and the household investment behavior, also enhancing the understanding of the household portfolio heterogeneity.

KEYWORDS:

• Political background
• financial market participation
• household financial
• China

JEL CLASSIFICATION:

• D14
• D91
• G11

1. Introduction

Household finance is the study of ways which households participate in the financial markets and employ various types of financial instruments to achieve their desired goals. This field of study has drawn increasing attentions from academics and practitioners in recent years, despite the fact that it is difficult to measure household investment behavior. Campbell (2006) documents some evidences on participation, diversification, and mortgage refinancing, suggesting that many households appear to invest effectively in the U.S. However, Cooper and Zhu (2018) reveal that the financial choices of households in China are driven by institutional factors such as labor market risks and cost of asset market participation, as well as by differences in preferences. This leads to a relatively low stock market participation rate and low share of stocks in wealth conditional on participation in China. Traditional corporate finance deals with owners’ wealth maximization while having to resolve agency problems, household finance has many unique features that offer its special characteristic. Campbell (2006) identifies households’ investment plan to be long but with finite horizon while holding certain important nontraded assets such as human capital and illiquid asset such as housing. In addition, households usually experience significant constraints on their ability to borrow. Many studies have reported that age, education, income, and financial literacy may affect household financial market participation (Campbell 2006; Guiso, Sapienza, and Zingales 2008; Jia et al. 2019; Zhou and Xiao 2018). However, research on intangible assets such as political background is insufficient, as intangible assets may also impact investment behavior. Cao and Qian (2020) investigate the important role of political background on household income in China. Political background is a metaphor used to conceptualize the accumulation of resources and power build through relationships, trust, goodwill, and influence between stakeholders. Therefore, we investigate the effect of political background on household financial asset allocation, an area has been drawing increasing attention in China.

Political background refers to the power and influence provided by political status. It reflects people’s social status, economic resources, and future development. The important role of political background on economic behavior has been widely recognized. Most studies measure the political background in terms of cadres or Party identity and find that political background positively affects household income, bank loans, and employment opportunities (Cao and Qian 2020; Li, Hermes, and Lensink 2020; Wei and Zhong 2017). However, there are two main limitations from previous studies. Firstly, many focus on the household income or bank loans and ignore the investment behavior. Secondly, there is a lack of acknowledgment of an individual’s ability due to sectional data.

Therefore, our article fills the gap and contributes to the existing literature on three important aspects: First, we provide new insights into the relationship between the political background and the household investment behavior, using panel data from households in 2014, 2016, and 2018. Second, we analyze the mechanisms of the political background affecting the investment behavior and further considered the heterogeneity between geographical regions. Third, we also consider the endogeneity concern of the political background and employ “distance from the provincial capital city” to construct the instrumental variable, mitigating the endogeneity issue. Thus, our results expand the existing research to examine the political background effects on the household investment behavior, while taking into consideration of the household portfolio heterogeneity.

The structure of the article is as follows. Section 2 provides a review on the existing literature and develops our research hypotheses. Section 3 introduces our models and defines the variables. Section 4 reports the empirical results and the robustness test. Section 5 conducts empirical research on the working mechanism, and Section 6 concludes.

2. Literature Review and Hypotheses

2.1. Political Background and Household Asset Allocation Behavior

Previous work by sociologists has recognized the role of the political background on the economic status acquisition, and thus, individuals with certain political status may have an advantage in the asset allocation behavior. First, households with political background have higher cognitive ability, better leadership qualities, and more household capital, affecting household investment behavior (Tong, Zhu, and Yang 2011; Wei and Zhong 2017; Yang and He 2019). For example, they are doing well to collect and process financial information (Smith, Mcardle, and Willis 2010), diversify investment risks, and reduce the risk of losses (Liu and Zhang 2018). These advantages help to reduce the participation cost in the financial market, thereby promoting household investment.

In addition, previous studies have found that households with political background have more advantages in economic activities. Zhang, Giles, and Rozelle (2012) document that the household with political background appears to have a higher income than other households because of political status. Moreover, cadres can turn their political privileges into economic advantages and seizing the opportunities. For example, they are able to obtain additional economic benefits from the redistribution of land, house, and other public assets (Walder 1994). Households with political background benefit more from the reform of housing privatization and benefit within the system, thus reducing the cost of living (Walder and He 2014). It is also documented that their relatives and children can also benefit from political connections (Zhang, Giles, and Rozelle 2012).

However, some households do not invest in risky financial assets even though they have more wealth (Guiso and Sodini 2013). There may be due to other factors that lead to differences in asset allocation behavior. The household with political background usually has better social networks because of their political status, thus promoting the household financial market participation by expanding financial information acquisition (He et al. 2017) and enhancing the household risk-sharing ability (Fafchamps and Gubert 2007). On the other hand, political background helps households obtain loans quickly (Li, Hermes, and Lensink 2020), thus, households are more likely to invest in the risky financial market with lower liquidity but higher yield. Based on this, we propose the following hypothesis:

Hypothesis 1: Political background has a positive effect on the risky financial market participation.

2.2. The Mediating Role of Household Wealth

Households with political background not only have higher income than others, but also exploit personal political status for economic benefit through legal or illegal ways, thus promoting wealth accumulation (Wu 2002; Zhang, Giles, and Rozelle 2012). For instance, they may have allocated fertile land, better equipment, and profitable enterprises to their household members at favorable prices (Friedman and Oi 1990); they may also obtain property in the housing privatization reform at a lower price (Yi, Huang., and Fan 2016). In addition, political advantages can also be passed from generation to generation with positive spillover effects on household members (Tan 2015; Zhang, Giles, and Rozelle 2012). For example, cadres’ relatives and children receive highly paid positions through political connections (Walder 1994).

Wealthy households are more willing to participate in financial markets and take more risks in their portfolios. Campbell (2006) uses the U.S. Consumer Finance Survey (SCF) data and concludes that current assets and cars dominate poor households’ asset allocation, while equity investment plays an important role in middle-class households’ asset allocation. According to the dynamic optimization model, there exists a positive relationship between the financial market participation from households and the increase of households wealth positions (Calvet and Sodini 2010). Based on this, we propose the following hypothesis:

Hypothesis 2: Household wealth mediates the relationship between political background and the risky financial market participation.

2.3. The Mediating Role of Social Networks

Political background brings some information advantages, which positively impact investment decisions (Christiansen, Joensen, and Rangvid 2008).

On the one hand, households with political background have more opportunities to build personal relationships with the government, financial institutions, and enterprises. Therefore, these enable them to obtain more financial information, investment opportunities, and professional knowledge through communication, which in turn reduce the cost of financial market participation (Brown et al. 2008; Durlurf 2004). On the other hand, the expansion on social networks can also strengthen households’ risk-sharing ability, and thus more willing to participate in risk-taking activities (Fafchamps and Gubert 2007). Based on this, we propose the following hypothesis:

Hypothesis 3: Social networks mediate the relationship between political background and risky financial market participation.

2.4. The Mediating Role of Credit Constraints

Political background provides easy access for households to obtain bank loans (Li, Hermes, and Lensink 2020). On the demand side, households with political background are more likely to apply credits than households without such background because they are more confident during banks’ scrutiny. On the supply side, the government’s employees have relatively stable income, this may affect loan applications. Moreover, participation in political activities offer them opportunities to cultivate personal relationships with others, such as bank officials. Frequent contacts facilitate mutual trust and eliminate information asymmetry between households and bank officers, thus helping households receive preferential treatment (Kung and Ma 2016) and get loans (Banerji, Duygun, and Shaban 2016; Cole 1998).

Political background transforms intangible assets into tangible credit resources. Therefore, households with political background are more likely to allocate their assets to the risky financial market with low liquidity but high yield (Wu and Yin 2019). Based on this, we propose the following hypothesis:

Hypothesis 4: Credit constraint mediates the relationship between political background and the risky financial market participation.

3. Data and Methodology

3.1. Data Sample

We collect our data on China Family Panel Studies (CFPS) from the Peking University Open Research Data Platform for 2014, 2016, and 2018. CFPS is a nationwide annual longitudinal survey first introduced by the Institute of Social Science Survey (ISSS) of Peking University in 2010. It is designed to collect individual-level, family-level, and community level longitudinal data. The survey focuses on the economic and well-being of the Chinese population, including economic activities, education outcomes, family dynamics and relationships, migration, and health. Specifically, our data sample includes household demographic characteristics, income, assets, and liabilities. In the 2014 CFPS survey, 14,237 households from 25 provinces, municipalities, and autonomous regions were conducted through face-to-face interviews. Until 2018, 86% of the samples have been followed up. We only include household-head who are 18 years old and above. Therefore, our final-balanced panel data includes 6,317 households for our sample period and results 18,951 household-year observations.

3.2. Variable Constructions

3.2.1. Measuring the Household Investment Behavior

We measure household asset allocation behaviors from two aspects: whether households participate in the risky financial market and the proportion of risky financial assets households hold. Whether households participated in the risky financial market is a dummy variable that takes the value of one if households invest in risky financial assets, such as stocks, funds, corporate bonds, derivatives, gold, and foreign currencies. In the survey, households report the market value of the financial assets they hold. We compute the proportion of risky financial assets in households’ total assets and take this as an indicator of the risky financial assets proportion.

3.2.2. Measuring Political Background

Political background is the power and influence provided by political status. Many studies measure political background in terms of cadres or Party identity (Cao and Qian 2020; Li, Hermes, and Lensink 2020). Households that have a member work in the government/Party/people’s organization/military and hold a management position are labeled as households with political background. It takes the value of one for households with political background and zero otherwise. We also use Party identity to conduct the robustness test.

3.2.3. Measuring Control Variables

We include three categories of control variables: household-head characteristics, household characteristics, and regional characteristics. Household-head characteristics include age, gender, health condition, marital status, and education level. Household characteristics include the child dependency ratio, the elderly dependency ratio, household size, business, and household income. We also use GDP and urban as a regional variable to represent the level of local development.

In addition, the financial decision-makers are defined as the family member familiar with the household financial situation and make financial decisions in the questionnaire. Therefore, we choose the household financial decision-maker as the household-head to alleviate the problem of Intra-household decisions.

3.3. Sample Statistics

Panel A in Table 1 provides summary statistics of data and how each variable is defined. Among our sample households, 22.38% (4,241) of households have political background. Twenty percent of households without political background participate in the risky financial market, while this figure increases to 31% in households with political background. The risky assets proportion of the nonpolitical background households is 10%, while this proportion increases to 13% in households with political background. Panel B in Table 1 shows the correlation matrix for the variables in this article. Risky market participation and Risky assets proportion are positively correlated, and both are positively correlated with Political background. This finding is consistent with our expectations. We also compare households with political background and those without political background. We present the univariate test across the two groups in Panel C in Table 1. The result suggests that households with political background are more likely to participate in the risky financial market participation compared to households without political background.

Table 1. Summary description of variables

		Full sample		Political background		Nonpolitical background
Variable	Variable definition	Mean	SD.	Mean	SD.	Mean	SD.
Panel A: Descriptive statistic
Dependent Variables
Risky market participation	=1 if household participates in the risky market, 0 otherwise	0.23	0.42	0.31	0.46	0.20	0.40
Risky assets proportion	=the proportion of risky assets to total financial assets	0.10	0.24	0.13	0.27	0.09	0.23
Other control variables
Age	= the age of household-head	49.33	12.23	48.31	11.49	49.63	12.42
Gender	1-male; 0-female	0.51	0.50	0.49	0.50	0.52	0.50
Health condition	1-vary bad; 5-very well	2.87	1.22	2.95	1.13	2.84	1.24
Marriage condition	=1 if married	0.89	0.32	0.91	0.29	0.88	0.32
Education	= the years of education	7.24	4.67	9.43	4.14	6.61	4.62
Child dependency ratio	=the number of children under 14 years old/household size	0.14	0.17	0.12	0.15	0.14	0.17
Old dependency ratio	=the number of people over 65 years old/household size	0.15	0.25	0.11	0.20	0.17	0.26
Household size	=the number of a household member	4.05	1.80	4.04	1.73	4.05	1.82
Business	=1 if household member set up a business	0.08	0.28	0.06	0.23	0.09	0.29
Household income	=annual per capita household income in RMB/10000	4.53	5.26	6.59	6.16	3.94	4.81
Urban	=1 if urban	0.46	0.50	0.69	0.46	0.39	0.49
GDP	=provincial Gross Domestic Product/trillion Yuan	31.31	21.73	30.59	20.71	31.52	22.01
Panel B: Correlation coefficient
Variables	Risky market participation	Risky assets proportion			Political background
Risky market participation	1
Risky assets proportion	0.741***	1
Political background	0.110***	0.080***			1
Panel C: T-test of difference in means
Variables	Political background = 0	Political background = 1			T-test of diff. in means
Risky market participation	0.201	0.311			−0.110***
Risky assets proportion	0.086	0.131			−0.046***

Table 1 presents summary statistics of our sample and illustrates how each variable is defined. The risky market participation takes the value of 1 if a household participates in the risk financial market and 0 otherwise. The proportion of risky financial assets is the ratio between the investment of the risky financial assets to the total financial assets. Control variables include age, gender, health condition, marriage condition, education, child dependency ratio, old dependency ratio, household size, business, household income, urban, and GDP. ***Indicate significance at the 1% level.

Regarding the control variables, for households without political background, the household head’s average educational level is 6.61 years (equal to junior high school). For households with political background, this figure is 9.43 years (equal to senior high school). Furthermore, households with political background earn twice as much as those without political background. The results indicate that the human capital of political households is generally higher than that of other households. In addition, we find that households without political background are more likely to set up a business.

3.4. Econometric Approach

To assess the effect of political background on household asset allocation behavior, we employ the Logit and OLS regressions. The panel data enable us to control for unobserved dimensions of ability. Because higher risky financial market participation of cadre household may not be driven by political background, instead that high ability individuals are recruited into public service (Morduch and Sicular 1998).

First, we use the Hausman test to choose between fixed effects and random effects. The estimator was 187.20, and the original hypothesis that the random effect model was rejected at the significance level of 1%. Therefore, we use the fixed effect model.

Moreover, the sample is balanced panel data, and the dependent variables are dummy variables. The fixed effect Probit model has the problem of associated parameters, so we employ the Logit model for regression. For the risky market proportion, since the model is nonlinear, standard panel data methods, such as fixed effects, would not produce consistent estimators. A commonly recommended solution that is found to be reasonably successful in empirical applications is to use a linear probability model. Therefore, we employ the OLS regression.

In the first regression, the dependent variable is households’ risky financial market participation probability, which is a dummy variable. Therefore, the Logit model is used as follow:

(1) $P\left(Y_{i,t}=1\right)=\varphi \left({\alpha }_1+{\beta }_1\times Politica{l}_{i,t}+{\beta }_2\times X_{i,t}+{\sigma }_i+{\epsilon }_{i,t}\right)$(1)

where $Y_{i,t}$ equals one if households participate in the risky financial market, 0 otherwise. $Politica{l}_{i,t}$ equals one if households have political background and 0 otherwise. $X_{i,t}$ is a set of control variables, ${\sigma }_i$ represents time fixed effect. ${\epsilon }_{i,t}$ is a random disturbing term.

In the second regression, the dependent variable is the proportion of risky financial assets that households held. We construct the following OLS model:

(2) $Y_{i,}^{\prime }={\alpha }_2+{\beta }_3\times Politica{l}_{i,t}+{\beta }_4\times X_{i,t}+{\sigma }_i+{\epsilon }_{i,t}$(2)

where $Y_{i,t}^{\prime }$ equals the proportion of risky financial assets to total financial assets. $Politica{l}_{i,t}$, $X_{i,t}$, ${\sigma }_i$, and ${\epsilon }_{i,t}$ have the same meaning as the Logit model.

To explore possible mechanisms, we set the mediating model as follows (Baron and Kenny 1986):

(3) $Investmen{t}_{i,t}={\alpha }_3+{\beta }_5\times Politica{l}_{i,t}+{\beta }_6\times X_{i,t}+{\sigma }_i+{\epsilon }_{i,t}$(3)

(4) $Me{d}_{i,t}={\alpha }_4+{\beta }_7\times Politica{l}_{i,t}+{\beta }_8\times X_{i,t}+{\sigma }_i+{\epsilon }_{i,t}$(4)

(5) $Investmen{t}_{i,t}={\alpha }_5+{\beta }_9\times Politica{l}_{i,t}+{\beta }_{10}\times Me{d}_{i,t}+{\beta }_{11}\times X_{i,t}+{\sigma }_i+{\epsilon }_{i,t}$(5)

The first step is to perform a regression using Model (3). ${\beta }_5$ is the total effect of political background on household investment behavior. The second step is to perform a regression using Model (4). ${\beta }_7$ measures the effect of political background on mediating variables. If the coefficients are statistically significant, the political background is then able to explain the variation of the mediating variables. The third step is to perform a regression using Model (5). ${\beta }_{10}$ measures the effect of the mediating variables on household investment behavior after controlling the political background variable. Thus, if ${\beta }_{10}$ is statistically significant, the mediating effect exists. We also use the Sobel test to verify the mediation effect.

4. Empirical Results

4.1. Baseline Results

Table 2 presents our baseline results using the Logit and OLS regressions to examine the impact of political background on household asset allocation behavior. Columns (1) to (3) use the households’ risky financial market participation as the dependent variable to perform the Logit regressions. Column (1) performs the logit model without other control variables. In Column (2), we gradually add household-head characteristics, household characteristics, and regional characteristics. In Column (3), we control for time fixed effects. Similarly, Column (4) to (6) use the proportion of risky financial assets households hold as the dependent variable to perform OLS regressions.

Table 2. Effects of political background on household asset allocation behavior

	Risky market participation			Risky assets proportion
Variables	(1)	(2)	(3)	(4)	(5)	(6)
Political background	0.044*	0.036*	0.036*	0.019***	0.014**	0.014**
	(0.023)	(0.020)	(0.020)	(0.007)	(0.007)	(0.007)
Age		−0.005***	−0.005***		−0.001***	−0.001***
		(0.001)	(0.001)		(0.000)	(0.000)
Gender		0.070***	0.070***		0.013***	0.013**
		(0.016)	(0.016)		(0.005)	(0.005)
Health condition		0.002	0.002		0.000	−0.000
		(0.006)	(0.006)		(0.002)	(0.002)
Marriage condition		0.020	0.019		−0.004	−0.006
		(0.031)	(0.031)		(0.010)	(0.010)
Education		0.041***	0.012**		0.001*	0.001*
		(0.008)	(0.005)		(0.001)	(0.001)
Child		0.017	0.024		0.006	0.004
		(0.063)	(0.064)		(0.022)	(0.022)
Old		−0.022	−0.011		−0.054***	−0.018*
		(0.029)	(0.034)		(0.009)	(0.011)
Household size		−0.014**	−0.015**		−0.003	−0.002
		(0.007)	(0.007)		(0.002)	(0.002)
Business		0.070***	0.068**		0.033***	0.034***
		(0.026)	(0.027)		(0.010)	(0.009)
Household income		0.003**	0.004***		0.000	0.000
		(0.001)	(0.001)		(0.000)	(0.000)
Urban		0.018	0.018		−0.019	−0.012
		(0.040)	(0.041)		(0.013)	(0.013)
GDP		0.001	0.001		0.002***	0.000
		(0.001)	(0.001)		(0.000)	(0.000)
Constant				0.092***	0.252***	0.164***
				(0.002)	(0.023)	(0.024)
Year Fixed Effect	No	No	Yes	No	No	Yes
N	18,951	18,951	18,951	18,951	18,951	18,951
R-squared	-	-	-	0.01	0.17	0.29

Table 2 presents our baseline results on the relationship between the political background and household investment behavior. Columns (1) to (3) are Logit models, the dependent variable is the risky financial market participation, which takes the value of 1 if a household participates, and 0 otherwise. Columns (4) to (6) are OLS regressions, the dependent variable is the risky asset proportion households invest. ***, **, and *Indicate significances at the 1%, 5%, and 10% levels, respectively. Numbers inside the brackets are p-values for heteroskedasticity robust standard errors.

The coefficients on Political background from Columns (1) to (3) are all positive and statistically significant at the 10% level, indicating that households with political background are more likely to participate in the risky financial market. Coefficients in Columns (4) to (6) remain positive and statistically significant at the 1% level in Column (4) and at the 5% level in Columns (5) and (6), suggesting that households with political background not only participate in the risky financial market but also invest and hold more risky financial markets. Results from Table 1 support our first hypothesis.

Regarding the control variables, the results of household-level variables in Column (3) show that households with more entrepreneurial activity or higher income have a higher risky financial market participation rate, while households with more members and have aged household-heads have lower risky financial market participation. The household-head’s characteristics indicated that households with elderlies have lower participation in the financial market, while higher education levels promote the risky financial market participation. A plausible interpretation is that educated people have better acceptance and understanding of financial knowledge and financial products, more likely to participate in the financial market. Moreover, higher income enables households to afford the participation cost (Vissing-Jorgensen 2002), thereby increasing the financial market participation.

4.2. Mechanism Analysis

Results in Section 4.1 suggest that households with political background have positive impacts on risky financial participation and asset allocation. In this section, we attempt to clarify the possible mechanism between political background and household asset allocation behavior. We choose three potential mechanisms: household wealth, social networks, and credit constraints. We therefore test the mediation effect using Models (3) to (5), and the results are presented in Table 3.

Table 3. The mediating mechanisms on the political background and household asset allocation behavior

Panel A: The mediating role of wealth
	(1)	(2)	(3)	(4)	(5)
	Risky market participation	Risky assets proportion	Wealth	Risky market participation	Risky assets proportion
Political background	0.169*	0.160*	0.061***	0.021*	0.014**
	(0.093)	(0.093)	(0.019)	(0.011)	(0.007)
Wealth				0.025***	0.010***
				(0.005)	(0.003)
Control variables	Yes	Yes	Yes	Yes	Yes
N	18,951	18,951	18,951	18,951	18,951
R-squared	0.40	-	0.59	0.20	0.29
Sobel test, Z				10.84	9.88
P-value				0.000	0.000
Effect size				18.91%	21.05%
Panel B: The mediating role of social networks
	(1)	(2)	(3)	(4)	(5)
	Risky market participation	Risky assets proportion	Social networks	Risky market participation	Risky assets proportion
Political background	0.169*	0.160*	0.021***	0.023**	0.014**
	(0.093)	(0.093)	(0.003)	(0.011)	(0.007)
Social networks				0.019**	0.003***
				(0.008)	(0.000)
Control variables	Yes	Yes	Yes	Yes	Yes
N	18,951	18,951	18,951	18,951	18,951
R-squared	-	-	-	0.19	0.29
Sobel test, Z				8.692	8.165
P-value				0.000	0.000
Effect size				13.12%	16.31%
Panel C: The mediating role of credit constraints
	(1)	(2)	(3)	(4)	(5)
	Risky market participation	Risky assets proportion	Credit constraints	Risky market participation	Risky assets proportion
Political background	0.169*	0.160*	−0.020**	0.022*	0.017**
	(0.093)	(0.093)	(0.009)	(0.012)	(0.008)
Credit constraints				−0.017***	−0.009***
				(0.004)	(0.002)
Control variables	Yes	Yes	Yes	Yes	Yes
N	18,951	18,951	18,951	18,951	18,951
R-squared	-	-	0.21	0.29	0.34
Sobel test, Z				−2.144	−2.433
P-value				0.032	0.015
Effect size				6.80%	1.32%

Table 3 presents the mediating mechanisms of the political background and the household investment behavior through three channels: credit constraints in Panel A, social networks in Panel B, and household wealth in Panel C. The mediation effects are tested using Models (3) to (5). ***, **, and *Indicate significances at the 1%, 5%, and 10% levels, respectively. Numbers inside the brackets are p-values for heteroskedasticity robust standard errors.

4.2.1. The Mediating Role of Wealth

Following previous studies, we measure wealth based on household assets. In Panel A of Table 3, political background affects household asset allocation behavior through the mediator of wealth. Results indicate that households with political background can obtain economic benefits using personal political status, thus increasing household wealth and participation in the risky financial market. Hypothesis 2 is supported.

4.2.2. The Mediating Role of Social Networks

Social networks are relatively stable systems composed of interpersonal relations (Wellman and Berkowitz 1988), through which people can share information and risk (Fafchamps and Gubert 2007), thus increasing the risky financial market participation. We use “the total amount of money that households have given/received in gifts and cash” to measure the social network, reflecting the number of relatives and friends owned by the household. The more number of relatives and friends, the more extensive social networks they may have. The social networks may help to mediate the financial market participation in the households.

Panel B in Table 3 reports the results of the role social networks play on mediating the financial market participation from households. The coefficients on the social networks are positive and statistically significant at the 1% level both in Columns (4) and (5). This suggests that households with political background have more extensive social networks because of their status, which help them obtain financial information, reduce information acquisition costs, and affect their investment behavior. In addition, social networks based on political background allow households to share risk within social networks, improve household risk-sharing ability, and promote household investment behavior. This in turn supports our hypothesis 3.

4.2.3. The Mediating Role of Credit Constraints

The credit constraint indicates that the availability of household credit is limited. We measure the credit constraints based on the question: “whether there is any experience of loan rejection”. We take the value of 1 if the respondent is “yes”, and 0 otherwise. The results are shown in Panel C of Table 3. The coefficients on the credit constraints are negative and statistically significant at the 5% level in Column (4) and at the 1% level in Column (5). These results indicate that political background helps households obtain credit support, enabling them to invest in the risky financial markets, which support our hypothesis 4.

4.3. Heterogeneity Analysis

In this section, we investigate the regional heterogeneous impacts of political background on households investment behavior. The effect of political background on household investment behavior is likely to be heterogeneous across regions and depends on the local economy’s development (Oi and Rozelle 2000). Cadre election competition is fierce in developed regions, which indicates the returns to cadre are higher in absolute and relative terms than in poor regions (Zhang, Giles, and Rozelle 2012). On the other hand, cadres in developed regions also have a better ability, thus promoting households risky market participation (Wei and Zhong 2017). However, in remote and impoverished regions, cadres may not have much power or advantages based on their political status (Song and Zhong 2009). Therefore, we further split our sample of households into two groups, depending on whether they are living in the Eastern region or urban region. Table 4 presents the heterogeneity analysis on the political background and the households investment behavior.

Table 4. Heterogeneity analysis

	Risky market participation		Risky assets proportion
Variables	(1)	(2)	(3)	(4)
	0.472***	0.360***	0.003	0.015
	(0.075)	(0.103)	(0.008)	(0.010)
Eastern region	0.149		0.012
	(0.093)		(0.048)
Political background* Eastern	0.278*		0.002
	(0.147)		(0.016)
Urban		0.217***		0.011
		(0.071)		(0.014)
Political background* Urban		0.154**		0.009**
		(0.061)		(0.004)
Control variables	Yes	Yes	Yes	Yes
Year Fixed Effect	Yes	Yes	Yes	Yes
N	18,951	18,951	18,951	18,951

Table 4 presents the heterogeneity analysis using the subsample of households. We split our sample households into the Eastern and Urban regions to examine households’ investment behaviors from different regions. ***, **, and *Indicate significances at the 1%, 5%, and 10% levels, respectively. Numbers inside the brackets are p-values for heteroskedasticity robust standard errors.

Firstly, we add the interaction term of political background and dummy variables in Eastern China for the regression analysis. According to the regression results reported in Columns (1) of Table 4, the interaction coefficient between the political background and Eastern region is 0.278, which is statistically significant at the level of 10%, suggesting that the effect of political background on the households investment behavior is more pronounced in Eastern region.

We also test the difference between urban and rural areas and add the interaction term of political background and urban area dummy variable to carry out the regression analysis. The coefficients of the interaction of political background and urban are 0.154 and 0.009, respectively. They are both statistically significant at the 5% levels, indicating that the marginal impact of political background on household risk financial allocation is more significant in urban areas. This is not surprising given the fact that the economic development in the Eastern region and urban is higher compared to other regions in China. Thus, with the development of the local economy, households with political background experience better performance in wealth accumulation, social networks, and investment behavior.

4.4. Robustness Test

In this section, we perform a number of robustness tests, including an alternative proxy for the political background, subsample test, and the IV approach. Results are reported in Table 5 for the alternative proxy and the subsample test, and in Table 6 for the IV approach.

Table 5. Robustness test

	Risky market participation			Risky assets proportion
Variables	(1)	(2)	(3)	(4)	(5)	(6)
Party	0.437***			0.008*
	(0.071)			(0.004)
Status		0.052**			0.009**
		(0.025)			(0.004)
Political background			0.438***			0.013*
			(0.067)			(0.008)
Age	−0.027***	−0.027***	−0.026***	−0.001***	−0.001***	−0.001**
	(0.003)	(0.003)	(0.003)	(0.000)	(0.000)	(0.000)
Gender	0.233***	0.244***	0.252***	0.013**	0.012**	0.014**
	(0.053)	(0.053)	(0.055)	(0.005)	(0.005)	(0.006)
Health condition	0.027	0.024	0.030	0.000	0.000	0.001
	(0.022)	(0.022)	(0.023)	(0.002)	(0.002)	(0.002)
Marriage condition	0.328***	0.353***	0.360***	−0.005	−0.005	−0.001
	(0.092)	(0.092)	(0.100)	(0.010)	(0.010)	(0.012)
Education	0.047***	0.053***	0.051***	0.001	0.001	0.000
	(0.006)	(0.006)	(0.007)	(0.001)	(0.001)	(0.001)
Child	0.094	0.032	0.081	0.003	0.002	0.007
	(0.189)	(0.190)	(0.197)	(0.022)	(0.022)	(0.024)
Old	−0.208*	−0.220*	−0.100	0.017	0.016	0.010
	(0.125)	(0.125)	(0.144)	(0.011)	(0.011)	(0.013)
Household size	−0.069***	−0.063***	−0.073***	−0.002	−0.002	−0.003
	(0.018)	(0.018)	(0.019)	(0.002)	(0.002)	(0.003)
Business	0.696***	0.693***	0.757***	0.033***	0.033***	0.033***
	(0.087)	(0.087)	(0.089)	(0.009)	(0.009)	(0.010)
Household income	0.028***	0.028***	0.025***	0.000	0.000	0.000
	(0.004)	(0.004)	(0.004)	(0.000)	(0.000)	(0.000)
Urban	0.328***	0.356***	0.238***	0.011	0.011	0.015
	(0.063)	(0.064)	(0.067)	(0.013)	(0.013)	(0.015)
GDP	0.000	0.000	0.000	0.000	0.000	0.001*
	(0.001)	(0.001)	(0.001)	(0.000)	(0.000)	(0.000)
Year Fixed Effect	Yes	Yes	Yes	Yes	Yes	Yes
N	18,951	18,919	16,972	18,951	18,919	16,972

Table 5 presents robustness tests using an alternative proxy for political background in Columns (1), (2), (4) and (5), also using a subsample in Columns (3) and (6). ***, **, and *Indicate significances at the 1%, 5%, and 10% levels, respectively. Numbers inside the brackets are p-values for heteroskedasticity robust standard errors.

Table 6. Robustness test

	Risky market participation	Risky assets proportion
Variables	(1)	(2)
Political background	0.246*	0.105**
	(0.061)	(0.035)
N	18,951	18,951
Control variables	Yes	Yes
Year Fixed Effect	Yes	Yes
Under identification test	300.692***	300.692***
F-value	245.086	245.086

Table 6 presents a further robustness test using the IV approach. We use the distance from the provincial capital city as the instrumental variable (IV). ***, **, and *Indicate significant at the 1%, 5%, and 10% level. Numbers inside the brackets are p-values for heteroskedasticity robust standard errors.

4.4.1. Alternative Proxy for Political Background

Party identity is a symbol of political status and considered to have higher political status in China. We use Party identity (i.e., whether there is a Party member in the households) and social status (i.e., the social status of the family) as alternative measures of the political background. First, we examine the relationship between Party identity and household investment behavior. Columns (1) and (4) in Table 5 present results for the Party Identity as an alternative proxy for the political background. The coefficients on Party identity are positive and statistically significant at the 1% level in Column (1), at the 10% level in Column (4). In addition, the coefficients on social status are positive and statistically significant at the 1% level in Column (2), at the 10% level in Column (5) as well.

4.4.2. Subsample Test

We drop the sample whose household head is older than 65 years old to reestimate our baseline regressions. Results are presented in Columns (3) and (6) of Table 5. The coefficients on Status (our subsample indicator) are positive and statistically significant at the 5% level, indicating our baseline results are robust.

4.4.3. Endogeneity Problem

Our results may suffer from the endogeneity problem. For example, there may be a reverse causality between the political background and the household investment behavior. The higher risky assets holding reflect the household’s advantages in individual ability as well. Thus, we use an IV approach to control for reverse causality and omitted variable biases. Results are presented in Table 6.

Geographical distance is regarded as an important indicator of the household political background. The longer distance may diminish the political power. Therefore, households further away from the political center are likely to have less political background. Following Cao and Qian (2020), we used the distance from the provincial capital city as the instrument variable (IV). Moreover, there is no direct relationship between geographical distance and household investment behavior, so the instrument variable is feasible. Table 6 presents the results using the instrument variable. The under identification test shows that Kleibergen-Paap rk L.M. statistic = 300.692, p = .000, strongly rejected the original hypothesis of unidentified, and instrumental variables are strongly correlated with endogenous variables. The F-statistics is 245.086, which is in line with the value recommended to avoid the problem of the weak instrument (Stock and Yogo 2005). The coefficients in Table 6 are positive and statistically significant at the 10% level in Column (1) and at the 5% level in Column (2). This is consistent with our baseline regression and suggesting that our baseline results are robust.

5. Conclusions

Political background is an important resource for households. This article used 2014–2018 household panel data from the China Family Panel Studies (CFPS) to investigate the impact of political background on household asset allocation behavior. Our results suggest that: First, political background has a significant positive impact on the household investment behavior. Second, mediation analysis indicates that political background results higher household wealth, better social capital, and fewer credit constraints, thus, leading to more investment. The results indicate that the impacts of political background on household investment behavior mainly relate to relieving liquidity constraints and participation cost. Third, heterogeneity analysis shows that the marginal impact of political background on household investment behavior is more significant in Eastern and urban areas. In addition, further robustness tests using the instrumental variable approach confirm our baseline results.

Our results expand the research on household investment behavior and enrich the development of household finance. Our conclusions help understand household portfolio heterogeneity and shed extra light on household asset allocations and promote wealth accumulation. In addition to demographic characteristics, household characteristics, and other factors, we believe political background as an intangible asset also affects household investment behavior. Households with political background are more likely to participate in the financial market. On the one hand, political background affects household investment behavior through wealth and credit constraints, which relaxes liquidity constraints. It provides the material basis and economic conditions for households to participate in the financial market. Political background promotes household investment behavior through social networks, which reduces information asymmetry and risk. Therefore, further study on the signaling effect of political background would help to build an information-sharing system between households and financial institutions so that households are able to obtain information on investment and promote the capital market development and economic growth.

Acknowledgments

We are grateful for helpful comments from referees. All errors remain our own.

Additional information

Funding

This work was supported by the National Social Science Fund of China [number 18AJY021]; and the Natural Science Foundation of Shandong Province [number ZR2018MG011].