Democracy, corruption, and endogenous entrepreneurship policy

Abstract

This paper endogenizes pro-entrepreneurship policies in a model where voters choose the strength of these policies and entrepreneurs generate social returns which benefit the median voter. In the model, incumbent firms who are harmed by the greater competition that this policy promotes can push back in two ways: via corruption and persuasion. Specifically, they can bribe elected politicians to break their campaign promises; and they can allocate some of their rents to corporate social responsibility (CSR) initiatives that also benefit voters. The model predicts that corruption which weakens pro-entrepreneurship policy can be completely neutralized by a forward-looking median voter—without removing the incentive among incumbent firms to bribe politicians. In this way, endogenizing entrepreneurship policy can destroy any relationship between corruption and entrepreneurship. Corporate social responsibility initiatives modify this prediction, which provides a novel rationale for CSR that appears to be new to the literature as well.

Appendices

Appendices

1.1 Appendix 1: Illustration with specific functional forms

To illustrate the model, consider the following specific functional forms:

$$\begin{aligned} s(p)= & {} p^\alpha \qquad \qquad \alpha \in [0,1) \\ \pi (x,p)= & {} xp \\ G_X(x)= & {} 1-(x_L/x)^\eta \qquad \qquad \eta >1\quad ,\quad x\in [x_L,\infty ) \\ \end{aligned}$$

The third equation specifies the distribution of X as Pareto, following Itskhoki and Moll (2019).

Using these functional forms, we obtain ${\hat{x}}=w/p$ and two reduced form simultaneous equations in p(z) and z(p), which are both positive, increasing and convex functions:

$$\begin{aligned} p(z)= & {} \left( \frac{\alpha +\eta }{\gamma _m}\lambda z\right) ^{\frac{1}{1-\alpha }} \\ z(p)= & {} \left( \frac{x_L}{{\hat{x}}}\right) ^\eta = \left( \frac{x_Lp}{w}\right) ^\eta \end{aligned}$$

These can be solved for the structural equations

$$\begin{aligned} p_m= & {} \left[ \frac{\gamma _m}{\lambda (\alpha +\eta )}\left( \frac{w}{x_L}\right) ^\eta \right] ^\frac{1}{\alpha +\eta -1} \\ z_m= & {} \left( \frac{x_L}{w}\right) ^\eta \left[ \frac{\gamma _m}{\lambda (\alpha +\eta )}\left( \frac{w}{x_L}\right) ^\eta \right] ^\frac{\eta }{\alpha +\eta -1} \end{aligned}$$

With these functional forms, the solution of the Nash bargaining game is

$$\begin{aligned} b^* = \frac{1}{2}\left[ \Pi \left( \frac{x_L}{w}\right) ^\eta p_m^\eta \left( 1-\mu ^\eta \right) + \frac{\xi w +\phi p_m(1-\mu )}{\psi } \right] \end{aligned}$$

Consistent with Proposition 1, we obtain

$$\begin{aligned} \frac{db^*}{d\Pi }= & {} \frac{1}{2}\left( \frac{x_L}{w}\right) ^\eta p_m^\eta \left( 1-\mu ^\eta \right)>0 \\ \frac{db^*}{d\mu }= & {} -\frac{1}{2}\left[ \eta \Pi \left( \frac{x_L}{w}\right) ^\eta p_m^\eta \mu ^{\eta -1} + \frac{\phi p_m}{\psi } \right] < 0 \\ \frac{db^*}{dp_m}= & {} \frac{1}{2}\left[ \eta \Pi \left( \frac{x_L}{w}\right) ^\eta p_m^{\eta -1} \left( 1-\mu ^\eta \right) + \frac{\phi (1-\mu )}{\psi } \right] > 0 \end{aligned}$$

with the other three derivatives identical to those in the text.

To illustrate Proposition 3, note that the solution when $\theta >0$ is

$$\begin{aligned} \frac{dp^*}{d\theta } \propto -\frac{\eta z(p^*)}{p^*} < 0 \end{aligned}$$

with

$$\begin{aligned} \frac{d^2 p^*}{d\theta ^2} \propto -\eta (\eta -1)\left( \frac{x_L}{w}\right) ^\eta (p^*)^{\eta -2} \frac{dp^*}{d\theta } > 0 \end{aligned}$$

since $\eta >1$.

1.2 Appendix 2: Cross-country correlations

Corruption occurs when public officials are neither required to disclose their finances and potential conflicts of interest, nor face serious consequences, when using their public offices for private gain. These outcomes and facilitators are captured by Transparency International’s Corruption Perceptions Index (CPI), a harmonized index that facilitates international comparisons, where higher values indicate lower levels of lobbying and corruption.⁵ Another corruption index is the World Governance Indicator, WGI, ’Control of Corruption, CC index, which is known to be highly correlated with the CPI Belitski et al., 2015; (Dutta & Sobel, 2016; Audretsch et al., 2021).

A harmonized cross-national measure of entrepreneurship is the index of total entrepreneurship activity (TEA). TEA is the percentage of adults aged 18–64 engaged in early-stage entrepreneurial activity at a given point in time. Data were taken from https://www.gemconsortium.org/file/open?fileId=50691. Figure 1 plots the CPI against TEA for 22 OECD countries for which data were available in 2019; Fig. 2 plots the WGI CC index against TEA for the same set of countries in the same year. Higher values of CPI and WGI indicate lower levels of corruption.

If corruption retards entrepreneurship we would expect to observe a positive empirical relationship between entrepreneurship and these two inverse corruption measures. However, Figures 1 and 2 show that if anything the opposite is the case. There is if anything a small negative relationship between CPI and TEA; but it is not statistically significant (in a double log regression, ${\hat{\beta }}=-0.09$, $p=0.44$). The same is true for the WGI CC index: the regression results also indicate a negatively signed and insignificant relationship (${\hat{\beta }} =-0.09$, $p=0.57$). Moreover, in additional plots (not reported) similar results apply using an ’opportunity entrepreneurship’ variant of the TEA measure. Thus, among developed countries where democratic institutions are generally well entrenched, we see no evidence of a cross-country relationship between entrepreneurship and political corruption.

Fig. 1

Cross-country relationship between corruption (CPI) and entrepreneurship

Fig. 2

Cross-country relationship between corruption (WGI) and entrepreneurship

1.3 Appendix 3: Conditions for optimal $\theta$

Absent CSR, the incumbent makes profits of $\Pi (1-z(p_m))$. With CSR, denote optimal policy by $p^* \le p_m$ [see Proposition 3]. The marginal cost to an incumbent of expending a fraction $\theta$ of profits on CSR is obviously $\theta \Pi (1-z(p_m))$. The marginal benefit to them of CSR which replaces $p_m$ with $p^*$ is

$$\begin{aligned} \Pi [(1-z(p^*))-(1-z(p_m))] = \Pi [z(p_m)-z(p^*)] \end{aligned}$$

Hence the incumbent’s objective function is

$$\begin{aligned} U_I = \Pi [z(p_m)-z(p^*)] - \theta \Pi (1-z(p_m)) \end{aligned}$$

With choice variable $\theta$, the Karush–Kuhn–Tucker conditions for a maximum are:

$$\begin{aligned} {{\mathcal {L}}}_\theta= & {} z(p_m)-1-z' \frac{dp^*}{d\theta } \le 0 \\ \theta\ge & {} 0 \\ \theta \left[ z(p_m)-1-z' \frac{dp^*}{d\theta }\right]= & {} 0 \end{aligned}$$

If $\theta ^*>0$, the second order condition for a maximum is

$$\begin{aligned} -z'\left[ \frac{d^2p^*}{d\theta ^2}+\left( \frac{dp^*}{d\theta } \right) ^2\right] < 0 \end{aligned}$$

A sufficient condition for this inequality to hold is $\frac{d^2 p^*}{d\theta ^2}>0$. As noted at the end of "Appendix 1", this condition holds for the specific functional forms analyzed there. In general, however, a corner solution $\theta ^*=0$ can arise if even a small profit share is too costly relative to the value of the policy change it induces via shifted voter choices. Then corporations do not find any CSR worthwhile, and complete policy neutralization can be expected.