Profit Constraint, Ownership Structure and Indigenous Innovation
LiYong(李勇)....................................................................................................................................................
Abstract: Under a logical self-consistent theoretical framework, this paper discusses SOEs’ innovation efficiency and innovation conundrum facing Chinese local firms. By creating a theoretical model of endogenous technology level, this paper finds that credit discrimination and soft budget constraint have both a crowding out effect and compensatory effect on corporate innovation. When firms engage in less profitable innovation, the compensatory effect outweighs crowding out effect, and a higher share of SOEs will promote the overall level of innovation. On the contrary, when firms engage in more profitable innovation, the compensatory effect is smaller than crowding out effect, and a higher share of SOEs will diminish overall innovation. In this sense, SOE innovation exhibits a threshold characteristic. Then, this paper carries out an empirical test using the inter-provincial panel data of 1997-2013, which proves our assumption. Finally, this paper arrives at conclusions and policy implications.
Keywords: SOEs, compensatory effect, crowding out effect, threshold characteristic JEL Classification Codes: B25, D22, E22
DOI:1 0.19602/j .chinaeconomist.2019.3.099602/ j .chinaeconomist.2018.09.02
1. Introduction
As China’s economy enters the new normal, indigenous innovation has become a key driver of sustained economic growth. But in reality, Chinese local enterprises (including SOEs, collective and private enterprises) are still dwarfed by manufacturers from developed countries and foreign-funded
1 enterprises in China in terms of R&D input and innovation (An, 2009). Overall, China’s indigenous innovation is far from meeting the needs of its economic development. China’s innovation-driven strategy remains largely unfulfilled. The causes of China’s lack of innovation have been widely discussed in literature.
Some recent studies suggest that lack of access to financing restricts corporate R&D input and innovation (Brown et al., 2009, 2011; Hall and Lerner, 2010, et al.). For large transition economies
like China, however, the level of financing constraint varies across enterprises of different ownerships. Access to finance is much easier for SOEs compared with collective and private enterprises (Cull and Xu 2002; Brandt and Li, 2003). A smaller proportion of state-owned shares corresponds to greater uncertainties in innovation and less corporate R&D input and innovation (Shen, et al., 2012; Zhang, et al., 2012). Financing constraint, which only faces collective and private enterprises, therefore should not be blamed for the lack of SOEs’ and overall innovations in China.
In a host of papers, Wu ( 2012a, 2015) et al. attribute the shortsightedness of SOEs to their ownership defects. Although reforms have enabled SOEs to claim and control their production surpluses, there remains a mismatch in their rights to claim and control innovation surpluses. This has led to losses in both production and innovation efficiency, particularly the latter. Dai, Zhang (2013) and Yang (2015) find that credit discrimination in favor of SOEs has “crowded out” “innovation funds” for private firms. Under the condition of SOE innovation inefficiency, China’s overall innovation is negatively correlated with credit discrimination and the share of SOEs.
Based on the above analysis, the key to China’s lack of innovation is the inefficiency of SOE innovation. However, some studies believe otherwise. For instance, public sector plays an important role in compensating for the lack of private sector investment (Arrow, 1962). Significant positive externalities of innovation will encourage SOEs, which boast guaranteed profit and soft budget constraint, to increase R&D spending, creating a spillover effect on the R&D input of private firms (Blanes, 2004; Hussinger, 2008, et al.). Therefore, innovation efficiency is not correlated with ownership structure. When the externalities are significant, SOEs are likely to become efficient at innovation (Li, Lu, 2014; Cheng et al., 2015).
SOE innovation efficiency is an essential part in the discussions on China’s innovation conundrum in the new normal. In the literature, neither discrimination nor inefficiency theory is able to explain SOE innovation inefficiency and China’s innovation challenges at the same time. We reckon that in addition to “crowding out effect,” credit discrimination and soft budget constraint also have a “compensatory effect” on innovation when firms engage in less profitable innovations. In this case, a higher share of SOEs will increase the overall level of innovation in China. For profitable innovations, however, credit discrimination and soft budget constraint will have a crowding out effect, and a higher share of SOEs will reduce the level of innovation in China. Hence, SOE innovations are subject to a threshold effect. To prove this hypothesis, this paper creates a theoretical model of endogenous technology level, and discusses SOE innovation efficiency and challenges under a logically self-consistent framework. This theoretical model offers a new explanation on the innovation conundrum facing Chinese firms, together with a theoretical basis for evaluating SOE innovation efficiency - and its critical conditions - in the new
2 normal. As a marginal contribution, this paper identifies the criteria for SOEs to become more efficient at innovation, and discusses SOE innovation efficiency and challenges under a logically self-consistent framework.
The rest of this paper is arranged as follows: Part 2 creates a theoretical model for the analysis of the threshold effect of SOE innovation, and arrives at relevant propositions; Part 3 verifies these propositions using panel threshold regression model; and Part 4 offers conclusions and policy implications.
2. Theoretical Model
As mentioned before, credit discrimination and soft budget constraint have a crowding out effect and a compensatory effect on SOE innovations, which exhibit a threshold effect. Crowding out effect
means that SOEs may use government resources or connections to influence banks’ credit decisions to their advantage despite being less efficient at innovation. Compensatory effect means that if innovation derives significant externalities, credit discrimination and soft budget constraint will guarantee profits that induce SOEs to spend more on R&D; these externalities also promote innovation in the private sector and beyond.
To verify SOE innovation’s threshold effect, this paper considers an economic environment as follows: Firms are divided into SOEs (S) and private firms (P) by ownership, and both engage
3
in innovation ( H) and routine production ( L). We assume that their production functions4 are
and respectively. Where, YS , YP
are the output levels of SOEs and private firms. , , and respectively denote the technology levels of SOEs and private firms in their innovation and routine production activities. , , and respectively denote the capital inputs of SOEs and private firms.
To examine the relationship between innovation and capital demand, we further assume that:
( , ). In this equation, is the capital efficiency of SOEs and private firms for innovation and routine production. Under the effect of credit discrimination and soft budget constraint, it is likely that SOEs only devote a portion of their funds to innovation and daily operations, and use the remainder for perquisite consumption, additional return on control rights, rent-seeking and other nonproductive activities (Dai, Zhang, 2013). Compared with non-SOEs, SOE capital efficiency is poor, i.e. . The production functions of SOEs and private firms can be finally expressed as:
(1)
2.1 Technology Level under the Baseline Scenario
Under the baseline scenario, credit allocation is free from discrimination against or in favor of firms of any ownership type, and both SOEs and private firms choose their capital inputs and corresponding levels of technology. Due to hefty early-stage investments and sunk cost, short-term profits for SOE innovation are limited. When innovation is less profitable than routine business operations ( ), SOEs are likely to suspend innovation programs. By such definition, the profit function of enterprises of different ownership types can be re-expressed as: Then, programming equation (5) is employed to solve the technology levels of SOEs and private 7 firms, which are: It can be proven that under the baseline scenario, difference in the technology level of firms of different ownership types ( ) can be expressed as: As can be found from equation (9), under the baseline scenario, neither credit discrimination nor
soft budget constraint exist, and capital efficiency is the same for firms of different ownership types, i.e.
. Thus, , and the technology levels of SOEs and private firms converge. But the impact of financing constraint on innovation is also discernable. Namely, when innovation is less profitable than routine production, all firms will be affected by financing constraint and give up innovation programs.
2.2 Technology Level under Credit Discrimination and Soft Budget Constraint
Under the effects of credit discrimination and soft budget constraint, SOEs may use government resources or avenues to influence banks’ credit decisions. This forces private firms to take into account the capital input and technology level of SOEs when deciding their own capital input and technology level. In selecting capital input and technology level, SOEs enjoy a “first mover advantage.” The structure of such a game can be illustrated as follows:
As Figure 1 shows, under credit discrimination, competition for credit resources between SOEs and private firms are divided into two stages:
Stage 1: SOEs first determine an amount of capital input, and then decide the corresponding level of technology;
Stage 2: private firms determine their own amount of capital input according to the capital input of SOEs, and then determine the relevant level of technology.
In addition, with the “soft budget constraint,” if innovation is less profitable than routine production, SOEs may apply for government subsidies and grants to keep profits at the level, whereas private firms have to abandon their innovation programs. In this case, the profit function of SOEs and private firms can be reformulated as follows: By substituting equation (10) and (11) into the programming equation (5), we may obtain the
It can be found through equations (14) and (15) that compared with the baseline scenario, the level of innovation for SOEs has increased under the financing environment of credit discrimination and soft budget constraint ( ). However, this comes at the expense of private firms’ innovation ( ).
Furthermore, we assume that the share of SOEs is , and the share of private firms is . Then, it can be proven that:
As discussed above, under the credit discrimination and soft budget constraint, the capital efficiency of SOEs is relatively low ( ). In this case, it can be found based on equations (14), (15) and (16) that:
(1) When innovation is more profitable than routine production ( ), the crowding out effect of credit discrimination and soft budget constraint on innovation outweighs compensatory effect8. While easing SOE financing constraint and partially increasing SOE innovation (
), this also crowds out innovation funds that would have otherwise been available to private firms, causing the latter’s financing constraints and diminishing innovation ( ). Therefore, with less capital efficiency of SOEs, credit discrimination and soft budget constraint will reduce the overall level of innovation ( ). A higher share of SOEs means a greater reduction in the overall level of innovation ( );
When innovation is less profitable than routine production ( ), the crowding out effect of
9 credit discrimination and soft budget constraint on innovation is smaller than the compensatory effect. Credit discrimination and soft budget constraint allow SOEs to maintain innovations profitability at the same level with routine production ( ), while private firms are forced to quit their innovation programs. Despite capital inefficiency, SOEs maintain the level of innovation at , thus making up for the innovation gaps that resulted from financing constraint. In this case, credit discrimination and soft budget constraint have increased the overall level of innovation. A higher share of SOEs means a more significant improvement in the overall level of innovation ( ).
Based on the above discussions, this paper puts forward Proposition 1 and Proposition 2 that can be tested: Proposition 1: When innovation is less profitable, credit discrimination and soft budget constraint will increase the overall level of innovation; otherwise, credit discrimination and soft budget constraint will reduce the overall level of innovation.
Proposition 2: When innovation is less profitable, a higher share of SOEs means faster growth in overall innovation; otherwise, a higher share of SOEs means slower growth in overall innovation.
3. Empirical Test
It can be found through the theoretical model in Part 2 that when innovation is less profitable,
SOE innovation will promote the overall level of innovation. Otherwise, SOE innovation will restrain the overall level of innovation. To verify this theoretical hypothesis, this paper carries out an empirical analysis of the interprovincial panel data of 29 provinces, municipalities and autonomous regions over the 1997- 2013 period, excluding Hainan, Tibet, Hong Kong, Macao and Taiwan. The purpose is to investigate SOE overall innovation efficiency range and critical conditions.
3.1 Definition of Data and Variables
Based on Table 1, indicators selected by this paper include:
3.1.1 Explained variable
Considering the drawback of patent applications in measuring the level of domestic firms’ innovation, recent studies (Fan et al., 2008; Zhang et al., 2014) have employed R&D investment as a proxy indicator for the level of innovation. According to our analysis, credit discrimination and soft budget constraint affect not only R&D spending as innovation input but the output of innovation - invention patents - as well. Hence, this paper uses the growth rate of R&D spending and the growth rate of valid patent inventions to reflect the level of innovation.
3.1.2 Core explanatory variable
Based on the above propositions, SOE innovation activities exhibit a threshold effect, which is manifested through credit discrimination and soft budget constraint. Therefore, this paper’s core explanatory variables include credit discrimination, soft budget constraint and the share of SOEs. Specifically:
(1) Relative share of loans to SOEs (RSLOAN): This paper uses the share of loans to SOEs as a proxy indicator for credit discrimination. Regretfully, there is no original RSLOAN data in available publications. Referencing Wang’s (2015) et al. treatment method, this paper carries out an estimation based on the panel fixed effect model of first-order residue autocorrelation through the following steps: a. First, we collected the original data of “gross industrial output value of large state-owned industrial enterprises in a region” and “gross industrial output value of large enterprises in the region” to calculate the share of large SOEs in total regional output value. b. We then collected the original data of “total bank credit in the region.” c. Assuming that there exists a fixed relationship between total credit and the output value of SOEs in a region, we create a panel model using the proportion between total bank credit and the output value of SOEs in a region to further arrive at the coefficient of fixed proportion between
11
total bank credit and the share of SOEs in total industrial output value. d. Total bank credit in a region is multiplied by the fixed proportion coefficient to obtain the absolute share and relative share of loans issued to SOEs.
(2) Intensity of government support (GOV): Based on our survey of literature, discussions on soft budget constraint focus on the qualitative level. From a quantitative point of view, Li et al. (2007) uses general and administrative expenses divided by sales income and general and administrative expenses as alternative indicators for agency cost to discuss the corporate efficiency impact of soft budget constraint, but their study is confined to a microscopic level. From a macro level, the allocation of government subsidies is also biased in respect of corporate ownership. Assuming that the majority of government R&D funds are received by SOEs, this paper uses the level of government support as a proxy indicator for soft budget constraint to further analyze its impact on innovation.
(3) Share of SOEs (SOE and SOE1): From both input and output perspectives, this paper identifies the share of fixed asset investments by the state sector of economy and the share of output value from the state sector of economy as two variables to reflect how change in the share of the state sector of economy influences overall innovation level. Specifically, the fixed asset investment and output value of large state-owned and state-controlled enterprises in a region are divided by the investment and output value
12
of large enterprises in the region.
3.1.3 Threshold variable
Based on prior analysis, when profits from innovation vary across firms, credit discrimination, soft
budget constraint and innovation will exhibit a threshold effect. Hence, this paper selects profitability as threshold indicator, denoted as the profit growth of large industrial enterprises in a region / total profits of large enterprises in the region.
3.1.4 Control variable
For other important factors that influence the level of innovation, this paper adopts the following control variables referencing relevant literature (Zhang, Zhou, 2011; Cheng et al., 2015): foreign direct investment (FDI), trade openness (OPEN), human capital (HUMCAP), scale (SCALE) and market segmentation (MSEG) to denote the innovation effects of trade openness, company characteristics and other institutional factors.
All the above data is from China Compendium of Statistics 1949-2004, China Economic Statistical Yearbook, Statistical Yearbook of China’s Industrial Economy, and China Financial Statistical Yearbook. The original data for total regional import and export volume has also been adjusted by the central parity rate. After conversion of baseline price index with 1998 as base period, we have made exclusions for relevant variables using the fixed baseline price index.
3.2 Test model
This paper’s estimation equations for Propositions 1 and 2 are as follows13:
3.3 Estimation Result
To estimate the specific threshold value , this paper first conducts a grid search. Specifically, calculated profitability (PRO) is ranked in an ascending order, and 10% observations at the top and bottom are removed based on Hansen’s (2000) suggestion. On such a basis, profitability is selected as the threshold value to conduct an estimation using equation (18) and obtain its residual. Then, equation (18) is used to find the estimated threshold value, and then simulate the likelihood test using the bootstrap method (repeated for 3,000 times in this paper) to determine whether the threshold effect exists. The test result of threshold effect is shown in the following table:
Estimation result of the threshold effect test shows that at 5% significance level (except for SOE and R&D at 10%), there exists a significant threshold effect for credit discrimination (RSLOAN), intensity of government support (GOV), the share of SOEs (SOE and SOE1) and growth rate of patent applications ( IPR), which explains that using panel threshold model for estimation is appropriate. Estimated threshold value is tabulated below:
With the estimated threshold values in Table 4, we may conduct a panel threshold estimation with
14 the following result:
As can be found from the estimation result of Table 5, the adjusted coefficient of determination for the equations are generally between 40% and 50%, which indicates a good explanatory power of the regression model. F test and Hausmann test suggest that the estimation result of fixed effect is relatively reliable. Judging by the estimation coefficient, most regression coefficients of the core explanatory variable have passed significance test at 5%, which indicates that the threshold effect exists between credit discrimination, soft budget constraint, share of SOEs and innovation. Specifically:
(1) When profitability is low, the regression coefficients of the share of loans to SOEs (RSLOAN) and the intensity of government support (GOV) are significantly positive, which indicates that when firms engage in less profitable innovation, credit discrimination and soft budget constraint will allow SOEs to maintain innovation profitability at a critical level, so as to be able to pursue innovation that would otherwise not happen. Hence, credit discrimination and soft budget constraint have a compensatory effect on China’s overall R&D spending and innovation, and SOEs still have a certain efficiency range. This verifies the conclusions of Lach (2002) and Hussinger (2008). When profitability is high, the share of SOEs and the regression coefficient of government support become negative (take IPR for instance, the coefficients change from 0.1823 and 0.0921 to -0.0324 and -0.1132). This shows that with the increase in profitability of innovation activities, credit ownership discrimination
and soft budget constraint will crowd out funds that otherwise would go to private firms, resulting in a financing constraint and lack of innovation among them, i.e. the crowding out effect. This finding is also consistent with the conclusions of Dai, Zhang (2013) and Yang (2015). Under the assumption of capital inefficiency of SOEs, credit discrimination and soft budget constraint will lead to a reduction in the level
of innovation ( IPR and R&D), i.e. Proposition 1 is proven.
(2) As can be seen from the relationship between the share of SOEs and innovation growth rate, when firms engage in more profitable innovations, a greater share of SOEs corresponds to higher growth rates in patent applications (IPR) and R&D spending (R&D) (coefficients are 0.2067, 0.0721, 0.3322 and 0.1831); on the contrary, a smaller share of SOEs corresponds to smaller growth rates in patent applications ( IPR) and R&D spending ( R&D) (coefficients are -0.1928 and - 0.2467), i.e. Proposition 2 is proven.
( 3) In the regression result of control variables, the regression coefficients of foreign direct investment (FDI) and trade openness (OPEN) are insignificant, which indicates that China’s foreign trade is still dominated by labor-intensive industries and is yet to optimize the factor endowment structure of firms and further promote their innovation. The regression coefficient of scale is significantly negative, which shows that China’s local firms still follow a crude pattern in expanding market share and pursuing monopolistic return. This has led to a lack of innovation and restricted the improvement of innovation (Wu, 2012; Dai, Zhang, 2013, et al.). The coefficient of market segmentation (MSEG) is also negative,
which indicates that the segmentation of product market has restricted innovation to some extent. This finding is consistent with the conclusions of Zhang and Zhou (2011), and will not be elaborated in this
15 paper. The coefficient of human capital (HUMCAP) is significant positive, highlighting the importance of human capital to innovation.
3.4 Robustness Test
Based on the above estimations, this paper identifies the threshold characteristics of credit discrimination, soft budget constraint and innovation. However, the two-way causality between core explanatory variable and explained variable in the model may lead to deviations in model estimation. Hence, this paper’s robustness test includes two parts: (1) Based on Kremer’s (2009) conclusions, equation (17) is estimated using panel threshold GMM to eliminate the endogeneity of core explanatory variable and explained variable; (2) after removing control variables, the endogeneity problem of control variables, core explanatory variable and explained variable is further resolved through further estimation using panel threshold GMM model. Then, this paper estimates panel threshold GMM of equation (18). Specifically, the first step is to search the threshold value using TSLS method (instrumental variable is one-period and two-period lagged profit), and simulates the likelihood test using bootstrap method; (2) based on the searched threshold value, re-estimation is conducted with GMM method. In the specific
16 regression result, the regression coefficient of core explanatory variable remains significant in the regression result of panel threshold GMM model irrespective of whether control variables are included or not, and follows the trend of change in the previous research propositions, i.e. when profitability is low, coefficient is positive; and vice versa. This indicates that after controlling for the endogeneity of control variables, core explanatory variable and explained variable, the model’s conclusions remain robust.
4. Concluding Remarks
Recent studies (Zhang et al., 2012) stress that the backwardness of financial development and financing constraint are key reasons behind the lack of R&D spending and innovation among local firms. However, financing constraint may only explain for the lack of innovation among private firms, and cannot explain for the innovation conundrum of SOEs and the economy as a whole. Hence, this paper discusses SOE innovation efficiency and challenges facing China’s local firms in their innovation endeavors using a theoretical model of endogenous technology level. Our findings suggest that when firms engage in less profitable innovation, SOEs still generate innovations that otherwise would not happen. Therefore, there is a positive correlation between the share of SOEs and the level of innovation. When firms engage in more profitable innovation, credit discrimination and soft budget constraint will crowd out R&D funds that otherwise would go to private firms. Under the assumption of SOE inefficiency, the overall level of innovation will decrease, and the higher share of SOEs, the more substantial the reduction will be. To verify this proposition, this paper carries out an empirical analysis using panel threshold model, and the conclusion verifies our assumption. Subsequent robustness analysis also proves the reliability of this paper’s conclusions.
This conclusion provides a new approach for our understanding of China’s innovation conundrum in the present stage: Credit discrimination and soft budget constraint not only lead to SOE capital inefficiency, but also cause a financing constraint facing private firms, thus compromising the overall level of innovation. Compared with financing constraint, the underlying institutional reasons (credit
discrimination and soft budget constraint) undoubtedly have more explanatory power. Moreover, this paper also finds that when firms engage in less profitable innovation, SOE innovation still have some
17 efficiency range. Namely, credit discrimination and soft budget constraint may only be sufficient conditions for the lack of innovation in the present stage. Even with capital inefficiency, SOEs and their financing arrangements may still promote overall innovation. The policy implication is that marketbased financial reforms should be part of ownership reform, and that without significant improvement in SOEs’ capital inefficiency, financial instrument innovations and financial liberalization are not conducive to innovation among Chinese local firms; on the contrary, they will exacerbate China’s innovation conundrum in the new normal.