MASTER THESIS
INNOVATION ACTIVITIES AND THE ROLE OF MANDATORY CSR SPENDING, AN EXAMINATION OF INDIAN FIRMS
By
Lucas Lenselink
Supervisor: dr. S. Homroy Co-assessor: dr. A. Dalò
MSc International Financial Management Faculty of economics and business
University of Groningen S2705400
L.j.l.lenselink@student.rug.nl
Abstract
This research examines the influence of mandatory CSR spending, introduced by the Indian government, and the effects on innovation activities of Indian firms in the period 20102019.
Following previous studies regarding corporate taxes and the negative influence on innovation activities, this research tries to explain the negative influence of “CSR tax” on the innovation activities of Indian firms. Using a Difference-in-Difference analysis no statistical evidence is found that shows firms applying for less patents because of the regulation. On the contrary, this research finds evidence for a very small increase of R&D expenses because of the regulation, however the economic impact is very low.
Field Key Words: Innovations, Patent application, CSR regulation, R&D expenses
1 Introduction
One of the most crucial drivers of economic growth is Innovation (Solow, 1957, He and Tian 2013). Though, innovation involves substantial initial cost and a long-term perspective (Cook et al., 2019; Eccles et al., 2014). Due to the long-term perspective, high risk and unpredictable investment, innovation is frequently highest on the list managers consider sacrificing. So, on the one hand, managers are careful when deciding on innovations. On the other hand, governments face a tradeoff between austerity and future growth (Mukherjee et al., 2017). In recent years, more interest is paid to the influence of taxes on innovation activities of firms (Mukherjee et al., 2017; Atanassov and Liu, 2019; Cai et al, 2018). This will be the key focus of this research, with the distinction of mandatory Corporate Social Responsibility (CSR) expenses as a form of tax instead of corporate tax
rates.
Mukherjee et al., (2017) studied the effect of state-level corporate tax rates and proved that increase in taxes reduced future innovation. Similarly, Atanassov and Lui,
(2019) find that tax increases has negative effects on corporate innovation. Additionally, Howell (2016) find that reduction in corporate tax burdens increases new product introduction. Another example of the influence tax has on innovation is the study of Cai et al., (2018). The research confirms that a decrease in effective tax rate has a positive influence on both patent application and Research & Development (R&D) expenses. Not only effective and corporate tax rates have an influence on innovation, but also tax credit. Several studies find that after the introduction of tax credits, R&D expenses increases significantly (Czarnitzki et al, 2011; Yang et al, 2012.;
Becker, 2015). This research will contribute to that debate in a different way. I examine the
influence of compulsory CSR spending by the government on innovation activities of firms
rather than using corporate tax or tax credits as influencers of innovations.
This study uses a Difference-in-Difference approach that exploits the innovation activities among Indian firms in the period 2010-2019. The sample consist of 6,190 firms over the years 2010-2019 with a total of 48,423 firm year observations. I control for firmlevel factors (e.g.
size, leverage, profitability and ownership), firm and year fixed effects to address the heterogeneity problem.
I found no statistical evidence that firms who are eligible for regulation applied for less patents in the post regulation period compared to the pre-regulation period. To analyze whether firms who were further away from the two percent threshold were affected more, i.e. the higher the
“CSR tax” the more severe the impact on innovation, I analyzed firms who were spending below one percent of net profits to CSR and firms spending not at all to CSR. In both cases I found the same results, there was no statistical evidence that firms applied for less patents in the post regulation period. Lastly, to analyze the real effect of the regulation, I analyze the actual distance from the CSR threshold for firms in the whole sample, again I did not find statistical evidence that firms applied for less patents in the post regulation period. The results are robust for two subsamples, only innovative firms and only firms who reported CSR spending, no significant results are found when using R&D scaled to Sales and innovation efficiency (Patents/R&D) as measures for innovation.
I do find significant results when using R&D expenses as measure for innovation. For all firms eligible for regulation and the subset of firms below one percent of CSR spending to net profits, the average treated firm has slightly more R&D expenses in the post regulation period compared to the pre-regulation period. On the contrary, when measuring the actual distance for all the firms in the main sample, for a one-unit percentage increase in actual distance to the two percent CSR threshold in the post regulation period, R&D expenses decrease with 0.070 million USD.
Compared to previous research, regarding R&D expenses as measure for innovation, I found some inconsistent results. First, I would expect for all the firms eligible for regulation to have lower R&D expenses in post regulation period. This due to firms experiencing an increase in “taxes” in the form of mandatory CSR spending. However, the results show a small increase of R&D expenses in the post regulation period and therefore imply the opposite of what previous literature has shown. On the other hand, for all firms in the sample, the further firms are away of the two percent threshold, the lower the R&D expenses, which confirms previous literature. Lastly, compared to previous literature, I cannot say anything about the application of patents since I do not find any significant results. The results neither confirm nor deny the existing literature.
This research adds to the existing literature as it is, to the best of my knowledge, the first to analyze the effect of mandatory CSR spending on innovation. The results of this research are relevant especially for other governments, i.e. countries, who are considering imposing a similar form of “CSR tax”. Introducing this form of tax rather than a straightforward tax of two percent is worthwhile assuming that firm employees are better in deciding and spending to CSR activities than government employees, because firms could earn goodwill from CSR activities (Dharmapala and Khanna, 2018). Furthermore, the results show that firms who are further away from the threshold experience less R&D in the post regulation period, which is an important feature of innovation input.
Although my results do not show that firms apply for less patents, it does show a small increase in R&D expenses because of the regulation, for firms eligible for regulation. The regulation does, to a very small extent, boosts innovation and since innovation is one of the key drivers of future economic growth (Solow 1957), it is important to both firms and governments.
The remainder of this paper is structured as follows. I provide a review of the relevant literature
and develop the hypothesis in section two. In section three I discuss the methodology and data
for the research. In section four I discus the main empirical results, section five checks for the robustness of my results and section six concludes.
2. Literature and Hypothesis 2.1. Literature review
Although innovations are costly and of high risk, they still are very important drivers for the increase of firm value and future long-term economic growth (Solow, 1957; Romer, 1990; Hall et al, 2005; Kogan et al, 2017; Atanassov and Liu, 2019). The impact of taxes on a firm’s innovation is not yet well explored (Cai et al., 2018). Mukherjee et al., (2017) did examine the impact of corporate tax rates on innovation and provided evidence that taxes do not only effect patenting and R&D investments, but also new product introductions. An increase and emphasize on taxes might discourage more risky innovation projects (Mukherjee et al., 2017). Furthermore, taxes influence the capital structure of firms with firms taking on more debt following an increase of corporate income tax (Heider and
Ljungqvist, 2015) and debt is not an encouraging form of finance for innovation projects (Mukherjee et al. 2017). Innovative firms intend to save their after-tax profits and transform it to internal cash. This internal cash allows a firm to have greater flexibility and able to experiment more, which is a key motivation for innovation (Atanassov and Liu, 2019; Manso, 2011). In addition, taxes lower this internal cashflow follow an increase of taxes and since this internal cashflow is a popular source for financing innovative projects, innovation will be harmed by these taxes (Mukherjee et al., 2017).
Additionally, Cai et al., (2018) find that a decrease of the effective tax rate leads to a
substantial increase in patent applications and improved R&D expenses. Furthermore, Chen,
Liu, Serrate, and Xu (2018) find evidence for increase in R&D investments when Chinese firms
received a substantial corporate tax cut. The basis for this research comes from Hall and Reenen
(2000) who provided evidence for the influence of tax credit on R&D. The essence of their research is that one dollar in tax credit results in one dollar of additional R&D. Other studies also examined the effect of tax credit on R&D spending, these studies are done in different countries regarding different industries and all found compelling evidence for increase in R&D spending after the introduction of tax credits (Czarnitzki, et al., 2011; Yang, et al., 2012.;
Becker, 2015).
Furthermore, a reduction in Chinese private sector corporate tax burden triggered an increase in new product and process sales (Howell, 2016). Also, low access to liquidity resulted in lower innovation both in the intensity of R&D and the likelihood of pursuing R&D (Howell 2016) and financial constraints are the major barrier for firms to innovation (Nee and Opper, 2012).
Finally, a combination of the research by Jaimovich and Rebelo (2017) and Mukherjee et al., (2017) they hypothesize that firms for which taxes are imposed on net profits, innovative employees are more likely to leave. The idea behind this is that following a reduction of net profits by the increase of taxes means less money available for spending on R&D. This in turn leads to a standoff for innovative employees. They either stay at the firm and have fewer incentives to innovate or they leave the firm and take their innovative ideas to another firm which do has the capital and is more likely to encourage their innovative projects.
On the one hand, there is a wide variety of literature supporting the negative influence
of corporate tax on innovation. On the other hand, there are also studies who do not find
evidence for this relationship. Thomson (2010) examined the top 500 Australian firms and
found no evidence that tax incentives increased R&D investment, but rather see sales growth as
the primary determinant of investments in R&D. Furthermore, Krugman (2016) argues
Scandinavian countries, which are highly innovative but also have high tax rates suggesting
high taxes and high innovation can complement each other.
This paper will be different from the papers of Mukherjee et al., (2017), Atanassov and Lui, (2019) and the papers about Chinese tax reforms (Cai et al., 2018; Chen Liu, Serrate, and Xu, 2018; Howell, 2016) in that I will not use corporate income tax or any tax reforms in India, but I will use mandatory spending on CSR as a form of tax. I can use this as a tax since the definition of tax is “a compulsory charge by the government imposed on a taxpayer”. In this case the taxpayers are the firms in India. Following the previous literature and the paper of Mukherjee et al., (2017), I propose the following question:
Does mandatory CSR spending influence innovation activities?
2.2. Institutional Background and Hypothesis
India is a first mover when it comes to introducing regulation requiring firms to spend a percentage of their profits on socially responsible activities (Dharmapala and Khanna, 2018).
This rule is part of section 135 of the firms act 2013 and consists of two parts. First, the section specifies which firms are affected by the regulation and second, the obligation firms have when affected by the regulation. Firms affected by the regulation are firms who surpass one of the following three thresholds in any fiscal year (i) having net worth (the face value of shares originally issued adjusted for retained earnings and various reserves) exceeding INR 5 billion, which is around USD 70 million, (ii) a turnover exceeding INR 10 billion, USD 140 million or (iii) net profits exceeding INR 50 million, USD 700 thousand. Furthermore, any firm with operations in India (foreign or domestic) crossing one of these thresholds is subject to section 135 (Dharmapala and Khanna, 2018). The following actions are required when crossing one of the thresholds: (i) it must constitute a Corporate Social Responsibility Committee consisting of at least three directors, out of which at least one must be independent
1, (ii) needs to disclose the
1 Section 135, sub section 1 (MCA, 2018)
composition of the CSR Committee
2, (iii) the CSR committee must formulate a CSR policy which shall indicate activities to be undertaken by the firm, recommend amount of expenses on the activities and monitor the CSR policy
3, (iv) the board is to approve the CSR policy and disclose contents of the CSR policy in its annual report
4, (v) the board must ensure the firm spends at least two percent of average net profits of last three years on activities listed in the CSR policy, or specify reasons when the firm fails to spend the amount
5. When failing to meet the requirements and no reason is provided why, both the firm and every officer of the firm is penalized with a fine of INR 10,000, 141 USD for the first day of violation and additionally any proceeding day a fine of IND 1,000, 14 USD if the violation continues (Dharmapala & Khanna, 2018). The list of activities which satisfy the requirements of CSR spending is very broad (e.g.
spending on education, health, environment gender equality, designated government programs) and therefore leave considerable directions for firms to designate the mandatory CSR spending (Dharmapala and Khanna, 2018).
There are several effects of this mandatory spending on CSR activities for firms.
Manchiraju and Rajgopal (2017) examined the effect of the law on shareholder value and found that the law caused a drop-in stock prices for firms who needed to spend money on
CSR. Secondly, Dharmapala and Khanna, (2018) investigated the impact of the law for Indian firms. They found that firms, who were required by the law and who did spend less than two percent of net profits to CSR, increased their spending to two percent. But, firms that spend more than two percent of net profits to CSR reduced their spending to two percent since they were only required to spend two percent. The introduction of the regulation introduces a unique setting in which we can examine the effect of the law on a treated and a control group. I form
2 Section 135, sub section 2 (MCA, 2018) 3
Section 135, sub section 3 (MCA, 2018) 4 Section 135, sub section 4 (MCA, 2018) 5 Section 135, sub section 5 (MCA, 2018)
the following hypothesis based on the literature discussed in the previous part of the paper.
Firms who spend less than two percent and must increase their spending, so an increase of “tax”, will experience a drop in innovation activities.
Hypothesis 1: an increase in CSR expenses due to the regulation will reduce innovation activities
3. Methodology and Data 3.1. Research method
To examine both hypotheses, I will perform firm level Difference-in-Difference (DiD) analysis, over the period 2010-2019, to see what the effect is from the treated group (the group who are required to spend two percent on CSR) to a control group regarding there innovation activities. This method is used extensively by prior studies all concerned with the influence of an exogeneous shock (Dharmapala and Khanna, 2018; Chen, Hung, and Wang, 2018; Krüger 2015; Mukherjee et al., 2017). First, I will analyze the main result for firms affected by regulation. Secondly, I will distinct the firms affected by regulation in two categories: firms, in pre-regulation period, spending less than one percent of net profits to CSR activities and firms spending zero percent of net profits to CSR activities. Lastly, I will analyze the real effect of the regulation by using the actual distance from the two percent threshold for firms.
3.2. Variables
The data used in this research comes from two different sources. The patent data I retrieve from the Controller General of Patents, Designs & Internal Trade marks database also called CGPDT. This database is the general place where information about patents applied in India is stored. The data for financial variables is retrieved from the Prowess
dxdatabase. This database contains financial information for over 40 thousand Indian firms, either privately, listed or unlisted.
3.2.1 Innovation
There are several ways to measure innovation. First, a common measure of innovation following Cook et al. (2019), Mukherjee et al., (2017) who use (i) number of patents held (ii) number of new patents generated by the firm during the year and (iii) the number of citations each patent has as of the end of the year. Secondly, other measures for innovation are R&D expenses or R&D investments (Cai et al., 2018; Mukherjee et al., 2017; Howell, 2016). I intend to follow the article of Helmers et al, (2017), who use three firm specific forms of innovation.
First, they use patent count like Mukherjee et al., (2017); Cook et al., (2018) this will be my main indicator for innovation. Secondly, they use R&D expenditure like (Cai et al., 2018) and lastly, they use the ratio of patent count to R&D, which is a measure for innovation efficiency (Hirshleifer et al, 2013; Helmers et al., 2017)). I will use R&D expenses and innovation efficiency as a robustness checks for my main
results.
I retrieve the data for patents from the online web page iPairs CGPDT which contains
all patent data in India filled from 2005 onwards. I rely on a sample period from April 2010 till
April 2019 and with the help of a scrapping tool (scrapping code) patents will be matched to
firms. All the data available from 2010-2019 with applicant country India is retrieved, which
comes down to almost 70 thousand patents, the distribution of these patents is reported in table 1.
From all the patents applied for, 26,631 are filled by firms. These patents are matched with the firms in the main database and counted for every firm in every year. This resulted in 12,337 patents matched to firms in the primary data base, which contains all the financial variables, ranging from 0 patent applications in one year to 209 patent applications in one year.
Table 1 Distribution of patents
Type of applicant # of patents % of total
Firms 26,631 38.3%
Colleges 1,013 1.5%
Universities 2,771 4.0%
Institutes 4,070 5.8%
Individuals
Only name of the patent, no further
30,872 44,4%
information 4,239 6.1%
Total 69,596 100.0%
Notes: Table 1 contains all the patents applied for during the period April 2010-April 2019. I divided the applications in five broad categories, with an additional sixth category for patents without any information.
3.2.2 CSR variable
To indicate whether firms are affected by regulation, a dummy variable is created (Comppre2013) which presents firms who were surpassing the threshold of turnover of USD 140 million or net profit of USD 700 thousand. I left out the net worth restriction since I could not find out which reserves are specified for the restriction. From those firms, two separate variables are created which indicate if a firm has below one percent spending of net profits to CSR before the regulation or no spending at all to CSR activities. Furthermore, I add a continuous variable which measures the actual distance of firms from the two percent threshold.
This variable is constructed as 0.02 (the threshold of CSR spending) minus the total CSR of a
firm in a year by the total net profits of a firm in a year.
3.2.3 Control variables
To address the omitted variable problem, my DiD analysis will control for timevarying firm-level variables, used in prior research which effect innovation. For firm characteristics is controlled for firm size as the natural logarithm of total assets (SIZE) (Kim et al, 2018), leverage as total debt/total assets (LEVERAGE), profitability as EBITDA/total assets (PROFITABILITY) however I will use net profit instead of EBITDA (Mukherjee et al., 2017;
Atanassov and Liu, 2019). Ownership measured in two different ways, percentage of shares held by promoters (FAMILY_OWNERS) and percentage of shares held by institutional investors (INSTITUTIONAL_OWNERS). Lastly, I winsorized all financial variables at .01 and .99 as suggested by previous research to avoid outliers who might affect the results (Gonenc and Scholtens, 2017).
3.3 Descriptive statistics and correlation
Table 2 represents the descriptive statistics for the main analysis. The number of patents applied for in one-year ranges from 0 to 209, with a mean of .156. This indicates that a lot of firms did not apply for patents in each of the years in the sample. The firms eligible for regulation in the pre-regulation period is 51.6 percent. So, a little more than half of the firms in the sample are affected by the regulation. From all the firms in the sample 49.9 percent had CSR spending lower than one percent of net profits in pre-regulation period and are eligible for regulation.
Next, of the firms eligible for regulation in the sample, 26.7 percent did not have CSR
spending in pre-regulation period. Furthermore, the average distance for firms in the sample to
the two percent threshold is two percent. This is for all the firm year observations in the analysis,
when I look at all the firm year observations in the sample after the regulation, the average
remains two percent distance from the threshold.
Table 2 Descriptive Statistics
Variable Obs. Mean Std.Dev. Min Max
# Patents 48,423 0.156 2.997 0 209
Comppre2013 48,423 0.516 0.5 0 1
Company 1% 48,423 0.499 0.5 0 1
Company 0% 48,423 0.267 0.443 0 1
CSR distance 45,938 0.02 0.001 0.006 0.021
Size 48,423 2.476 2.371 -4.51 9.272
Leverage 48,423 0.462 1.496 0 30.092
Profitability 48,423 0.073 1.119 -10.355 3.798
Institutional owners 48,423 3.955 8.109 0 44.77
Family owners 48,423 44.677 25.584 0 100
R&D expenses 48,402 0.492 5.085 0 166.923
CSR total expenses 48,423 0.094 1.492 -0.112 276.616
Notes: the table presents the descriptive statistics for the main analysis. # Patents is the total number of patents applied per firm in each year of the sample. Comppre2013 are the firms who pass the thresholds and thus are eligible for regulation in the pre-regulation period. Company 1% are the firms eligible for regulation but spending less than 1% of net profits to CSR in pre-regulation period. Firms 0% are the firms eligible for regulation, but not at all spending to CSR activities in pre-regulation period. CSR distance is the actual distance of a firm from the two percent CSR threshold. Size is the log of total assets. Leverage is the total debt/total assets. Profitability is the net profit/total assets. Family owners is the percentage of shares held by promoters. Institutional owners is the percentage of shares held by institutions. R&D expenses are the expenses to R&D activities per year for each firm.
CSR total is the total CSR per year for each firm.
Table 3A Pearson Correlation Matrix
Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
(1) # Patents 1.000
(2) Comppre2013 0.045 1.000
(3) Company 1% 0.046 0.967 1.000
(4) Company 0% 0.013 0.596 0.617 1.000
(5) Size 0.099 0.579 0.563 0.196 1.000
(6) Leverage -0.010 -0.102 -0.100 -0.044 -0.120 1.000
(7) Profitability 0.051 0.463 0.454 0.261 0.199 -0.188 1.000
(8) Institutional owners 0.071 0.196 0.196 0.017 0.370 0.017 0.074 1.000
(9) Family owners 0.024 0.181 0.174 0.057 0.205 0.007 0.057 0.108 1.000
(10) Total CSR 0.131 0.059 0.054 -0.033 0.126 -0.013 0.049 0.112 0.027 1.000
Notes: table 3A present the correlation matrix. Where # Patents is the total number of patents applied per firm in
each year of the sample. Comppre2013 are the firms who pass the thresholds and thus are eligible for regulation.
Company 1% are the firms eligible for regulation but spending less than 1% of net profits to CSR in pre-regulation period. Firms 0% are the firms eligible for regulation, but not at all spending to CSR activities in pre-regulation period. Size is the log of total assets. Leverage is the total debt/total assets. Profitability is the net profit/total assets.
Family owners is the percentage of shares held by promoters. Institutional owners is the percentage of shares held by institutions and Total CSR is the total CSR for a company in each year.
Table 3B frequency table CSR/Net profits
% CSR of Net profits 0% or lower
# Observations 37,585
Between 0%-1% 8,556 17.67
Between 1%-2% 286 0.59
Above 2% 1,996 4.12
Total 48,423 100
Notes: table 3B presents the frequency table. The table presents all the firm year observations as percentage of net profits and the frequency per category of total observations. It is important to note that this contains all the observations used in the main analysis and not only the firms eligible for regulation.
This is because a lot of firms do not have any CSR reported in the sample period, which is reported in the frequency table 3B. Furthermore, the total average total CSR spend during the year is .094 million USD. Lastly, the average R&D expenses during a year are .492 million USD, ranging from 0 to 166.9 million USD.
Table 3A presents the correlation matrix. The correlation of the variables most used in the literature regarding the number of patents is low, except for Size, which indicates that they are relevant for my DiD model and will explain a part of the variation in number of patents applied for each year. The correlation between Comppre2013 and Size is logical since the dummy only represents the bigger firms in the sample, which surpass the thresholds of the law.
Furthermore, the correlation between Comppre2013 and both Company 1% and Company 0%
and the correlation of Company 1% and Company 0% is explained by the frequency table. The frequency table shows that most of the observations are in the category’s zero percent or lower and between zero and one percent.
Percent
7 7. 62
4. Main empirical findings
4.1.1 Empirical findings for all firms eligible for regulation
I start my analysis with the main DiD regression, following the same method as Mukherjee et al., (2017). All firms eligible for regulation are included as the variable comppre2013. Control variables, year and firm fixed effects (𝛾
𝑡,𝑖) are added as well to address the omitted variable problem. The main explanatory variable will be the natural logarithm of patents applied for each firm (i) in each year(t), ln(1+#patents). Since the regulation is only present in India and I examine only firms in India, I do not have to control for country fixed effects. Standard error estimates are clustered by each firm to deal with correlation amongst firms. The regression equation is as follows:
Ln(1 + 𝑃𝑎𝑡𝑒𝑛𝑡𝑠)
𝑡,𝑖= 𝛽
0+ 𝛽
1𝑐𝑠𝑟𝑟𝑒𝑔𝑢𝑙𝑎𝑡𝑖𝑜𝑛 + 𝛽
2𝑐𝑜𝑚𝑝𝑝𝑟𝑒2013
+ 𝛽
3𝑐𝑠𝑟𝑟𝑒𝑔𝑢𝑙𝑎𝑡𝑖𝑜𝑛 × 𝑐𝑜𝑚𝑝𝑝𝑟𝑒2013 + 𝛽
4𝑆𝑖𝑧𝑒
𝑡,𝑖+ 𝛽
5𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒
𝑡,𝑖+ 𝛽
6𝑃𝑟𝑜𝑓𝑖𝑡𝑎𝑏𝑎𝑖𝑙𝑖𝑡𝑦
𝑡,𝑖+ 𝛽
7𝐼𝑛𝑠𝑡𝑖𝑡𝑢𝑡𝑖𝑜𝑛𝑎𝑙 𝑜𝑤𝑛𝑒𝑟𝑠
𝑡,𝑖+ 𝛽
8𝐹𝑎𝑚𝑖𝑙𝑦 𝑜𝑤𝑛𝑒𝑟𝑠
𝑡,𝑖+ 𝛾
𝑡,𝑖+ 𝜀
𝑡,𝑖(1)
Results of the first DiD are reported in table 4 column 1, where all firms affected by regulation are included. I find no evidence that following the regulation, firms applied less patents then before regulation. The coefficient is statistically insignificant; therefore, I cannot infer anything from the regression results.
4.2 Firms spending one percent or less to CSR activities
To more thoroughly analyze the impact of the regulation on innovation, I perform a
second DiD, which follows the same methodology as the previous analysis. The exception for
this DiD is that only firms are included which spend less than one percent to CSR activities before regulation. I do this analysis to examine whether the impact of the “CSR tax” is higher when firms must adjust more, i.e. they must compensate more because there
CSR spending was below 1 percent of net profits before the regulation was introduced. This DiD analysis has the following regression equation:
Table 4 DiD with ln(1+patents) as measure for innovation
(1) (2) (3) (4)
VARIABLES Ln(1+
#Patents)
Ln(1+
#Patents)
Ln(1+
#Patents)
Ln(1+
#Patents)
Csrregulation*Comppre2013
-0.001 (0.004)
Csrregulation*Company 1%
-0.002 (0.004)
Csrregulation*Company 0%
0.001 (0.004)
Csrregulation*CSR distance
-0.084 (2.004)
Csrregulation
-0.003***
-0.002**
-0.003
-0.001
(0.001) (0.001) (0.002) (0.040)
Comppre2013
0.002 (0.003)
Company 1%
0.002 (0.003)
Company 0%
0.001 (0.004)
CSR distance
1.466 (1.319)
Size -0.001 -0.001 -0.001 -0.001
(0.001) (0.001) (0.001) (0.001)
Leverage -5.56e-05 -5.59e-05 -5.46e-05 -3.14e-05
(0.000) (0.000) (0.000) (0.000)
Profitability 0.002** 0.002** 0.002** 0.002**
(0.001) (0.001) (0.000) (0.001)
Institutional Owners -0.000 -0.000 -0.000 -0.000
(0.000) (0.000) (0.000) (0.000)
Family Owners 1.52e-05 1.56e-05 1.53e-05 1.28e-05
(3.74e-05) (3.74e-05) (3.72e-05) (3.86e-05)
Constant 0.034*** 0.034*** 0.035*** 0.006
(0.003) (0.00336) (0.00362) (0.027)
Firm fixed effects YES YES YES YES
Observations 48,423 48,423 48,423 45,938
R-squared 0.000 0.000 0.000 0.001
Number of Firms 6,190 6,190 6,190 6,144
Notes: table 4 reports the regression results of the Difference-in-Difference approach. Where ln(1+#patents) is the measure of innovation, the natural logarithm of 1+# Patents per year per firm. Csrregulation is the indicator for the start of the regulation. Comppre2013 are the firms who pass the thresholds and thus are eligible for regulation.
Company 1% are the firms eligible for regulation but spending less than 1% of net profits to CSR in pre-regulation period. Firms 0% are the firms eligible for regulation, but not at all spending to CSR activities in pre-regulation period. CSR distance is the actual distance of a firm from the two percent CSR threshold. Size is the log of total assets. Leverage is the total debt/total assets. Profitability is the net profit/total assets. Family owners is the percentage of shares held by promoters. Institutional owners is the percentage of shares held by institutions. Firm and year fixed effects are included, Robust standard errors in parentheses . *** p<0.01, ** p<0.05, * p<0.
Ln(1 + 𝑃𝑎𝑡𝑒𝑛𝑡𝑠)
𝑡,𝑖= 𝛽
0+ 𝛽
1𝑐𝑠𝑟𝑟𝑒𝑔𝑢𝑙𝑎𝑡𝑖𝑜𝑛 + 𝛽
2𝐶𝑜𝑚𝑝𝑎𝑛𝑖𝑒𝑠 1%
+ 𝛽
3𝑐𝑠𝑟𝑟𝑒𝑔𝑢𝑙𝑎𝑡𝑖𝑜𝑛 × 𝐶𝑜𝑚𝑝𝑎𝑛𝑖𝑒𝑠 1% + 𝛽
4𝑆𝑖𝑧𝑒
𝑡,𝑖+ 𝛽
5𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒
𝑡,𝑖+ 𝛽
6𝑃𝑟𝑜𝑓𝑖𝑡𝑎𝑏𝑎𝑖𝑙𝑖𝑡𝑦
𝑡,𝑖+ 𝛽
7𝐼𝑛𝑠𝑡𝑖𝑡𝑢𝑡𝑖𝑜𝑛𝑎𝑙 𝑜𝑤𝑛𝑒𝑟𝑠
𝑡,𝑖+ 𝛽
8𝐹𝑎𝑚𝑖𝑙𝑦 𝑜𝑤𝑛𝑒𝑟𝑠
𝑡,𝑖+ 𝛾
𝑡,𝑖+ 𝜀
𝑡,𝑖(2)
Where Firms 1% represents the treated firms. Results are reported in table 4 column 2.
Consistent with the previous results, the coefficient of interaction is statistically insignificant, so I cannot infer anything about the regression results.
4.3 Firms who did not spend at all to CSR activities
Third, for the main empirical findings, I look at firms who did not spend at all to CSR activities in the pre-regulation period. Again, following the same methodology as before. For these firms, I expect to largest difference between treated and control group, since these firms must adjust more and has the following regression equation:
Ln(1 + 𝑃𝑎𝑡𝑒𝑛𝑡𝑠)
𝑡,𝑖= 𝛽
0+ 𝛽
1𝑐𝑠𝑟𝑟𝑒𝑔𝑢𝑙𝑎𝑡𝑖𝑜𝑛 + 𝛽
2𝐶𝑜𝑚𝑝𝑎𝑛𝑖𝑒𝑠 0%
+ 𝛽
3𝑐𝑠𝑟𝑟𝑒𝑔𝑢𝑙𝑎𝑡𝑖𝑜𝑛 × 𝐶𝑜𝑚𝑝𝑎𝑛𝑖𝑒𝑠 0% + 𝛽
4𝑆𝑖𝑧𝑒
𝑡,𝑖+ 𝛽
5𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒
𝑡,𝑖+ 𝛽
6𝑃𝑟𝑜𝑓𝑖𝑡𝑎𝑏𝑎𝑖𝑙𝑖𝑡𝑦
𝑡,𝑖+ 𝛽
7𝐼𝑛𝑠𝑡𝑖𝑡𝑢𝑡𝑖𝑜𝑛𝑎𝑙 𝑜𝑤𝑛𝑒𝑟𝑠
𝑡,𝑖+ 𝛽
8𝐹𝑎𝑚𝑖𝑙𝑦 𝑜𝑤𝑛𝑒𝑟𝑠
𝑡,𝑖+ 𝛾
𝑡,𝑖+ 𝜀
𝑡,𝑖(3)
Where Firms 0% represents the treated firms. The results are reported in table 4 column 3. I find no evidence that firms who did not spend on CSR activities before the regulation are filing lower patents after the regulation, while they are required to spend the biggest amount of net profits to CSR activities.
4.4 Continuous measure from two percent CSR
Lastly, to analyze the real effect of the introduction of regulation, I use a continuous measure which measures the actual distance of firms CSR spending regarding the two percent rule. I follow the same methodology as in the previous analysis. The equation for this model is a follow:
Ln(1 + 𝑃𝑎𝑡𝑒𝑛𝑡𝑠)
𝑡,𝑖= 𝛽
0+ 𝛽
1𝑐𝑠𝑟𝑟𝑒𝑔𝑢𝑙𝑎𝑡𝑖𝑜𝑛 + 𝛽
2𝐶𝑆𝑅 𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑒
𝑡,𝑖+ 𝛽
3𝑐𝑠𝑟𝑟𝑒𝑔𝑢𝑙𝑎𝑡𝑖𝑜𝑛 × 𝐶𝑆𝑅 𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑒
𝑡,𝑖+ 𝛽
4𝑆𝑖𝑧𝑒
𝑡,𝑖+ 𝛽
5𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒
𝑡,𝑖+ 𝛽
6𝑃𝑟𝑜𝑓𝑖𝑡𝑎𝑏𝑎𝑖𝑙𝑖𝑡𝑦
𝑡,𝑖+ 𝛽
7𝐼𝑛𝑠𝑡𝑖𝑡𝑢𝑡𝑖𝑜𝑛𝑎𝑙 𝑜𝑤𝑛𝑒𝑟𝑠
𝑡,𝑖+ 𝛽
8𝐹𝑎𝑚𝑖𝑙𝑦 𝑜𝑤𝑛𝑒𝑟𝑠
𝑡,𝑖+ 𝛾
𝑡,𝑖+ 𝜀
𝑡,𝑖(4)
Where CSR distance represents the actual distance, a firm has from the two percent threshold. The results are reported in table 4 column 4, the coefficient of the interaction is statistically insignificant. Therefore, I cannot infer anything from the regression results.
5. Robustness checks
For the robustness of my results, I perform additional analysis to test the impact of
“CSR tax” on the innovation of firms. First, considering different measures of innovation.
Following the articles of Helmers, Patnam and Rau (2017); Cai et al., (2018); Mukherjee et al., (2017) other measure of innovation, besides number of patents applied for, are R&D expenses.
I will use two different measures of R&D expenses, the ln(1+R&D) expenses and the ln(R&D/Sales) (Mukherjee et al., 2017). Additionally, I will use innovation efficiency as proxy for innovation following the article of Hirshleifer, et al., (2013). The measure for innovation efficiency is patents scaled by R&D expenses, ln(1+Patents/R&D).
5.1.1 Innovation measure as ln(1+R&D), for all firms eligible for regulation
Following the same methodology as the main empirical findings and thus the same regression equation, table 5 columns 1 reports the results for all the firms eligible for regulation.
Columns 1 shows a significant and positive results. These results imply that, for all firms eligible for regulation, following the regulation experience an increase in R&D expenses. This, in turn, suggest the firms spend more on innovation. This is in contrast with the main results on application of patents and the literature. To determine the economic significance, I follow the articles of Mukherjee et al. (2017) and Atanassov and Lui (2019), where the latter also takes the distribution into account (Edelen and Warner, 2001; Becker, 2006). For the average treated firm, R&D expenses increased with 0.019
3million USD. To assess whether this change is small or big depends on the distribution (Atanassov and Lui, 2019). For the average treated firm, R&D expenses increased with 0.019
7million USD. To assess whether this change is small or big depends on the distribution (Atanassov and Lui,
3 Post regulation is followed by an increase of 0.038 times 0.492 mean R&D expenses = 0.019 R&D expenses 7 Post regulation is followed by an increase of 0.038 times 0.492 mean R&D expenses = 0.019 R&D expenses 8 Increase in R&D expenses 0.019 divided by SD R&D 5.085 = .37%
2019). Therefore, I compare this change in R&D expenses to the standard deviation of the R&D expenses (5.085). This increase in R&D expenses is approximately 0.37
8percent of the standard deviation above the mean. So, the variation attributable to the introduction of the regulation represents .37 percent of the variation in innovation, when using R&D expenses as measure for innovation, in the post regulation period, which is an extremely small part. Although the economic significance is low, an explanation for these results could be the fact that the law does not specify CSR, it rather gives a possible list of CSR activities, which could be innovative activities (Manchiraju and Rajgopal, 2017).
Table 5 DiD with ln(1+R&D) as measure for innovation
(1) (2) (3) (4)
VARIABLES Ln(1+R&D) Ln(1+R&D) Ln(1+R&D) Ln(1+R&D)
Csrregulation*Comppre2013 0.038***
(0.005)
Csrregulation*Company 1% 0.035***
(0.006)
Csrregulation*Company 0% -0.006
(0.00557)
Csrregulation*CSR distance -14.30***
(2.753)
Csrregulation -0.005*** -0.00274 0.0149*** 0.296***
(0.002) (0.00184) (0.00298) (0.0552)
Comppre2013 -0.016***
(0.004)
Company 1% -0.015***
(0.004)
Company 0% -0.010*
(0.005)
CSR distance 1.205
(1.524)
Size 0.015*** 0.015*** 0.017*** 0.017***
(0.003) (0.003) (0.003) (0.003)
Leverage 0.009*** 0.009*** 0.009*** 0.009***
(0.002) (0.002) (0.002) (0.002)
Profitability 0.003** 0.003** 0.005*** 0.004**
(0.001) (0.001) (0.001) (0.001)
Institutional Owners 0.003*** 0.003*** 0.003*** 0.003***
(0.000) (0.000) (0.000) (0.000)
Family Owners 2.96e-05 3.26e-05 5.24e-05 2.59e-05
(7.06e-05) (7.04e-05) (7.06e-05) (7.33e-05)
Constant 0.045*** 0.043*** 0.033*** 0.008
(0.007) (0.008) (0.008) (0.031)
Firm fixed effects YES YES YES YES
Observations 48,402 48,402 48,402 45,919
R-squared 0.021 0.020 0.019 0.022
Number of Firms 6,189 6,189 6,189 6,143
Notes: table 5 reports the regression results of the Difference-in-Difference approach. Where ln(1 +R&D) is the measure of innovation, the natural logarithm of 1 + R&D expenses for each firm in each year. Csrregulation is the indicator for the start of the regulation. Comppre2013 are the firms who pass the thresholds and thus are eligible for regulation. Company 1% are the firms eligible for regulation but spending less than 1% of net profits to CSR in pre-regulation period. Firms 0% are the firms eligible for regulation, but not at all spending to CSR activities in pre-regulation period. CSR distance is the actual distance of a firm from the two percent CSR threshold. Size is the log of total assets. Leverage is the total debt/total assets. Profitability is the net profit/total assets. Family owners is the percentage of shares held by promoters. Institutional owners is the percentage of shares held by institutions.
Firm and year fixed effects are included, Robust standard errors in parentheses . *** p<0.01, ** p<0.05, * p<0.
Therefore, it is possible that firms report existing innovative activities as CSR activities, while in the past they did not report those activities as CSR. So, it is possible firms spend more to those activities because they are now required to by the regulation and therefore R&D expenses will increase. Another considerable explanation for more R&D expenses after the introduction of the regulation is the important difference between corporate tax rates and the
“CSR tax” from the regulation. Instead of collecting the “CSR tax” similar to corporate taxes and spend these taxes as a government to CSR projects, the government lets the firms choose to which CSR activities they spend their “CSR tax”
(Dharmapala and Khanna 2018). This, assuming firm employees do a better job at selecting
CSR activities compared to government employees because, for example, firms can earn
goodwill of effective CSR activities. Also, especially in emerging countries, the risk for
government employees spending to for them more favorably activities and thus misusing the
funds is greater than that of firm employees (Dharmapala and Khanna 2018). Because firms
can earn goodwill from CSR activities, therefore have incentives to innovate and as a result R&D expenses could increase.
5.1.2 Innovation measure as ln(1+R&D), for firms spending one percent or less
Following the same methodology of the main empirical results, table 5 column 2 reports the results for all firms eligible for regulation and spending less than one percent to CSR activities prior to regulation. The coefficient of interaction is statistically significant which implies that firms eligible for regulation with one percent or less CSR have more R&D expenses in post regulation period. The economic significance of these firms is slightly less than all firms eligible for regulation in the sample. Following the same methodology as in Section 5.1.1, for the average treated firm R&D expenses increased with
0.017
4million USD. This increase in R&D expenses is approximately 0.33
10percent of the standard deviation above the mean. So, the variation attributable to the introduction of the regulation represents .33 percent of the variation in innovation, when using R&D expenses as measure for innovation, in the post regulation period, which is again an extremely small part.
Additionally, to the explanations mentioned in Section 5.1.1, another explanation for more R&D expenses could be that these firms must invest a substantial amount in CSR activities and therefore there CSR performance regarding the pre-regulation period significantly increases.
When the CSR performance is associated with innovative CSR activities, then the CSR performance has a positive influence on R&D expenses (Cook et al., 2019). However, the economic significance is still extremely low.
4 Post regulation is followed by an increase of 0.035 times 0.492 mean R&D expenses = 0.017 R&D expenses 10 Increase in R&D expenses 0.017 divided by SD R&D 5.085 = .33%
5.1.3 Innovation measure as ln(1+R&D), for firms spending not at all to CSR
Following the same methodology of the main empirical results, table 5 column 3 reports the results for all firms eligible for regulation and spending not at all to CSR activities prior to regulation. The coefficient of interaction is statistically insignificant, which is consistent with the main empirical findings. Since the coefficient of interaction is statistically insignificant I cannot infer anything from the regression results.
5.1.4 Innovation measure as ln(1+R&D), for distance from CSR threshold
Following the same methodology of the main empirical results, table 5 column 4 reports the results for the actual distance of firms to the two percent CSR threshold. The coefficient of interaction is statistically significant, which implies for firms who are further away from the two percent CSR threshold in the post regulation period have less R&D expenses. More specifically, in the post regulation period, for a one-unit percentage increase
5in actual distance to CSR threshold, R&D expenses decrease with 0.070
6million USD. This is 1.38 percent of the variation in standard deviation below the mean.
An explanation for the different result between R&D expenses and patent filing is the fact that the application process for patents is a rather long process, while R&D expenses could be adjusted more on the short term. Also, R&D expenses considered to be the innovation input, where patents considered to be the innovation output (Chang et al., 2019). Therefore, these results are as expected since it is likely that the impact of the regulation on patents takes longer to show an effect, where the impact on R&D is directly shown because of the short-term adjustable nature of R&D expenses. This result is in line with the previous literature. The decrease in R&D expenses has important implications for firms in my sample, since
5 A one-unit percentage increase in this case is an increase of 0.01 in CSR distance 6 Post regulation is followed by a decrease of 14.30% times 0.492 = 0.070 R&D expenses
approximately 95 percent of the sample observations have less than one percent total CSR of net profits.
5.2 Innovation measure as ln(R&D/Sales)
Following the article of Mukherjee et al., (2017) innovation can also be measured as the R&D to sales ratio or ln(R&D/Sales). I use ln(R&D/Sales) to ensure that changes in R&D are driving the results and not changes in sales (Mukherjee et al., 2017). I follow the same methodology as the main empirical findings, I report the results in table 6. For all the four different regressions the coefficients of interaction are statistically insignificant which means I cannot infer anything from the results.
5.3 Innovation measure as innovation efficiency ln(1+Patent/R&D)
Following the article of Hirshleifer et al., (2013) and Helmers et al., (2017), another measure for innovation is the efficiency of the innovation. The idea behind this measure is that it captures the ability to generate patents per dollar of R&D investment and connects innovation output (i.e. patents) with innovation input (i.e. R&D expenses) (Chang et al., 2019). I follow the same methodology as the main empirical results. The results are reported in table 7. For all the four regressions the coefficients of interaction are statistically insignificant, which means I cannot infer anything from the results.
5.4. Only innovative firms and only CSR firms
5.4.1 Subsamples with ln(1+ # Patents) as innovation measure
I perform two different subsample analysis to test for robustness. First, I use only
innovative firms. I consider a firm to be innovative if the firm filled for at least one patent in at
least one year of the sample (Atanassov and Lui, 2019), also because my main measure of innovation is patent application. The results are reported in table 8. Only 298 of the total 6,190 firms in the main analysis filled for patents during the sample period. The coefficient of interaction in column 1 for Comppre2013 is statistically significant and positive. For only innovative firms and eligible for regulation, the average treated firm has 0.033
7more patents filled in the post regulation period. However, the economic significance is small since the 0.033 increase in patents applied is approximately 1.1
14percent of the standard deviation above the mean. So, when using patent application as measure for innovation, the variation attributable to the introduction of the regulation represents 1.1 percent of the variation in innovation, in the post regulation period.
Table 6 DiD with ln(R&D/Sales) as measure for innovation
(1) (2) (3) (4)
VARIABLES Ln(R&D/
Sales)
Ln(R&D/
Sales)
Ln(R&D/
Sales)
Ln(R&D/
Sales)
Csrregulation*Comppre2013
0.469 (0.393)
Csrregulation*Company 1%
0.442 (0.371)
Csrregulation*Company 0%
0.257 (0.181)
Csrregulation*CSR distance
-50.08 (51.99)
Csrregulation
-0.550
-0.526
-0.363
0.700
(0.471) (0.451) (0.307) (0.778)
Comppre2013
-0.383 (0.348)
Company 1%
-0.361 (0.336)
Company 0%
-0.251 (0.223)
7 Post regulation is followed by an increase of .212 times .156 = .033 more patents 14 Increase in patents of .033 divided by SD of # patents (2.997) = 1.1%
