## Abstract

One of the most rigorous methodologies in the corporate governance literature uses firms' reactions to industry shocks to characterize the quality of governance. This methodology can produce the wrong answer unless one considers the ways firms compete. Because macro-level shocks reverberate differently at the firm level depending on whether a firm has a cost structure that requires significant adjustment, the quality of governance can only be elucidated accurately analyzing a firm's business strategy and their corporate governance. These differences can help one determine whether the fruits of a positive macro-level shock have been expropriated by insiders. Using the example of Indian firms, we show that an influential finding is reversed when these differences are considered. We further argue that the conventional wisdom about tunneling and business groups will need to be reformulated in light of the data, methodology, and findings presented here.

In emerging economies, business groups are responsible for the vast majority of sales, assets, and value added. A business group is in essence a leverage device: firms within the group band together to fund investments and start-ups and to share production, R&D, and marketing knowledge. The group also enables a single entrepreneur to control vast knowledge-creating resources with a fraction of the capital that would be needed by a stand-alone entity. Such groups thus make possible complex recombinations of inputs beyond the reach of a stand-alone firm. But are business groups in emerging economies actually being used to accomplish such a value-creating, sophisticated recombination of inputs? Or are their owner-managers pilfering the invested resources of minority investors by exploiting their control over group affiliates and developing countries' weak legal institutions (Johnson et al. 2000; Siegel 2005)?

Deep empirical and theoretical insights are needed to determine whether business groups serve primarily as theft devices (as proposed by Bertrand, Mehta, and Mullainathan's influential 2002 article) or as an agglomeration technology that enables scale-and-scope efficiencies (the dominant view in the business strategy literature). This is a multi-trillion-dollar question, since its answer may influence how legislators, regulators, and law-enforcement officials in emerging economies prioritize their institutional efforts to ensure high-quality corporate governance. Thus, the development of rigorous methodology to study corporate governance is not merely an academic issue; it has enormous real-world consequences.

The most influential methodology in the corporate governance literature is that of Bertrand, Mehta, and Mullainathan (2002). By examining changes in the profitability of a firm's industry peers, they argue, one can measure whether the firm was reacting to sudden changes in industry profitability and then use that measure to assess the quality of the firm's governance. This insight has power and simplicity because it uses an external reference point to evaluate governance. The methodology also allows researchers to use firm-level fixed effects and thus to control for time-invariant firm heterogeneity. Bertrand, Mehta, and Mullainathan's (2002) study has been hailed for cementing the conventional wisdom about tunneling and business groups (see Morck, Wolfenzon, and Yeung 2005; Bebchuk and Weisbach 2010), and cited over 90 times in 2002–2010 according to the Social Science Citation Index.

But Bertrand, Mehta, and Mullainathan's (2002) methodology overlooks the crucial role of business strategy in corporate governance. The choice of business strategy determines corporate governance in logical ways, and only by directly analyzing business strategy can the quality of corporate governance be effectively assessed. Our understanding of corporate governance will be misleadingly incomplete if we do not take into account the different ways that firms in a given industry compete. Firms make different choices about the degree to which they recombine inputs purchased elsewhere to create something more valuable. Some firms are little more than resellers of touched-up inputs; at the other end of the spectrum, firms recombine inputs in complex, knowledge-based ways to add functionality and increase consumers' willingness to pay. The decision to be a reseller or a recombiner is a core choice in applied microeconomics; in fact, it is a primary focus of the field of academic inquiry known as strategy. This article presents a rigorous new statistical model that corrects for Bertrand, Mehta, and Mullainathan's (2002) shortcomings by taking usiness strategy into account. It also presents new empirical findings.

Four shortcomings of Bertrand, Mehta, and Mullainathan's (2002) methodology contributed to inaccurate results. First, the database they used, the Prowess database compiled by the Center for Monitoring Indian Economy (CMIE), had an explicit survivorship bias during the period of their study: in that early phase of its development, the CMIE systematically omitted all historical observations of companies that eventually failed, ceased to exist independently, or failed to provide disclosure for three consecutive years. In other words, Bertrand, Mehta, and Mullainathan's (2002) study of performance differences did not acknowledge that the least successful firms had been eliminated from their sample. Our sample consists of over 47,000 observations for the years 1989–1999; for the same period, Bertrand, Mehta, and Mullainathan (2002) had under 19,000 observations because their sample omitted a very large number of firms. Another reason for the difference in sample size is that over time CMIE expanded its inclusion criteria to encompass an incrementally larger percentage of the economy. (Our findings are consistent; however, even when using the sample-inclusion criteria from the time of Bertrand, Mehta, and Mullainathan's [2002] study along with the graveyard set from that era.) After their study, the survivorship-bias problem was eliminated when the CMIE publicly announced reincorporation of the historical graveyard set into its data. More recent observations no longer have a survivorship bias.

Second, Bertrand, Mehta, and Mullainathan (2002) did not deal with heteroscedasticity concerns. Though their dependent variable was raw operating profit unscaled by size or assets, no attempt was made to deal with the serious problem of heteroscedasticity. When the residuals are plotted against the fitted values, and when the Breusch-Pagan test is conducted, we see clear evidence of heteroscedasticity in the Indian corporate-performance data.

Third, Bertrand, Mehta, and Mullainathan (2002) did not deal with serial autocorrelation issues. Despite a 1989–1999 panel data set characterized by as many as 11 observations of the same firm, their implemented methodology did not take into account the likelihood of serial autocorrelation in performance at the firm level. There is a certain irony here, in that two of the study's authors circulated an influential article the same year faulting other academics for failing to take serial autocorrelation seriously (Bertrand, Duflo, and Mullainathan 2002). Serial autocorrelation, heteroscedasticity, and survivorship bias jointly led to sharply inflated coefficient estimates and underestimated standard errors in Bertrand, Mehta, and Mullainathan (2002).

Fourth, Bertrand, Mehta, and Mullainathan's (2002) models linked precise percentages of insider cash-flow ownership within a business group to expropriation, but did not disclose the serious shortcomings of their corporate-ownership data. Specifically, they did not disclose that in that era a very high percentage of eligible Indian firms—even publicly listed firms—only reported precise percentages of insider cash-flow ownership occasionally. The holes in the data are severe, thus introducing significant sample-composition bias. In part because of those holes, the data provider no longer distributes those corporate-ownership variables from the 1990s. The data provider referred us to a respected Indian academic who had previously downloaded all the corporate-ownership data from the 1990s, and as a robustness check we ran the regressions and found results materially different from those reported by Bertrand, Mehta, and Mullainathan (2002).

We differ from Bertrand, Mehta, and Mullainathan (2002) in (1) our emphasis on differences in firms' business strategies, (2) a methodology that simultaneously analyzes strategic-activity choices and corporate governance, and (3) a theory of business groups newly focused on unique business strategies. The statistical model we present here illustrates the critical importance of analyzing business strategy in the service of identifying superior or deficient corporate governance. It turns out that a neighboring firm's windfall may not predict the focal firm's windfall, even within the same industry, if the neighboring firm is pursuing a different business strategy. Suppose the neighboring firm's business strategy is essentially to resell inputs, while the focal firm uses firm-specific knowledge to recombine inputs into something more sophisticated. In the event of a positive shock to demand, the neighboring firm's operating costs do not change as the focal firm's do. Only the focal firm will respond to a sudden positive shock with an increase in recombination activities, thus incurring associated costs. We can only assess the quality of the focal firm's corporate governance accurately, therefore, by taking into account its choice of business strategy. When we do so, we exonerate many of the firms implicated in the literature's well-known finding that Indian business groups are typically expropriators: instead, we find that they are honest actors engaged in value creation.

In the corporate governance literature, our study most closely resembles that of Dharmapala and Khanna (2008), who tested for the effect on tunneling of post-1999 changes in Indian governance institutions and did not find a significant group effect. The authors attributed the lack of statistical significance in their regression result to a time-period effect and did not question the earlier result. Nor did they posit or investigate any explanation for the non-result, except to suggest that corporate governance might have changed radically after 1999 and that groups might have become more constrained.

Section 2 of this article describes the data and presents a series of summary statistics, Section 3 presents the model and results, Section 4 discusses the results, and Section 5 presents our conclusion.

## 1. Data and Summary Statistics

### 1.1. Data

Our data come from the Prowess database, a subscription-based data source maintained by the Center for Monitoring Indian Economy (CMIE), an independent think tank based in Mumbai. CMIE was founded in 1976 by the economist Narottam Shah; by 2009 it employed 330 researchers and others in 17 offices across India. CMIE is the most trusted source of data on Indian companies' financial statements and business group affiliations, and its data have been used in numerous academic studies in corporate finance and strategic management, including that of Bertrand, Mehta, and Mullainathan (2002). CMIE's historical product manuals provide an exhaustive description of every variable, and CMIE is meticulous in its efforts to correct for differences in how Indian companies combine or segment financial items.

We utilize longitudinal data from Prowess on public and privately held firms from fiscal years 1988–1989 through 2007–2008. For simplicity, we will refer to this period as 1989–2008, the years in which these companies issued their annual reports. By the time of our data download in October 2009, the Prowess data set for this period covered over 22,000 Indian firms; it included all firms listed on India's main stock exchanges and the vast universe of medium-sized and large privately held firms, and it tracked firm-level death and entry. The Prowess data set encompasses over 1,500 financial variables and ratios on each firm's balance sheets, profit-and-loss statements, and cash-flow statements, starting in 1989. Like Bertrand, Mehta, and Mullainathan (2002), we exclude state-owned and foreign-owned firms (until we arrive at a series of robustness checks) because our primary goal is to compare firms owned by Indian business groups to stand-alones. Again like Bertrand, Mehta, and Mullainathan (2002), we rely on CMIE's designation of a given firm as independent or as a member of a business group. By checking all merger and acquisition (M&A) transactions reported in the Thomson Reuters SDC database for the 1989–2008 period, we found a handful of uncoded changes in group affiliation; those few cases had no significant effect on the results.

We utilize the following variables, measured in Indian crore (1 crore = 10 million rupees) and converted to constant 1995 Indian crore following Bertrand, Mehta, and Mullainathan (2002). For firm size, we use total assets over time (Total assets) by taking the log of total assets (Ln assets). In other models, we utilize total annual revenues (Total sales). Further following Bertrand, Mehta, and Mullainathan (2002), we frequently focus on annual profits before depreciation/amortization, interest, and taxes (PBDITA).2 Because this measure of profit includes one-time transactions and extraordinary income, we also run robustness checks using annual profits before depreciation/amortization, interest, and taxes net of extraordinary income and one-time transactions (PBDITA net of NOI and NNRT). We also control for year of incorporation (Year of incorporation).

Further following Bertrand, Mehta, and Mullainathan (2002), we measure industry shocks and then calculate a predicted profit for each firm based on the profit shock experienced by the other firms in the same industry. The first step is to add up the profits and total assets of a given industry in each year (subtracting out those of the focal firm). The next step is to take the industry's ROA for a given year (subtracting out that of the focal firm) and multiply it by the focal firm's asset size in that year to predict what the firm would earn given the industry shock (Own shock). This variable is then multiplied by a group-affiliation dummy to separately test the effect of being a group affiliate (as opposed to being a stand-alone firm). The Own shock variable is also interacted with firm size and year of incorporation to replicate Bertrand, Mehta, and Mullainathan's (2002) results. For a series of robustness checks, we further include key control variables in the model by controlling for leverage, leverage interacted with Own shock, export orientation, export orientation interacted with Own shock, trading-based sales, trading-based sales interacted with Own shock, excise tax paid, and excise tax paid interacted with Own shock.

Again following Bertrand, Mehta, and Mullainathan (2002), we sought to test whether tunneling increases when insiders with a low percentage of cash-right ownership exercise control, and whether this phenomenon is compounded when insiders control a business group. Since 2002, Prowess has stopped providing historical shareholder-ownership data for the period 1989–2001 (and for most of the 2001–2006 time period for most firms), in part because few firms voluntarily reported shareholder ownership. (A large percentage of publicly listed firms, which were required to report, also failed to do so.) We were given a separate download of 1989–1999 shareholder-ownership data by a referral from the CMIE, and we ran a robustness check on that data. Following Bertrand, Mehta, and Mullainathan (2002), we examined the percentages of ownership held by directors and by outside others.

To understand how firms in the same industry pursue significantly different activities, we examine the extent to which firms engage in value-creating recombinations of inputs as proxied by excise taxes paid divided by total sales. It is noteworthy that excise taxes during the period in question were imposed at the federal level, with the exception of state-level taxes on alcohol and narcotic substances.3 A comparable proxy for these within-industry differences is sales of finished goods divided by total sales.

We contribute to the corporate governance literature by acknowledging the differences in firms' strategic activities. Specifically, we examine how industry shocks affect firms differently depending on their strategic activities. Thus, we examine the effect of shocks on firm-level changes in advertising; marketing; power; fuel-and-water expense; kWh usage of electricity; repairs and maintenance of plant and machinery; miscellaneous expenditure/residual selling, general, and administrative (SG&A); other amortization; provisions for bad/doubtful advances; total provisions; and directors' remuneration, while controlling for such key variables as size, year, and firm fixed effects. Most of our analysis excludes government- and foreign-owned firms (as did Bertrand, Mehta, and Mullainathan [2002]), but we run a series of robustness checks in which they are included.

We also examine the predictions of prior schools of thought that business groups must retrench as market institutions develop. Specifically, we determine the number of each group's affiliates and the number of three-digit industries they represent. We also look at how the relatedness of Indian business groups' diversification changed over the 1989–2008 time period, using information on industries' input-output relatedness from the 2006–2007 Indian Industry Input-Output Tables.4

### 1.2. Summary statistics

As Table 1 shows, group-affiliated firms in India are typically larger, more profitable, and slightly older than stand-alones. Foreign- and government-owned firms are typically larger than both group-affiliated firms and stand-alones; they also include more outliers in terms of performance. When we exclude performance outliers, we find foreign-owned firms to be the most profitable on average, by a slim margin over group-affiliated firms. Foreign- and government-owned firms also tend to be older than stand-alones, but foreign-owned firms are on average similar in age to group affiliates.

Table 1

Summary statistics

Our analyses (1989–2008)

(Group-Affiliated + Stand-alones) Groups Stand-alones Foreign Affiliates Government-Owned
Total assets 96.99 222.94 33.66 478.45 3253.90
(1193.758) (2026.551) (254.615) (2801.388) (13199.720)
Total sales 55.48 123.93 21.07 171.63 868.04
(439.866) (745.372) (88.916) (605.846) (4789.135)
PBDITA 10.51 25.11 3.17 48.92 279.21
(118.422) (200.873) (25.057) (246.589) (1063.351)
Ratio of PBDITA to total 0.06 0.07 0.06 -0.07 0.01
assets (1.927) (2.571) (1.502) (12.157) (1.882)
Ratio of PBDITA to total 0.08 0.09 0.07 0.11 0.08
assets (with outliers ≤ – 0.20 and ≥ 0.40 temporarily excluded) (0.093) (0.096) (0.090) (0.109) (0.094)
PBDITA net of NOI and 10.11 24.17 3.05 47.25 271.33
NNRT (117.205) (198.868) (24.729) (242.748) (1044.745)
Ratio of (PBDITA net of 0.04 0.04 0.05 -0.08 -0.03
NOI and NNRT) to total assets (1.863) (3.017) (0.798) (12.162) (1.805)
Ratio of (PBDITA net of 0.08 0.09 0.07 0.11 0.07
NOI and NNRT) to total assets (with outliers < = – 0.20 and > = 0.40 temporarily excluded) (0.092) (0.097) (0.090) (0.108) (0.093)
Year of incorporation 1981.08 1976.44 1983.42 1976.84 1969.09
(19.070) (22.117) (16.857) (22.945) (23.434)
Sample size with asset size information 141788 47435 94353 9588 7350
Our analyses (1989–2008)

(Group-Affiliated + Stand-alones) Groups Stand-alones Foreign Affiliates Government-Owned
Total assets 96.99 222.94 33.66 478.45 3253.90
(1193.758) (2026.551) (254.615) (2801.388) (13199.720)
Total sales 55.48 123.93 21.07 171.63 868.04
(439.866) (745.372) (88.916) (605.846) (4789.135)
PBDITA 10.51 25.11 3.17 48.92 279.21
(118.422) (200.873) (25.057) (246.589) (1063.351)
Ratio of PBDITA to total 0.06 0.07 0.06 -0.07 0.01
assets (1.927) (2.571) (1.502) (12.157) (1.882)
Ratio of PBDITA to total 0.08 0.09 0.07 0.11 0.08
assets (with outliers ≤ – 0.20 and ≥ 0.40 temporarily excluded) (0.093) (0.096) (0.090) (0.109) (0.094)
PBDITA net of NOI and 10.11 24.17 3.05 47.25 271.33
NNRT (117.205) (198.868) (24.729) (242.748) (1044.745)
Ratio of (PBDITA net of 0.04 0.04 0.05 -0.08 -0.03
NOI and NNRT) to total assets (1.863) (3.017) (0.798) (12.162) (1.805)
Ratio of (PBDITA net of 0.08 0.09 0.07 0.11 0.07
NOI and NNRT) to total assets (with outliers < = – 0.20 and > = 0.40 temporarily excluded) (0.092) (0.097) (0.090) (0.108) (0.093)
Year of incorporation 1981.08 1976.44 1983.42 1976.84 1969.09
(19.070) (22.117) (16.857) (22.945) (23.434)
Sample size with asset size information 141788 47435 94353 9588 7350

All financial figures in Tables 1–13 are in Indian Rs. Crore (1 crore = 10 million rupees) and have been converted to 1995 constant currency.

This table presents summary statistics on our sample, which covers the 1989–2008 period. We provide means with the standard deviations below them in parentheses. We start with a column showing the summary statistics for the combined sample of group affiliates and stand-alones. We next show the summary statistics for just the groups. We then present a column of summary statistics for the stand-alones only. To this point, we have excluded foreign-owned and government-owned companies; however, in the last columns we show the summary statistics for the foreign affiliates only and then for the government-owned firms only. For the purposes of a robustness check, we add the foreign-owned firms and government-owned firms to the sample in Tables 8–9.

The sample size reported above is for firms with asset size information, yet some firms are missing information on some variables, and therefore the sample size in the subsequent models will typically be somewhat smaller.

## 2. Model and Results

Table 2 replicates Bertrand, Mehta, and Mullainathan's (2002) results. The model regresses a firm's realized profit on its predicted profit (based on that of its industry) and on predicted profit interacted with group affiliation, as well as firm size, firm size interacted with predicted profit, and year of incorporation interacted with predicted profit. Time-invariant characteristics are controlled for via firm fixed effects, and the effect of time-period effects is controlled for through year dummies. Thus, the basic model is

(1)
$$perf_{kt} = a + b(pre{d_{kt}}) + c(grou{p_k}*pre{d_{kt}}) + d(control{s_{kt}}) + Fir{m_k} + Tim{e_t}.$$
where $$perf_{kt}$$ represents firm $$k$$'s PBDITA (or PBDITA net of prior-period and extraordinary income) at time $$t, pred_{kt}$$ is the firm's predicted PBDITA (or predicted PBDITA net of prior-period and extraordinary income) at time $$t$$, and $$group_{k}*pred_{kt}$$ is the interaction term between $$pred_{kt}$$ and the dummy variable for whether a firm is a member of a business group, $$group_{k}$$. We also include control variables (which increase in number as we proceed toward the full specification), firm fixed effects, and year dummies. Table 2 shows that, using Bertrand, Mehta, and Mullainathan's (2002) main model, we can replicate their result (that groups are expropriation devices) when not using clustering and robust standard errors. The negative and significant coefficient for group affiliation critically relies on the absence of clustering and robust standard errors.

Table 2

Replicating the prior result

Panel A: The 1989–2008 time period

Our analyses for the 1989–2008 time period (using PBDITA) and using 1995 constant rupees and focusing on group affiliates and stand-alone firms

Our analyses for the 1989–2008 time period (using PBDITA net of NOI and NNRT) and using 1995 constant rupees and focusing on group affiliates and stand-alone firms

Own shock 0.946 -0.310 1.591 11.069 0.909 -0.371 1.884 11.721
(0.015) (0.024) (0.269) (0.289) (0.015) (0.024) (0.280) (0.300)
Own shock* group 0.076 -0.206 0.077 -0.289 0.114 -0.173 0.114 -0.260
(0.015) (0.015) (0.015) (0.015) (0.015) (0.015) (0.015) (0.015)
Ln assets 0.375 3.930 0.377 5.000 0.743 4.230 0.742 5.273
(0.241) (0.242) (0.242) (0.243) (0.239) (0.241) (0.240) (0.241)
Own shock* ln assets  0.137  0.181  0.140  0.185
(0.002)  (0.002)  (0.002)  (0.002)
Own shock* year of incorporation   – 3.257E – 04 – 0.006   – 4.914E – 04 – 0.006
(1.356E – 04) (1.505E – 04)   (1.408E – 04) (1.56E – 04)
Sample size 141039 141039 140356 140356 141039 141039 140356 140356
R2 0.500 0.518 0.500 0.524 0.494 0.512 0.494 0.519

Panel B: The 1989–1999 time period

Our analyses for the 1989–1999 time period (using PBDITA) and using 1995 constant rupees and focusing on group affiliates and stand-alone firms

Our analyses for the 1989–1999 time period (using PBDITA net of NOI and NNRT) and using 1995 constant rupees and focusing on group affiliates and stand-alone firms

Own shock 0.671 -2.152 31.888 18.320 0.671 -2.057 32.880 19.076
(0.024) (0.034) (0.602) (0.554) (0.024) (0.034) (0.595) (0.557)
Own shock* group 0.136 -0.291 0.167 -0.240 0.125 -0.301 0.160 -0.243
(0.025) (0.022) (0.024) (0.022) (0.025) (0.022) (0.024) (0.022)
Ln assets 0.664 6.385 2.147 6.945 0.756 6.230 2.272 6.765
(0.231) (0.210) (0.225) (0.207) (0.224) (0.206) (0.218) (0.203)
Own shock* ln assets  0.333  0.309  0.321  0.294
(0.003)  (0.003)  (0.003)  (0.003)
Own shock* year of incorporation   – 0.016 – 0.010   – 0.016 – 0.011
(3.069E – 04) (2.795E – 04)   (3.036E – 04) (2.808E – 04)
Sample size 47936 47936 47918 47918 47936 47936 47918 47918
R2 0.339 0.491 0.383 0.509 0.349 0.489 0.395 0.508
Panel A: The 1989–2008 time period

Our analyses for the 1989–2008 time period (using PBDITA) and using 1995 constant rupees and focusing on group affiliates and stand-alone firms

Our analyses for the 1989–2008 time period (using PBDITA net of NOI and NNRT) and using 1995 constant rupees and focusing on group affiliates and stand-alone firms

Own shock 0.946 -0.310 1.591 11.069 0.909 -0.371 1.884 11.721
(0.015) (0.024) (0.269) (0.289) (0.015) (0.024) (0.280) (0.300)
Own shock* group 0.076 -0.206 0.077 -0.289 0.114 -0.173 0.114 -0.260
(0.015) (0.015) (0.015) (0.015) (0.015) (0.015) (0.015) (0.015)
Ln assets 0.375 3.930 0.377 5.000 0.743 4.230 0.742 5.273
(0.241) (0.242) (0.242) (0.243) (0.239) (0.241) (0.240) (0.241)
Own shock* ln assets  0.137  0.181  0.140  0.185
(0.002)  (0.002)  (0.002)  (0.002)
Own shock* year of incorporation   – 3.257E – 04 – 0.006   – 4.914E – 04 – 0.006
(1.356E – 04) (1.505E – 04)   (1.408E – 04) (1.56E – 04)
Sample size 141039 141039 140356 140356 141039 141039 140356 140356
R2 0.500 0.518 0.500 0.524 0.494 0.512 0.494 0.519

Panel B: The 1989–1999 time period

Our analyses for the 1989–1999 time period (using PBDITA) and using 1995 constant rupees and focusing on group affiliates and stand-alone firms

Our analyses for the 1989–1999 time period (using PBDITA net of NOI and NNRT) and using 1995 constant rupees and focusing on group affiliates and stand-alone firms

Own shock 0.671 -2.152 31.888 18.320 0.671 -2.057 32.880 19.076
(0.024) (0.034) (0.602) (0.554) (0.024) (0.034) (0.595) (0.557)
Own shock* group 0.136 -0.291 0.167 -0.240 0.125 -0.301 0.160 -0.243
(0.025) (0.022) (0.024) (0.022) (0.025) (0.022) (0.024) (0.022)
Ln assets 0.664 6.385 2.147 6.945 0.756 6.230 2.272 6.765
(0.231) (0.210) (0.225) (0.207) (0.224) (0.206) (0.218) (0.203)
Own shock* ln assets  0.333  0.309  0.321  0.294
(0.003)  (0.003)  (0.003)  (0.003)
Own shock* year of incorporation   – 0.016 – 0.010   – 0.016 – 0.011
(3.069E – 04) (2.795E – 04)   (3.036E – 04) (2.808E – 04)
Sample size 47936 47936 47918 47918 47936 47936 47918 47918
R2 0.339 0.491 0.383 0.509 0.349 0.489 0.395 0.508

In this table, we are able to replicate the prior result from Bertrand, Mehta, and Mullainathan (2002), by focusing on either 1989–2008 or 1989–1999, but as they do: without the use of clustering or robust standard errors.

In both Panels A and B, the analysis is focused on the group-affiliated firms and stand-alone firms in the Prowess database. Also, in both Panels A and B, the analysis is restricted to industries with two or more firms in a given year. Models included firm fixed effects and year fixed effects.

Bertrand, Mehta, and Mullainathan's (2002) experimental design was incomplete in that it did not take into account the possibility that firms in a given industry could systematically differ in their strategic activities. A core thesis of the strategy literature is that firms must choose how to compete relative to their industry peers (Caves and Porter 1977; Porter 1996). They aim for a form of competition that will add value (either by raising willingness to pay or by lowering cost) and that will be difficult for competitors to imitate (Porter 1980, 1985, 1996).

Indian firms have chosen two main forms of competition: greater emphasis on reselling finished goods and greater emphasis on complex recombination of inputs. By the criteria of numerous seminal publications in the strategy literature, this fundamental choice is a core strategic decision. In fact, Caves and Porter (1977, p. 246) even say that a core investment in product differentiation can lead to higher fixed cost intensity, and in turn to a less favorable position for one's competitors. Thus, any differences in operating leverage among Indian firms are an outcome of fundamental choices about technologies, manufacturing process, which people to hire, customers, etc. Clearly, resellers compete more on cost of delivering the already-finished good (a low-cost strategy), whereas recombination firms compete more on product differentiation and willingness to pay (a differentiation strategy).5

As Table 3 powerfully shows, group affiliates and stand-alones differ systematically in the business activities they pursue. Group affiliates in India are more likely to recombine inputs to generate added value and then to pay taxes on that added value. Stand-alone firms, by contrast, are more likely to simply resell finished goods made by other firms; even when stand-alones manufacture goods, they perform less value-added recombination than do group affiliates, on average. The systematic differences are stark. Group affiliates pay two-thirds more in value-added excise tax as a percentage of sales than do stand-alone firms. Groups also derive significantly less of their total revenues from trading/reselling finished goods, but this distinction does not fully account for the difference between the two types of firms in excise taxes paid divided by sales. The two types of firms hardly differ in their export orientations, ruling out that factor as the cause of their different responses to domestic industry shocks.

Table 3

Heterogeneity in strategic activities between group-affiliated firms and other types of firms

Excise Tax Paid/Sales T test for differences of means between group-affiliated and stand-alone firms Sales From Trading/Total Sales T test for differences of means between group-affiliated and stand-alone firms Exports/Sales T test fordifferences of means between group-affiliated and stand-alone firms
Stand-alone firms 0.032  0.144  0.108
(0.058) 0.000 (0.317) 0.000 (0.246) 0.000
Group-affiliated firms 0.054  0.104  0.115
(0.074)  (0.260)  (0.227)
Government-owned firms 0.035  0.096  0.040
(0.057)  (0.264)  (0.134)
Foreign-owned firms 0.061  0.124  0.138
(0.077)  (0.273)  (0.251)
Excise Tax Paid/Sales T test for differences of means between group-affiliated and stand-alone firms Sales From Trading/Total Sales T test for differences of means between group-affiliated and stand-alone firms Exports/Sales T test fordifferences of means between group-affiliated and stand-alone firms
Stand-alone firms 0.032  0.144  0.108
(0.058) 0.000 (0.317) 0.000 (0.246) 0.000
Group-affiliated firms 0.054  0.104  0.115
(0.074)  (0.260)  (0.227)
Government-owned firms 0.035  0.096  0.040
(0.057)  (0.264)  (0.134)
Foreign-owned firms 0.061  0.124  0.138
(0.077)  (0.273)  (0.251)

This table reports the results from difference of means tests between group-affiliated firms and other types of firms in India.

We next link our analysis of how firms' strategic activities differ to an analysis of how they react differently to shocks. Table 4 shows the results of an experiment demonstrating how each strategic activity cost changes when an industry shock occurs. For purposes of this simple illustration, each key operational activity cost is now the alternative dependent variable. We thus estimate

(2)
$$Strategic\,Activity\,Cos{t_{kt}} = a + b(pre{d_{kt}}) + c(lnasset{s_{kt}}) + Fir{m_k} + Tim{e_t}.$$
where $$StrategicActivityCost_{kt}$$ represents an increase (or decrease) in the magnitude of firm $$k$$'s operational activity cost items, including advertising, marketing, repair and maintenance, power, fuel, and water costs, etc., at time $$t$$. In this model, $$pred_{kt}$$ is the firm's predicted PBDITA (or its predicted PBDITA net of prior-period and extraordinary income) at time $$t$$; we also control for the log of firm size, firm fixed effects, and year dummies. We ran these specifications separately for the groups-only subsample and for the stand-alones-only sample to illustrate how these subsamples respond to industry shocks. Table 4 presents the regression coefficient for the shock variable, which represents the impact of the shock on a given strategic activity cost item for that subsample. Thus, the larger the positive coefficient, the more a positive shock leads to an increase in that strategic activity cost. Conversely, in the case of negative shocks, the larger the coefficient, the more that strategic activity is reduced; this makes sense, since trading-focused stand-alones rarely have the same strategic activity costs to cut during a negative shock. The industry shock is still being defined by the total population of groups and stand-alones.

Table 4

How group-affiliated firms and stand-alone firms react differently to shocks

 Groups  Stand-alones Difference in Magnitude: Advertising 0.011 0.005 (0.002) (0.001) Marketing 0.055 0.023 (0.009) (0.007) Distribution (for industries with a minimum 1% 0.071 0.071 average distribution-to-sales ratio) (0.016) (0.034) Power, fuel, and water (for manufacturing sector) 0.157 0.061 (0.045) (0.033) Repairs and maintenance of plant and machinery 0.018 0.007 (for manufacturing sector) (0.005) (0.004) Miscellaneous expenditure 0.064 0.024 (0.009) (0.005) Other amortization 0.026 0.001 (0.009) (0.001) Provisions for bad/doubtful advances 0.071 0.022 (0.018) (0.008) Total provisions 0.078 0.024 (0.019) (0.009) kWh electricity utilization increase 1219744 732653.4 (528603.7) (226600.3) And It’s Not That the Groups Are Paying Their Executives Much More: Directors’ remuneration 0.001 0.001 (3.396E-04) (0.001)
 Groups  Stand-alones Difference in Magnitude: Advertising 0.011 0.005 (0.002) (0.001) Marketing 0.055 0.023 (0.009) (0.007) Distribution (for industries with a minimum 1% 0.071 0.071 average distribution-to-sales ratio) (0.016) (0.034) Power, fuel, and water (for manufacturing sector) 0.157 0.061 (0.045) (0.033) Repairs and maintenance of plant and machinery 0.018 0.007 (for manufacturing sector) (0.005) (0.004) Miscellaneous expenditure 0.064 0.024 (0.009) (0.005) Other amortization 0.026 0.001 (0.009) (0.001) Provisions for bad/doubtful advances 0.071 0.022 (0.018) (0.008) Total provisions 0.078 0.024 (0.019) (0.009) kWh electricity utilization increase 1219744 732653.4 (528603.7) (226600.3) And It’s Not That the Groups Are Paying Their Executives Much More: Directors’ remuneration 0.001 0.001 (3.396E-04) (0.001)

In this table, we run an illustrative experiment where we are able to see how each strategic activity cost changes when there is an industry shock. Therefore, for the purpose of this very simple illustration, each key operational activity cost is now the alternative dependent variable. We thus estimate $$StrategicActivityCostkt = a + b(predkt) + c(lnassetskt) + Firmk +Timet$$, where StrategicActivityCostkt represents the increase (or decrease) in the magnitude of a firm $$k’$$s operational activity cost item, including advertising, marketing, distribution, repair, and maintenance of plant and machinery, power, fuel, and water costs, etc., at time $$t$$, and predkt is the firm’s predicted PBDITA (or alternatively its predicted PBDITA net of prior-period and extraordinary income) at time $$t$$. We also control for the log of firm size, firm fixed effects, and year dummies. We run these specifications separately for the groups-only subsample and then for the stand-alones-only sample to get a simple illustration of how these subsamples respond quite differently to industry shocks. What is presented is the regression coefficient for the shock variable (which represents the impact of the shock on that strategic activity cost item for that subsample). Thus, the larger the positive coefficient, the more a positive shock leads to an increase in that strategic activity cost. Conversely, in the case of negative shocks, the larger the coefficient, the more that strategic activity is reduced, which makes sense since trading-focused stand-alones don’t have these same strategic activity costs to cut during a negative shock. Robust standard errors adjusted for clustering at the firm level appear below each coefficient in parentheses. The industry shock is still being defined by the total population of groups and stand-alones.

Robust standard errors that are clustered at the firm level appear below the coefficients. All models include firm size, year dummies, and firm fixed effects.

Consistent with the finding that group affiliates tend to add value while stand-alones tend to trade finished goods, we find in Table 4 that group affiliates react to positive industry shocks by systematically taking on greater expenditures: for advertising and marketing; power, fuel, and water; repair and maintenance; miscellaneous costs, including higher donations to charity but more importantly a function of higher overall SG&A expense (miscellaneous expenses are akin to SG&A in the United States); other amortization; total provisions (especially provisions for bad/doubtful advances to customers, which are treated as an operational expense); and greater utilization of plant and machinery, as evident in a greater increase in electricity consumed. Stand-alones do not react similarly to industry shocks because they are apt to resell finished goods and to do less recombination of inputs. Also, increases in directors' remuneration do not differ meaningfully between group affiliates and stand-alones.

These strategic differences have clear governance implications. Group affiliates are more likely to increase operational expenses to benefit from positive industry shocks; stand-alones can simply sell more volume without shouldering new operational expenses. For legitimate reasons, therefore, group affiliates should see less incremental profit from an industry shock than do stand-alones. Groups find that any incremental revenues they might enjoy from a positive industry shock will in effect be taxed twice, as intensified operational activity costs and via higher excise taxes. There is thus a legitimate reason why groups enjoy more modest windfall profits from a positive industry shock than stand-alone firms.

Even so, the negative group effect on profit disappears when one uses modern econometric methods for clustering at the firm level with robust standard errors. It should be obvious that when the dependent variable is the amount of profit (as opposed to something scaled along the lines of ROA), the standard errors of the uncorrected models in Table 2 are likely to have heteroscedasticity problems. Table 5 shows that the statistical significance of the business group dummy in the main model in fact disappears when we use clustering at the firm level with robust standard errors. Thus, policy interventions broadly targeted at all Indian groups and based on an experiment that did not employ modern econometric techniques could be misguided. We can now return to our models with a new understanding that groups might appear to be theft devices when they are simply pursuing positive value-added activities.

Table 5

Introducing clustering with robust standard errors

Our analyses for the 1989–2008 time period (using PBDITA) and using 1995 constant rupees and focusing on group affiliates and stand-alone firms

Our analyses for the 1989–2008 time period (using PBDITA net of NOI and NNRT) and using 1995 constant rupees and focusing on group affiliates and stand-alone firms

Own shock 0.952 -0.268 1.681 11.054 0.915 -0.333 1.973 11.719
(0.235) (0.597) (4.480) (4.375) (0.240) (0.628) (5.028) (4.276)
Own shock* group 0.072 -0.211 0.072 -0.295 0.110 -0.178 0.110 -0.265
(0.241) (0.287) (0.241) (0.295) (0.246) (0.301) (0.245) (0.305)
Ln assets 0.563 4.218 0.560 5.275 0.921 4.521 0.914 5.552
(1.193) (2.567) (1.213) (2.878) (1.241) (2.651) (1.271) (2.917)
Own shock* ln assets  0.134  0.177  0.137  0.182
(0.059)  (0.073)  (0.063)  (0.075)
Own shock* year of incorporation   -3.682 -0.006   -0.001 -0.006
(0.002) (0.002)   (0.003) (0.002)
Sample size 141039 141039 140356 140356 141039 141039 140356 140356
R2 0.500 0.517 0.500 0.524 0.494 0.512 0.494 0.518
Our analyses for the 1989–2008 time period (using PBDITA) and using 1995 constant rupees and focusing on group affiliates and stand-alone firms

Our analyses for the 1989–2008 time period (using PBDITA net of NOI and NNRT) and using 1995 constant rupees and focusing on group affiliates and stand-alone firms

Own shock 0.952 -0.268 1.681 11.054 0.915 -0.333 1.973 11.719
(0.235) (0.597) (4.480) (4.375) (0.240) (0.628) (5.028) (4.276)
Own shock* group 0.072 -0.211 0.072 -0.295 0.110 -0.178 0.110 -0.265
(0.241) (0.287) (0.241) (0.295) (0.246) (0.301) (0.245) (0.305)
Ln assets 0.563 4.218 0.560 5.275 0.921 4.521 0.914 5.552
(1.193) (2.567) (1.213) (2.878) (1.241) (2.651) (1.271) (2.917)
Own shock* ln assets  0.134  0.177  0.137  0.182
(0.059)  (0.073)  (0.063)  (0.075)
Own shock* year of incorporation   -3.682 -0.006   -0.001 -0.006
(0.002) (0.002)   (0.003) (0.002)
Sample size 141039 141039 140356 140356 141039 141039 140356 140356
R2 0.500 0.517 0.500 0.524 0.494 0.512 0.494 0.518

We show in this table that the statistical significance of the business group dummy in the main model of Table 2 in fact goes away when one uses clustering at the firm level with robust standard errors.

Robust standard errors that are clustered at the firm level appear below the coefficients. All models include firm fixed effects and year fixed effects. Firms in an industry containing just one firm are excluded.

We next submit the models to a further series of controls: leverage and export orientation are two strategic choices that could also influence how a firm will be impacted by an industry shock. Table 6 shows that the coefficient for group affiliation is insignificantly different from zero even when we further control for the effects of leverage, leverage interacted with predicted profit, export orientation, and export orientation interacted with predicted profit. Table 7 further shows that the coefficient for group affiliation is insignificantly different from zero even when we add further controls for trading/sales, trading/sales interacted with Own shock, excise/sales, and excise/sales interacted with Own shock.

Table 6

Introducing additional control variables

Our analyses for the 1989–2008 time period (using PBDITA) and using 1995 constant rupees and focusing on group affiliates and stand-alone firms

Our analyses for the 1989–2008 time period (using PBDITA net of NOI and NNRT) and using 1995 constant rupees and focusing on group affiliates and stand-alone firms

Plus Marketing-Related

Plus Energy and Marketing-Related

Plus Energy

Plus Marketing-Related

Plus Energy and Marketing-Related

Own shock 0.946 1.141 1.190 -8.755 -12.531 -14.244 0.909 1.099 1.169 -7.068 -8.517 -11.070 -12.519
(0.238) (0.229) (0.244) (7.988) (8.471) (8.966) (0.242) (0.241) (0.231) (7.922) (8.529) (8.441) (9.000)
Own shock* group 0.076 0.132 -0.278 -0.435 -0.372 -0.272 0.114 0.181 -0.265 -0.418 -0.319 -0.353 -0.254
(0.243) (0.213) (0.256) (0.220) (0.233) (0.253) (0.247) (0.226) (0.248) (0.237) (0.258) (0.251) (0.272)
Ln assets 0.375 1.011 1.984 13.515 14.504 15.654 0.743 1.388 2.598 13.701 14.961 14.742 16.001
(1.091) (1.160) (2.393) (8.080) (8.239) (8.741) (1.136) (1.209) (2.569) (8.062) (8.587) (8.230) (8.755)
Own shock*    0.463 0.477 0.484    0.463 0.472 0.479 0.488
ln assets    (0.241) (0.246) (0.259)    (0.243) (0.258) (0.249) (0.263)
Own shock* year    0.003 0.005 0.006    0.002 0.003 0.004 0.005
of incorporation    (0.003) (0.004) (0.004)    (0.003) (0.004) (0.004) (0.004)
Leverage  0.006  0.881 0.936 0.995  0.011  0.911 0.976 0.968 1.033
(0.011)  (0.558) (0.573) (0.607)  (0.013)  (0.557) (0.593) (0.573) (0.609)
Own shock*  -0.889  -1.832 -1.922 -2.006  -0.914  -1.837 -1.929 -1.931 -2.024
leverage  (0.371)  (0.646) (0.662) (0.702)  (0.395)  (0.650) (0.690) (0.667) (0.710)
Export orientation   0.281 0.247 0.295 0.345   0.249 0.217 0.265 0.265 0.313
(0.320) (0.288) (0.326) (0.373)   (0.311) (0.276) (0.321) (0.314) (0.359)
Own shock* export   -0.001 -0.001 -0.001 -0.001   -0.001 -0.001 -0.001 -0.001 -0.001
orientation   (0.002) (0.001) (0.002) (0.002)   (0.002) (0.001) (0.002) (0.002) (0.002)
Sample size 141039 141039 105511 105090 105090 105090 141039 141039 105511 105090 105090 105090 105090
R2 0.500 0.519 0.240 0.324 0.337 0.344 0.494 0.514 0.228 0.307 0.316 0.320 0.329
Our analyses for the 1989–2008 time period (using PBDITA) and using 1995 constant rupees and focusing on group affiliates and stand-alone firms

Our analyses for the 1989–2008 time period (using PBDITA net of NOI and NNRT) and using 1995 constant rupees and focusing on group affiliates and stand-alone firms

Plus Marketing-Related

Plus Energy and Marketing-Related

Plus Energy

Plus Marketing-Related

Plus Energy and Marketing-Related

Own shock 0.946 1.141 1.190 -8.755 -12.531 -14.244 0.909 1.099 1.169 -7.068 -8.517 -11.070 -12.519
(0.238) (0.229) (0.244) (7.988) (8.471) (8.966) (0.242) (0.241) (0.231) (7.922) (8.529) (8.441) (9.000)
Own shock* group 0.076 0.132 -0.278 -0.435 -0.372 -0.272 0.114 0.181 -0.265 -0.418 -0.319 -0.353 -0.254
(0.243) (0.213) (0.256) (0.220) (0.233) (0.253) (0.247) (0.226) (0.248) (0.237) (0.258) (0.251) (0.272)
Ln assets 0.375 1.011 1.984 13.515 14.504 15.654 0.743 1.388 2.598 13.701 14.961 14.742 16.001
(1.091) (1.160) (2.393) (8.080) (8.239) (8.741) (1.136) (1.209) (2.569) (8.062) (8.587) (8.230) (8.755)
Own shock*    0.463 0.477 0.484    0.463 0.472 0.479 0.488
ln assets    (0.241) (0.246) (0.259)    (0.243) (0.258) (0.249) (0.263)
Own shock* year    0.003 0.005 0.006    0.002 0.003 0.004 0.005
of incorporation    (0.003) (0.004) (0.004)    (0.003) (0.004) (0.004) (0.004)
Leverage  0.006  0.881 0.936 0.995  0.011  0.911 0.976 0.968 1.033
(0.011)  (0.558) (0.573) (0.607)  (0.013)  (0.557) (0.593) (0.573) (0.609)
Own shock*  -0.889  -1.832 -1.922 -2.006  -0.914  -1.837 -1.929 -1.931 -2.024
leverage  (0.371)  (0.646) (0.662) (0.702)  (0.395)  (0.650) (0.690) (0.667) (0.710)
Export orientation   0.281 0.247 0.295 0.345   0.249 0.217 0.265 0.265 0.313
(0.320) (0.288) (0.326) (0.373)   (0.311) (0.276) (0.321) (0.314) (0.359)
Own shock* export   -0.001 -0.001 -0.001 -0.001   -0.001 -0.001 -0.001 -0.001 -0.001
orientation   (0.002) (0.001) (0.002) (0.002)   (0.002) (0.001) (0.002) (0.002) (0.002)
Sample size 141039 141039 105511 105090 105090 105090 141039 141039 105511 105090 105090 105090 105090
R2 0.500 0.519 0.240 0.324 0.337 0.344 0.494 0.514 0.228 0.307 0.316 0.320 0.329

We take the specification from the prior Table 5, which utilized clustering and robust standard errors, and add further control variables related to leverage and export orientation because these time-varying variables could directly impact how firms will react to industry shocks.

Note that the sample size is smaller when export orientation is included because the variable is not available for all firms.

Robust standard errors clustered at the firm level appear below the coefficients. All models include firm fixed effects and year fixed effects. Firms in an industry containing just one firm are excluded.

Table 7

Introducing further controls for trading/sales and excise/sales

Our analyses for the 1989–2008 time period (using PBDITA) and using 1995 constant rupees and focusing on group affiliates and stand-alone firms

Our analyses for the 1989–2008 time period (using PBDITA net of NOI and NNRT) and using 1995 constant rupees and focusing on group affiliates and stand-alone firms

Own shock -8.755 -9.668 6.865 5.885 -7.068 -7.946 9.610 8.640
(7.988) (7.784) (7.589) (7.729) (7.922) (7.753) (8.571) (8.749)
Own shock group -0.435 -0.419 -0.257 -0.230 -0.418 -0.403 -0.228 -0.200
(0.220) (0.224) (0.199) (0.198) (0.237) (0.242) (0.207) (0.205)
Ln assets 13.515 13.562 11.101 11.119 13.701 13.740 11.142 11.148
(8.080) (8.071) (5.137) (5.116) (8.062) (8.057) (4.940) (4.922)
Own shock ln assets 0.463 0.458 0.367 0.357 0.463 0.458 0.356 0.346
(0.241) (0.242) (0.149) (0.148) (0.243) (0.245) (0.143) (0.143)
Own shock year of incorporation 0.003 0.004 -0.004 -0.004 0.002 0.003 -0.006 -0.005
(0.003) (0.003) (0.004) (0.004) (0.003) (0.003) (0.005) (0.005)
Leverage 0.881 0.875 0.733 0.721 0.911 0.905 0.750 0.737
(0.558) (0.561) (0.373) (0.371) (0.557) (0.560) (0.359) (0.356)
Own shock leverage -1.832 -1.805 -1.594 -1.549 -1.837 -1.807 -1.543 -1.491
(0.646) (0.652) (0.454) (0.451) (0.650) (0.658) (0.428) (0.424)
Export orientation 0.247 0.108 0.377 0.178 0.217 0.083 0.348 0.145
(0.288) (0.110) (0.375) (0.146) (0.276) (0.099) (0.363) (0.127)
Own shock export orientation -0.001 -1.025E-04 -0.001 -4.646E-04 -0.001 3.07E-05 -0.001 -2.998E-04
(0.001) (0.001) (0.002) (0.001) (0.001) (0.001) (0.002) (0.001)
Trading sales/total sales  1.169  2.032  1.213  2.144
(1.080)  (0.766)  (1.058)  (0.749)
Own shock trading sales/total sales  -0.449  -0.654  -0.425  -0.650
(0.189)  (0.132)  (0.203)  (0.137)
Excise/sales   8.208 8.381   8.131 8.307
(4.435) (4.450)   (4.268) (4.282)
Own shock excise/sales   -7.706 -7.878   -8.192 -8.375
(4.338) (4.332)   (4.549) (4.542)
Sample size 105090 105090 105090 105090 105090 105090 105090 105090
R2 0.324 0.326 0.385 0.389 0.307 0.309 0.373 0.378
Our analyses for the 1989–2008 time period (using PBDITA) and using 1995 constant rupees and focusing on group affiliates and stand-alone firms

Our analyses for the 1989–2008 time period (using PBDITA net of NOI and NNRT) and using 1995 constant rupees and focusing on group affiliates and stand-alone firms

Own shock -8.755 -9.668 6.865 5.885 -7.068 -7.946 9.610 8.640
(7.988) (7.784) (7.589) (7.729) (7.922) (7.753) (8.571) (8.749)
Own shock group -0.435 -0.419 -0.257 -0.230 -0.418 -0.403 -0.228 -0.200
(0.220) (0.224) (0.199) (0.198) (0.237) (0.242) (0.207) (0.205)
Ln assets 13.515 13.562 11.101 11.119 13.701 13.740 11.142 11.148
(8.080) (8.071) (5.137) (5.116) (8.062) (8.057) (4.940) (4.922)
Own shock ln assets 0.463 0.458 0.367 0.357 0.463 0.458 0.356 0.346
(0.241) (0.242) (0.149) (0.148) (0.243) (0.245) (0.143) (0.143)
Own shock year of incorporation 0.003 0.004 -0.004 -0.004 0.002 0.003 -0.006 -0.005
(0.003) (0.003) (0.004) (0.004) (0.003) (0.003) (0.005) (0.005)
Leverage 0.881 0.875 0.733 0.721 0.911 0.905 0.750 0.737
(0.558) (0.561) (0.373) (0.371) (0.557) (0.560) (0.359) (0.356)
Own shock leverage -1.832 -1.805 -1.594 -1.549 -1.837 -1.807 -1.543 -1.491
(0.646) (0.652) (0.454) (0.451) (0.650) (0.658) (0.428) (0.424)
Export orientation 0.247 0.108 0.377 0.178 0.217 0.083 0.348 0.145
(0.288) (0.110) (0.375) (0.146) (0.276) (0.099) (0.363) (0.127)
Own shock export orientation -0.001 -1.025E-04 -0.001 -4.646E-04 -0.001 3.07E-05 -0.001 -2.998E-04
(0.001) (0.001) (0.002) (0.001) (0.001) (0.001) (0.002) (0.001)
Trading sales/total sales  1.169  2.032  1.213  2.144
(1.080)  (0.766)  (1.058)  (0.749)
Own shock trading sales/total sales  -0.449  -0.654  -0.425  -0.650
(0.189)  (0.132)  (0.203)  (0.137)
Excise/sales   8.208 8.381   8.131 8.307
(4.435) (4.450)   (4.268) (4.282)
Own shock excise/sales   -7.706 -7.878   -8.192 -8.375
(4.338) (4.332)   (4.549) (4.542)
Sample size 105090 105090 105090 105090 105090 105090 105090 105090
R2 0.324 0.326 0.385 0.389 0.307 0.309 0.373 0.378

We take the specification from the prior Table 6, which utilized clustering and robust standard errors, and add yet further control variables related to trading emphasis and value-added recombination emphasis as proxied by the firm’s ratio of excise taxes paid/total sales, since these time-varying variables may directly impact how firms will react to industry shocks.

Robust standard errors clustered at the firm level appear below the coefficients. All models include firm fixed effects and year fixed effects. Firms in an industry containing just one firm are excluded.

In Table 8, we perform a robustness check by adding foreign- and government-owned firms. Again, there is no significantly negative effect of group affiliation, but government ownership is significantly associated with tunneling in the full models. In Table 9, we add more control variables while retaining foreign- and government-owned firms. In this model, group affiliation at first appears to be negative and to have some statistical significance. But keep in mind that any residual negative group effect is an artifact of increased value-added activity costs in response to a shock. Thus, we estimate

(3)
\begin{align} &ProfitPlusValueAddedActivityCos{t_{kt}} = a + b(pre{d_{kt}}) \\ & \quad + c(grou{p_k}*pre{d_{kt}}) + d(control{s_{kt}}) + Fir{m_k} + Tim{e_t}, \\ \end{align}
where $$ProfitPlusValueAddedActivityCost_{kt}$$ represents firm $$k$$'s PBDITA (or, alternatively, its PBDITA net of prior-period and extraordinary income) plus any one of a few core value-added activity costs like power, fuel, and water (energy) expenses, at time $$t; pred_{kt}$$ is the firm's predicted PBDITA (or its predicted PBDITA net of prior-period and extraordinary income) at time $$t$$; and $$group_{k}* pred_{kt}$$ is the interaction term between $$pred_kt$$ and the dummy variable for whether a firm is a member of a business group, $$group_{k}$$. This model also includes further control variables, firm fixed effects, and year dummies. As Table 9 shows, when one simply nets out the effect of increased energy expenditure as a natural response to an industry shock, the statistical significance of the group-affiliation variable disappears completely.

Table 8

With the inclusion of foreign-owned firms and government-owned firms

Our analyses for the 1989–2008 time period (using PBDITA) and using 1995 constant rupees and including both foreign-owned firms and government-owned firms

Our analyses for the 1989–2008 time period (using PBDITA net of NOI and NNRT) and using 1995 constant rupees and including both foreign-owned firms and government-owned firms

Own shock 1.057 -0.083 6.432 5.745 0.454 -0.373 4.482 4.224
(0.184) (0.501) (3.391) (3.125) (0.555) (0.697) (3.601) (3.454)
Own shock* group 0.077 -0.190 0.075 -0.196 0.689 0.441 0.688 0.434
(0.195) (0.221) (0.197) (0.221) (0.558) (0.497) (0.559) (0.499)
Own shock * foreign-owned -0.105 -0.258 -0.170 -0.331 0.480 0.320 0.432 0.261
(0.195) (0.208) (0.201) (0.215) (0.558) (0.508) (0.568) (0.520)
Own shock * government-owned -0.197 -0.552 -0.266 -0.632 0.490 0.162 0.437 0.094
(0.195) (0.253) (0.207) (0.264) (0.578) (0.518) (0.593) (0.536)
Ln assets -0.857 3.422 -0.685 3.689 0.470 3.557 0.598 3.779
(0.667) (2.004) (0.688) (2.010) (1.245) (1.842) (1.230) (1.848)
Own shock* ln assets  0.126  0.128  0.096  0.098
(0.051)  (0.050)  (0.032)  (0.032)
Own shock* year of incorporation   -0.003 -0.003   -0.002 -0.002
(0.002) (0.002)   (0.002) (0.002)
Sample size 158522 158522 157600 157600 158498 158498 157576 157576
R2 0.642 0.663 0.645 0.667 0.674 0.685 0.676 0.687
Our analyses for the 1989–2008 time period (using PBDITA) and using 1995 constant rupees and including both foreign-owned firms and government-owned firms

Our analyses for the 1989–2008 time period (using PBDITA net of NOI and NNRT) and using 1995 constant rupees and including both foreign-owned firms and government-owned firms

Own shock 1.057 -0.083 6.432 5.745 0.454 -0.373 4.482 4.224
(0.184) (0.501) (3.391) (3.125) (0.555) (0.697) (3.601) (3.454)
Own shock* group 0.077 -0.190 0.075 -0.196 0.689 0.441 0.688 0.434
(0.195) (0.221) (0.197) (0.221) (0.558) (0.497) (0.559) (0.499)
Own shock * foreign-owned -0.105 -0.258 -0.170 -0.331 0.480 0.320 0.432 0.261
(0.195) (0.208) (0.201) (0.215) (0.558) (0.508) (0.568) (0.520)
Own shock * government-owned -0.197 -0.552 -0.266 -0.632 0.490 0.162 0.437 0.094
(0.195) (0.253) (0.207) (0.264) (0.578) (0.518) (0.593) (0.536)
Ln assets -0.857 3.422 -0.685 3.689 0.470 3.557 0.598 3.779
(0.667) (2.004) (0.688) (2.010) (1.245) (1.842) (1.230) (1.848)
Own shock* ln assets  0.126  0.128  0.096  0.098
(0.051)  (0.050)  (0.032)  (0.032)
Own shock* year of incorporation   -0.003 -0.003   -0.002 -0.002
(0.002) (0.002)   (0.002) (0.002)
Sample size 158522 158522 157600 157600 158498 158498 157576 157576
R2 0.642 0.663 0.645 0.667 0.674 0.685 0.676 0.687

In this table, we take the basic specification from the prior Table 5, which included clustering and robust standard errors, and we now add foreign-owned firms and government-owned firms to the sample for the purpose of a robustness check.

Robust standard errors clustered at the firm level appear below the coefficients. All models include firm fixed effects and year fixed effects. Firms in an industry containing just one firm are excluded.

Table 9

With the inclusion of foreign-owned firms and government-owned firms and further control variables

Our analyses for the 1989–2008 time period (using PBDITA) and using 1995 constant rupees and including both foreign-owned firms and government-owned firms

Our analyses for the 1989–2008 time period (using PBDITA net of NOI and NNRT) and using 1995 constant rupees and including both foreign-owned firms and government-owned firms

Plus energy

Plus energy

Own shock 0.199 0.159 4.564 4.509 3.458 1.794 1.778 5.836 5.828 4.858
(10.744) (10.702) (10.690) (10.644) (10.909) (11.113) (11.095) (10.841) (10.817) (11.125)
Own shock* group -0.308 -0.302 -0.311 -0.306 -0.217 -0.282 -0.281 -0.282 -0.282 -0.190
(0.147) (0.144) (0.149) (0.146) (0.152) (0.151) (0.149) (0.153) (0.151) (0.158)
Own shock * foreign-owned -0.263 -0.265 -0.152 -0.154 -0.078 -0.289 -0.289 -0.190 -0.190 -0.147
(0.133) (0.133) (0.156) (0.156) (0.209) (0.133) (0.133) (0.166) (0.166) (0.247)
Own shock * government-owned -0.999 -0.996 -0.958 -0.956 -0.969 -0.952 -0.952 -0.921 -0.921 -0.941
(0.244) (0.239) (0.241) (0.237) (0.239) (0.265) (0.264) (0.259) (0.258) (0.261)
Ln assets 8.240 8.022 7.922 7.756 8.343 7.999 7.935 7.817 7.778 8.585
(2.683) (2.455) (2.663) (2.438) (2.499) (2.711) (2.547) (2.685) (2.521) (2.585)
Own shock* ln assets 0.230 0.228 0.222 0.221 0.206 0.222 0.221 0.216 0.215 0.202
(0.059) (0.057) (0.060) (0.058) (0.058) (0.064) (0.062) (0.064) (0.063) (0.062)
Own shock* year of incorporation -3.453E-04 -3.218E-04 -0.003 -0.002 -0.002 -0.001 -0.001 -0.003 -0.003 -0.003
(0.005) (0.005) (0.005) (0.005) (0.005) (0.006) (0.006) (0.005) (0.005) (0.006)
Leverage 0.436 0.428 0.394 0.389 0.373 0.421 0.421 0.396 0.398 0.391
(0.164) (0.150) (0.143) (0.131) (0.132) (0.159) (0.150) (0.140) (0.133) (0.135)
Own shock* leverage -1.543 -1.533 -1.400 -1.394 -1.491 -1.380 -1.378 -1.282 -1.281 -1.395
(0.334) (0.320) (0.281) (0.273) (0.273) (0.339) (0.333) (0.281) (0.279) (0.279)
Export orientation 0.100 0.149 0.127 0.164 0.172 0.092 0.104 0.118 0.123 0.129
(0.232) (0.247) (0.237) (0.244) (0.256) (0.232) (0.218) (0.236) (0.218) (0.229)
Own shock* export orientation -6.37E-05 -3.743E-04 -2.36E-04 -4.632E-04 -0.001 -1.72E-05 -9.43E-05 -1.83E-04 -2.131E-04 -2.526E-04
(0.001) (0.002) (0.001) (0.002) (0.002) (0.001) (0.001) (0.002) (0.001) (0.001)
Trading sales/total sales  -3.427  -3.244 -3.688  -2.255  -2.251 -2.716
(2.311)  (2.376) (2.419)  (2.046)  (2.172) (2.225)
Own shock * trading sales/total sales  0.157  0.115 0.022  0.038  0.014 -0.079
(0.284)  (0.289) (0.290)  (0.258)  (0.268) (0.269)
Excise/sales   2.361 2.306 2.064   2.093 2.044 1.635
(0.956) (0.933) (1.046)   (1.217) (1.204) (1.414)
Own shock * excise/sales   -1.363 -1.355 -1.221   -1.396 -1.395 -1.174
(0.408) (0.406) (0.475)   (0.651) (0.645) (0.804)
Sample size 118432 118432 118432 118432 118432 118432 118432 118432 118432 118432
R2 0.616 0.617 0.626 0.626 0.625 0.618 0.618 0.624 0.624 0.624
Our analyses for the 1989–2008 time period (using PBDITA) and using 1995 constant rupees and including both foreign-owned firms and government-owned firms

Our analyses for the 1989–2008 time period (using PBDITA net of NOI and NNRT) and using 1995 constant rupees and including both foreign-owned firms and government-owned firms

Plus energy

Plus energy

Own shock 0.199 0.159 4.564 4.509 3.458 1.794 1.778 5.836 5.828 4.858
(10.744) (10.702) (10.690) (10.644) (10.909) (11.113) (11.095) (10.841) (10.817) (11.125)
Own shock* group -0.308 -0.302 -0.311 -0.306 -0.217 -0.282 -0.281 -0.282 -0.282 -0.190
(0.147) (0.144) (0.149) (0.146) (0.152) (0.151) (0.149) (0.153) (0.151) (0.158)
Own shock * foreign-owned -0.263 -0.265 -0.152 -0.154 -0.078 -0.289 -0.289 -0.190 -0.190 -0.147
(0.133) (0.133) (0.156) (0.156) (0.209) (0.133) (0.133) (0.166) (0.166) (0.247)
Own shock * government-owned -0.999 -0.996 -0.958 -0.956 -0.969 -0.952 -0.952 -0.921 -0.921 -0.941
(0.244) (0.239) (0.241) (0.237) (0.239) (0.265) (0.264) (0.259) (0.258) (0.261)
Ln assets 8.240 8.022 7.922 7.756 8.343 7.999 7.935 7.817 7.778 8.585
(2.683) (2.455) (2.663) (2.438) (2.499) (2.711) (2.547) (2.685) (2.521) (2.585)
Own shock* ln assets 0.230 0.228 0.222 0.221 0.206 0.222 0.221 0.216 0.215 0.202
(0.059) (0.057) (0.060) (0.058) (0.058) (0.064) (0.062) (0.064) (0.063) (0.062)
Own shock* year of incorporation -3.453E-04 -3.218E-04 -0.003 -0.002 -0.002 -0.001 -0.001 -0.003 -0.003 -0.003
(0.005) (0.005) (0.005) (0.005) (0.005) (0.006) (0.006) (0.005) (0.005) (0.006)
Leverage 0.436 0.428 0.394 0.389 0.373 0.421 0.421 0.396 0.398 0.391
(0.164) (0.150) (0.143) (0.131) (0.132) (0.159) (0.150) (0.140) (0.133) (0.135)
Own shock* leverage -1.543 -1.533 -1.400 -1.394 -1.491 -1.380 -1.378 -1.282 -1.281 -1.395
(0.334) (0.320) (0.281) (0.273) (0.273) (0.339) (0.333) (0.281) (0.279) (0.279)
Export orientation 0.100 0.149 0.127 0.164 0.172 0.092 0.104 0.118 0.123 0.129
(0.232) (0.247) (0.237) (0.244) (0.256) (0.232) (0.218) (0.236) (0.218) (0.229)
Own shock* export orientation -6.37E-05 -3.743E-04 -2.36E-04 -4.632E-04 -0.001 -1.72E-05 -9.43E-05 -1.83E-04 -2.131E-04 -2.526E-04
(0.001) (0.002) (0.001) (0.002) (0.002) (0.001) (0.001) (0.002) (0.001) (0.001)
Trading sales/total sales  -3.427  -3.244 -3.688  -2.255  -2.251 -2.716
(2.311)  (2.376) (2.419)  (2.046)  (2.172) (2.225)
Own shock * trading sales/total sales  0.157  0.115 0.022  0.038  0.014 -0.079
(0.284)  (0.289) (0.290)  (0.258)  (0.268) (0.269)
Excise/sales   2.361 2.306 2.064   2.093 2.044 1.635
(0.956) (0.933) (1.046)   (1.217) (1.204) (1.414)
Own shock * excise/sales   -1.363 -1.355 -1.221   -1.396 -1.395 -1.174
(0.408) (0.406) (0.475)   (0.651) (0.645) (0.804)
Sample size 118432 118432 118432 118432 118432 118432 118432 118432 118432 118432
R2 0.616 0.617 0.626 0.626 0.625 0.618 0.618 0.624 0.624 0.624

In this table, we take the specification from the prior Table 8, and then add control variables related to leverage, export orientation, trading emphasis, and value-added recombination emphasis as proxied by the ratio of excise taxes paid/total sales.

Robust standard errors clustered at the firm level appear below the coefficients. All models include firm fixed effects and year fixed effects. Firms in an industry containing just one firm are excluded.

Thus far we have established three key empirical results. First, the effect on profit of business group affiliation in India is often slightly negative for purely legitimate business reasons, namely that business groups typically take on larger operational costs in response to positive industry shocks than stand-alones. Second, this legitimate negative effect of business group affiliation largely disappears when one applies modern econometric techniques for clustering at the firm level and robust standard errors. Third, even when the group affiliation effect is “resuscitated,” it disappears when one takes into account the effect of a few operational-cost items, notably energy.

As a supplementary exercise, we have performed numerous industry case studies as background research, all of which led to similar findings. What we have found through econometrics is no mere statistical artifact. It is confirmable through in-depth industry-level examinations. As a purely illustrative but not causal or motivating example, the plastic products industry is one where group firms and stand-alone firms within the same industry have very different levels of value-added activities and react to industry shocks differently.

In this industry, the two largest firms in the 2005–2007 time period were Nilkamal Ltd., a stand-alone firm, and Supreme Industries Ltd., a group affiliate firm. In 2006, Supreme and Nilkamal had sales of Rs. 1125 crore and Rs. 399 crore, respectively.6

The industry experienced a large, positive shock starting in the year 2005. The sum of industry profit had declined between 2003 and 2005 but then sharply increased between 2005 and 2006, and then again between 2006 and 2007, exceeding the 2003 levels.

We then checked how this positive industry shock affected the two firms. As Figure 1 indicates, both the stand-alone firm (Nilkamal) and the group affiliate (Supreme Industries) experienced positive sales growth in this period, though the group affiliate experienced a higher rate of growth in sales.

Figure 1

Sales

Figure 1

Sales

Next we compared the group affiliate and the stand-alone firm on two important measures related to the extent of value-added activities: (1) excise over sales and (2) trading sales over total sales. Both of these variables reflect the extent of value-added activities. As Figure 2 indicates, prior to the positive industry shock, the group firm, Supreme Industries, had a much higher excise over sales ratio compared to the stand-alone firm. However, the even more interesting fact here is that post the industry shock, the excise over sales measure at the group firm experienced a marginal increase while the same ratio experienced a decline at the stand-alone firm immediately post the shock.

Figure 2

Excise tax over sales

Figure 2

Excise tax over sales

Next we compared trading sales over total sales at the two firms over this period and find that prior to the shock, the stand-alone firm had a significantly higher ratio of trading sales over sales compared to the group firm. While the ratio for the stand-alone firm was around 32% in 2004, the same ratio was around 9% for the group firm. Figure 3 also indicates that post the positive industry shock, the ratio of trading sales to total sales experienced a marginal increase at the stand-alone firm and the same ratio experienced a decline at the group firm.

Figure 3

Trading sales over total sales

Figure 3

Trading sales over total sales

These findings indicate that in this industry, the leading group firm not only had higher levels of value-added activities compared to the leading stand-alone firm, the findings also indicate that post the positive industry shock, the group firm moved toward higher value-added activities while the stand-alone firm did not.

These facts are corroborated by an analysis of the annual reports of the two firms from that time period. The annual reports of Nilkamal, the stand-alone firm, indicate that the main focus of the firm was the moulded furniture business where there were “low entry barriers” and where the industry was dominated by the unorganized sector.7 The moulded furniture business comprised around 75% of the revenues for Nilkamal in 2007, and as the annual report indicates, a key challenge for the firm was to reduce its dependence on the “Monoblocks segment (basic chairs) wherein the competition is high and margins are lower.”8

In comparison, the group firm Supreme Industries was focused on products as diverse as plastic piping systems, material handling systems, and cross-laminated film.9 This focus on value-added products accelerated post the positive industry shock. To quote the annual report in 2007–08, “the company has made large investments in last two years in its plastic piping system, cross-laminated film, protective packaging products, industrial products, material handling systems and furniture product segments.”10 There were several product launches in areas as diverse as CPVC, hot and cold water plumbing systems, HDPE pipes, injection moulded inspection chambers, and rotational moulding.11 Supreme Industries also became the first Indian company to “launch injection moulded pallets in India successfully.”12

Though the above example is merely illustrative rather than a form of proof, we believe that looking at business groups' ability to do complex recombination of inputs is strongly supported by both small-sample qualitative and large-sample quantitative evidence. Given that the literature is primarily oriented toward the latter, we devote most of this study to large-sample quantitative evidence.

Returning to our large-sample quantitative focus, we perform robustness checks on randomly chosen subsamples and find substantively similar results—no matter what period we focus on. Even when the business-group-affiliation effect can be temporarily resuscitated in some of these subsamples, it almost always disappears when we examine the effect of a small set of logical operational expenditures for energy, wages, advertising, marketing, and plant repairs, either in isolation or in combination.

We also perform several robustness checks. A two-digit definition of an industry or a finer-grained four- or five-digit definition produces substantively similar results. We opt for three-digit codes because the two-digit formulation includes industries subject to very different shocks, and the four- and five-digit formulation differentiates firms facing the same shocks, thus eliminating pertinent firms from analysis and compromising the measure of the shocks themselves. In tables in which we control for export intensity, the financial sector is omitted; we have confirmed, however, that the results are substantively similar if we reinstate the financial sector. We separately find that our results are substantively similar even without the use of clustering. We also find substantively similar results when we artificially constrain the definition of groups to include only groups of at least 5, 10, 20, 25, 30, or even 50 affiliates. Our results are robust to the inclusion of advertising intensity, advertising intensity interacted with the shock variable, R&D intensity, and R&D intensity interacted with the shock variable. They are also robust to the inclusion of those variables while omitting the two excise-related variables.

We wondered whether groups' corporate governance would appear worse if we took into account the effects of changes in price dispersion; this does not prove to be the case. Using data from Prowess, we calculate average annual product prices in each industry and then calculate price dispersion as the standard deviation of product prices divided by average product price. Price dispersion turns out not to be statistically significant in its relationship to profit shocks, defined in terms of PBDITA or of PBDITA/assets and controlling for industry fixed effects and year dummies. Thus, branded firms do not enjoy significantly higher price premiums on average as an effect of profit shocks. Nor is there evidence of a differential group effect when examining industries above or below the price-dispersion median. When further controlling for price dispersion, the main group coefficient in the full model is not substantively affected; thus, an omitted variable for price dispersion is not driving the group coefficient.

Next we examine the effect of insider cash-rights ownership and ownership by minority shareholders. As noted, CMIE no longer sells historical ownership data, partly because of the difficulty of tracking changing government-mandated ownership categories over time, and partly because during much of the period in question firms only disclosed such information occasionally. We therefore affix a big caveat to historical shareholder-ownership analysis, and perform such analysis only to show that we have not neglected any of Bertrand, Mehta, and Mullainathan's (2002) ownership findings. Thus, with considerable caution, we estimate

(4)
\begin{align} per{f_{kt}} = & a + b(pre{d_{kt}}) + c(pre{d_k}*Director\,Ownership\,Percentag{e_{kt}}) \\ & + d(control{s_{kt}}) + Fir{m_k} + Tim{e_t}, \\ \end{align}
where $$perf_{kt}$$ represents firm $$k$$'s PBDITA (or its PBDITA net of prior-period and extraordinary income) at time $$t, pred_{kt}$$ is the firm's predicted PBDITA (or its predicted PBDITA net of prior-period and extraordinary income) at time $$t$$, and $$pred_{kt}* DirectorOwnershipPercentage_{kt}$$ represents the interaction term between $$pred_{kt}$$ and the percentage of the firm's ownership held by its directors. We also include a set of control variables, firm fixed effects, and year dummies.

Here our results contradict conventional wisdom. As Table 10 shows, firms actually benefit less from positive industry shocks when they are group-affiliated and their directors possess a lower percentage of cash-right ownership. There is no robust association between minority cash-flow rights and tunneling for group-affiliated firms. Because ownership data for the 1989–1999 period are largely restricted to the very largest Indian firms, and typically cover just one year, such analysis is provided only as a point of comparison with the earlier experiment.

Table 10

An analysis of insider and outside ownership stakes

Panel A: Director equity
Our analyses for the 1989–1999 time period (using PBDITA) and using 1995 constant rupees and focusing on group affiliates and stand-alone firms

Our analyses for the 1989–1999 time period (using PBDITA net of NOI and NNRT) and using 1995 constant rupees and focusing on group affiliates and stand-alone firms

Groups (1) Groups (2) Stand-alones (3) Stand-alones (4) Groups (1) Groups (2) Stand-alones (3) Stand-alones (4)
Own shock 0.256 35.618 1.172 -12.213 0.349 24.516 1.201 -18.776
(0.257) (17.303) (0.199) (21.002) (0.182) (12.996) (0.195) (22.608)
Own shock * director equity -0.016 -0.020 0.001 -3.972E-04 -0.017 -0.019 2.6E-05 -0.001
(0.007) (0.011) (0.005) (0.004) (0.006) (0.008) (0.005) (0.004)
Ln assets 24.075 23.518 -0.419 1.665 20.552 19.982 -0.902 1.063
(9.101) (6.755) (1.338) (1.551) (7.477) (7.349) (1.264) (1.643)
Own shock * ln assets  -0.148  0.131  -0.111  0.104
(0.164)  (0.122)  (0.115)  (0.133)
Own shock * year of incorporation  -0.017  0.006  -0.012  0.010
(0.008)  (0.011)  (0.006)  (0.012)
Sample size 1717 1717 1839 1839 1717 1717 1839 1839
Adjusted R2 0.072 0.125 0.619 0.641 0.085 0.107 0.638 0.659

Panel B: Other ownership
Our analyses for the 1989–1999 time period (using PBDIT) and using 1995 constant rupees and focusing on group affiliates and stand-alone firms

Our analyses for the 1989–1999 time period (using PBDIT net of NOI and NNRT) and using 1995 constant rupees and focusing on group affiliates and stand-alone firms

Groups (1 Groups (2) Stand-alones (3) Stand-alones (4) Groups (1) Groups (2) Stand-alones (3) Stand-alones (4)

Own shock 0.412 34.219 0.975 -20.631 0.459 23.295 0.987 -27.099
(0.298) (17.371) (0.187) (19.431) (0.212) (12.940) (0.168) (21.152)
Own shock * other ownership -0.006 -0.005 0.009 0.011 -0.004 -0.004 0.009 0.011
(0.004) (0.004) (0.004) (0.004) (0.003) (0.003) (0.004) (0.004)
Ln assets 22.359 22.455 -0.523 1.575 18.969 19.071 -1.023 0.942
(8.822) (6.733) (1.332) (1.492) (7.170) (7.386) (1.256) (1.591)
Own shock * ln assets  -0.116  0.116  -0.080  0.088
(0.154)  (0.112)  (0.113)  (0.124)
Own shock * year of incorporation  -0.017  0.010  -0.011  0.014
(0.008)  (0.010)  (0.006)  (0.011)
Sample size 1717 1717 1839 1839 1717 1717 1839 1839
R2 0.081 0.126 0.632 0.658 0.088 0.105 0.652 0.677
Panel A: Director equity
Our analyses for the 1989–1999 time period (using PBDITA) and using 1995 constant rupees and focusing on group affiliates and stand-alone firms

Our analyses for the 1989–1999 time period (using PBDITA net of NOI and NNRT) and using 1995 constant rupees and focusing on group affiliates and stand-alone firms

Groups (1) Groups (2) Stand-alones (3) Stand-alones (4) Groups (1) Groups (2) Stand-alones (3) Stand-alones (4)
Own shock 0.256 35.618 1.172 -12.213 0.349 24.516 1.201 -18.776
(0.257) (17.303) (0.199) (21.002) (0.182) (12.996) (0.195) (22.608)
Own shock * director equity -0.016 -0.020 0.001 -3.972E-04 -0.017 -0.019 2.6E-05 -0.001
(0.007) (0.011) (0.005) (0.004) (0.006) (0.008) (0.005) (0.004)
Ln assets 24.075 23.518 -0.419 1.665 20.552 19.982 -0.902 1.063
(9.101) (6.755) (1.338) (1.551) (7.477) (7.349) (1.264) (1.643)
Own shock * ln assets  -0.148  0.131  -0.111  0.104
(0.164)  (0.122)  (0.115)  (0.133)
Own shock * year of incorporation  -0.017  0.006  -0.012  0.010
(0.008)  (0.011)  (0.006)  (0.012)
Sample size 1717 1717 1839 1839 1717 1717 1839 1839
Adjusted R2 0.072 0.125 0.619 0.641 0.085 0.107 0.638 0.659

Panel B: Other ownership
Our analyses for the 1989–1999 time period (using PBDIT) and using 1995 constant rupees and focusing on group affiliates and stand-alone firms

Our analyses for the 1989–1999 time period (using PBDIT net of NOI and NNRT) and using 1995 constant rupees and focusing on group affiliates and stand-alone firms

Groups (1 Groups (2) Stand-alones (3) Stand-alones (4) Groups (1) Groups (2) Stand-alones (3) Stand-alones (4)

Own shock 0.412 34.219 0.975 -20.631 0.459 23.295 0.987 -27.099
(0.298) (17.371) (0.187) (19.431) (0.212) (12.940) (0.168) (21.152)
Own shock * other ownership -0.006 -0.005 0.009 0.011 -0.004 -0.004 0.009 0.011
(0.004) (0.004) (0.004) (0.004) (0.003) (0.003) (0.004) (0.004)
Ln assets 22.359 22.455 -0.523 1.575 18.969 19.071 -1.023 0.942
(8.822) (6.733) (1.332) (1.492) (7.170) (7.386) (1.256) (1.591)
Own shock * ln assets  -0.116  0.116  -0.080  0.088
(0.154)  (0.112)  (0.113)  (0.124)
Own shock * year of incorporation  -0.017  0.010  -0.011  0.014
(0.008)  (0.010)  (0.006)  (0.011)
Sample size 1717 1717 1839 1839 1717 1717 1839 1839
R2 0.081 0.126 0.632 0.658 0.088 0.105 0.652 0.677

This table is simply meant to show that the available ownership data from India from the 1990s were quite limited and that the results even for this limited available sample of (typically) the largest firms do not match with prevailing conventional wisdom.

Robust standard errors clustered at the firm level appear below the coefficients. All models include firm fixed effects and year fixed effects. Firms in an industry containing just one firm are excluded.

In Table 11, we take a fresh look at Bertrand, Mehta, and Mullainathan's (2002) suggestion that the stock market knew which public firms were being tunneled and which were receiving funds transferred surreptitiously from other firms. Using the most comprehensive data on market-to-book values from a data source, Capitaline, which collects capital-market information on publicly listed firms, we find results that differ from Bertrand, Mehta, and Mullainathan (2002). We follow their models in examining whether a firm's profits are predicted by comparing its Tobin's q to that of its publicly traded group affiliates, but find no evidence of tunneling per se. Like these authors, we do find that the firms with the highest valuations in their business groups tend to outperform their peers during a positive industry shock. It surprises us that they interpreted this finding as a sign of sinister tunneling; the simplest and most logical explanation is that the most skilled firms react the most constructively to positive opportunities. Unlike these authors, as Table 10 shows, we do not find evidence of firms being systematically affected by other group affiliates' shocks.

Table 11

An analysis of within-group relative valuation and related variables

Our analyses for the 1989–2008 time period (using PBDITA) and using 1995 constant rupees

Our analyses for the 1989–1999 time period (using PBDITA net of NOI and NNRT) and using 1995 constant rupees

(1) (2) (3) (4) (1) (2) (3) (4)
Own shock 1.015 1.025 0.998 1.021 1.038 1.025 1.006 1.045
(0.148) (0.074) (0.158) (0.145) (0.159) (0.080) (0.174) (0.158)
Own shock * firm Q 0.001   0.016 0.010   0.026
(0.054)   (0.065) (0.059)   (0.080)
Own shock * relative Q  0.105    0.119
(0.044)    (0.045)
Own shock * group Q   -0.014 -0.021   -0.010 -0.022
(0.067) (0.083)   (0.072) (0.094)
Group shock 0.008 0.013 0.028 0.026 0.009 0.015 0.029 0.026
(0.009) (0.012) (0.041) (0.039) (0.009) (0.013) (0.042) (0.039)
Group shock * firm Q -0.002   -0.002 -0.002   -0.002
(0.003)   (0.002) (0.003)   (0.003)
Group shock * relative Q  -0.003    -0.004
(0.003)    (0.003)
Group shock * group Q   -0.013 -0.011   -0.013 -0.011
(0.022) (0.020)   (0.022) (0.020)
Sample size 14253 14253 14253 14253 14253 14253 14253 14253
R2 0.441 0.447 0.442 0.442 0.435 0.441 0.435 0.435
Our analyses for the 1989–2008 time period (using PBDITA) and using 1995 constant rupees

Our analyses for the 1989–1999 time period (using PBDITA net of NOI and NNRT) and using 1995 constant rupees

(1) (2) (3) (4) (1) (2) (3) (4)
Own shock 1.015 1.025 0.998 1.021 1.038 1.025 1.006 1.045
(0.148) (0.074) (0.158) (0.145) (0.159) (0.080) (0.174) (0.158)
Own shock * firm Q 0.001   0.016 0.010   0.026
(0.054)   (0.065) (0.059)   (0.080)
Own shock * relative Q  0.105    0.119
(0.044)    (0.045)
Own shock * group Q   -0.014 -0.021   -0.010 -0.022
(0.067) (0.083)   (0.072) (0.094)
Group shock 0.008 0.013 0.028 0.026 0.009 0.015 0.029 0.026
(0.009) (0.012) (0.041) (0.039) (0.009) (0.013) (0.042) (0.039)
Group shock * firm Q -0.002   -0.002 -0.002   -0.002
(0.003)   (0.002) (0.003)   (0.003)
Group shock * relative Q  -0.003    -0.004
(0.003)    (0.003)
Group shock * group Q   -0.013 -0.011   -0.013 -0.011
(0.022) (0.020)   (0.022) (0.020)
Sample size 14253 14253 14253 14253 14253 14253 14253 14253
R2 0.441 0.447 0.442 0.442 0.435 0.441 0.435 0.435

In this table, we take a fresh look at Bertrand, Mehta, and Mullainathan’s (2002) results suggesting that the stock market knew which firms were being tunneled and which were receiving funds taken surreptitiously from other firms. Using the most comprehensive data on market-to-book values from a separate data source, Capitaline’s Data Set, which focuses on publicly listed firms and collects yet more detailed capital market information on those publicly firms, we find different results from what was found in the prior study. We follow Bertrand, Mehta, and Mullainathan’s (2002) models in seeing if the firm’s profits are predicted by whether the firm’s Tobin’s q is high relative to the rest of its publicly traded group affiliates. In contrast to their results, we don’t find any evidence of tunneling per se. Like Bertrand, Mehta, and Mullainathan’s (2002), we do find that firms that have the highest valuations in their business groups tend to outperform their peers in a positive industry shock. We find it surprising that Bertrand, Mehta, and Mullainathan’s (2002) interpreted this as a sign of sinister tunneling, whereas by Occam’s Razor, the simplest and most logical explanation is that the most skilled firms within a group are the ones that react the most constructively to positive opportunities.

Robust standard errors clustered at the firm level appear below the coefficients. All models include firm size, year fixed effects, and firm fixed effects.

Table 12

Superior profit creation and capabilities development in groups, including after liberalization and market development

Year Group: Stand-alone ROA BMM Ratio Group: Stand-alone ROA NOINNRT Ratio Group Multiplier In Terms of Advertising Intensity Group Multiplier In Terms of R&D Intensity
1989 1.08 1.08 1.44 253.83
1990 1.10 1.10 1.98 128.15
1991 1.08 1.12 1.79 2.14
1992 1.14 1.21 1.66 3.10
1993 1.20 1.19 1.50 2.60
1994 1.12 1.12 1.57 2.28
1995 1.17 1.18 1.61 2.28
1996 1.23 1.23 1.79 2.06
1997 1.24 1.21 1.70 2.69
1998 1.25 1.20 1.85 2.70
1999 1.23 1.21 1.92 5.17
2000 1.12 1.10 1.65 2.66
2001 1.30 1.27 1.47 2.06
2002 1.14 1.12 1.47 2.39
2003 1.25 1.23 1.37 2.44
2004 1.22 1.22 1.55 2.53
2005 1.18 1.19 1.40 2.08
2006 1.16 1.16 1.47 0.98
2007 1.18 1.18 1.43 1.74
2008 1.15 1.18 1.42 1.70
Year Group: Stand-alone ROA BMM Ratio Group: Stand-alone ROA NOINNRT Ratio Group Multiplier In Terms of Advertising Intensity Group Multiplier In Terms of R&D Intensity
1989 1.08 1.08 1.44 253.83
1990 1.10 1.10 1.98 128.15
1991 1.08 1.12 1.79 2.14
1992 1.14 1.21 1.66 3.10
1993 1.20 1.19 1.50 2.60
1994 1.12 1.12 1.57 2.28
1995 1.17 1.18 1.61 2.28
1996 1.23 1.23 1.79 2.06
1997 1.24 1.21 1.70 2.69
1998 1.25 1.20 1.85 2.70
1999 1.23 1.21 1.92 5.17
2000 1.12 1.10 1.65 2.66
2001 1.30 1.27 1.47 2.06
2002 1.14 1.12 1.47 2.39
2003 1.25 1.23 1.37 2.44
2004 1.22 1.22 1.55 2.53
2005 1.18 1.19 1.40 2.08
2006 1.16 1.16 1.47 0.98
2007 1.18 1.18 1.43 1.74
2008 1.15 1.18 1.42 1.70

Next we turn our attention to further developing our own knowledge-based, recombinative capabilities view of business groups. Building on our earlier tables, which showed that groups systematically are more engaged in value-added activities and react to shocks with a large increase in the utilization of those value-added activities, here we show that groups have maintained a profit edge over stand-alones during the entire 1989–2008 period. We also find that groups have far higher advertising intensity and R&D intensity through the entire period of liberalization during 1989–2008. These findings differ from the predictions of all prior schools of thought about business groups as these prior schools of thought, all for the most part, predicted that there would be an obsolescence of business groups’ profit and resource advantages with liberalization.

The first column above uses PBDITA when calculating ROA, whereas the second uses PBDIT net or NOI and NNRT.

In the last column, the first two cell values are so large because stand-alones were doing only a trivial amount of R&D in those two years. In that same last column, the multiplier suddenly dropped temporarily in 2006 because of one stand-alone making a particularly large R&D investment that year.

Business groups continue to tower over stand-alones in the magnitude of their investment in both marketing-related and R&D-related capabilities (Table 12). The group-capabilities multiplier effect, which measures group affiliates' investment in these capabilities divided by stand-alones' investment, is large through the end of the sample. At the end of the period, groups are still investing 42% more in marketing-related capabilities, and fully 70% more in R&D-related capabilities, than stand-alones.

The prevailing view in the corporate governance literature is that business groups will become obsolete with liberalization and corporate governance reform. Our results suggest, however, that such groups are becoming larger and more diversified. The average number of affiliates per business group, and the average number of industries in which a given business group competed, increased significantly between l989 and 2008. Nor are Indian business groups simply pursuing more unrelated diversification over time. As Table 13 shows, groups encompass more closely related affiliates over time, at least as measured by the average of their average pair-level input-output-table relatedness. Using publicly available data, we calculate a commercial-dependence index for each portfolio of group affiliates. As a proxy measure for commercial dependence, we (1) translate the industry codes from the 2006–2007 Indian national input-output tables into ISIC codes, (2) take the inputs and outputs of a given affiliate pair's two industries, (3) compute the percentage of each industry's inputs to and outputs from the other paired industry, (4) take the highest of those four percentages for each affiliate-pair in a given group, (5) take the average of the group's pairs of affiliates, and (6) take the annual average of that average for all groups. This process generates a country-level relatedness-of-diversification index for Indian business groups for every year between 1989 and 2008. The index serves to indicate whether Indian business groups engaged in increasingly related, or increasingly unrelated, diversification over time.13

Table 13

Indian business groups growing larger and more diversified after liberalization and after market development

Year Number of Groups Average Number of Group Affiliates Average Number of Industries Per Group Industry Relatedness Index based on Maximum Input-Output Industry Dependence
1989 347 2.4 1.9 9.879
1990 374 2.7 2.1 11.271
1991 443 2.9 2.2 10.252
1992 480 3.1 2.3 10.314
1993 520 3.3 2.4 11.351
1994 561 3.5 2.5 11.684
1995 568 3.9 2.8 12.489
1996 568 4.1 2.8 13.085
1997 556 4.2 2.9 12.813
1998 555 4.4 3.0 13.027
1999 568 4.6 3.0 13.003
2000 564 4.7 3.1 12.559
2001 564 4.6 3.0 12.399
2002 562 4.8 3.2 12.522
2003 558 5.6 3.5 12.707
2004 557 6.0 3.7 13.246
2005 554 6.0 3.7 12.930
2006 550 5.9 3.6 13.312
2007 543 5.9 3.6 13.285
2008 521 5.8 3.5 13.321
Year Number of Groups Average Number of Group Affiliates Average Number of Industries Per Group Industry Relatedness Index based on Maximum Input-Output Industry Dependence
1989 347 2.4 1.9 9.879
1990 374 2.7 2.1 11.271
1991 443 2.9 2.2 10.252
1992 480 3.1 2.3 10.314
1993 520 3.3 2.4 11.351
1994 561 3.5 2.5 11.684
1995 568 3.9 2.8 12.489
1996 568 4.1 2.8 13.085
1997 556 4.2 2.9 12.813
1998 555 4.4 3.0 13.027
1999 568 4.6 3.0 13.003
2000 564 4.7 3.1 12.559
2001 564 4.6 3.0 12.399
2002 562 4.8 3.2 12.522
2003 558 5.6 3.5 12.707
2004 557 6.0 3.7 13.246
2005 554 6.0 3.7 12.930
2006 550 5.9 3.6 13.312
2007 543 5.9 3.6 13.285
2008 521 5.8 3.5 13.321

In this table, we show that Indian business groups actually became larger and more diversified with liberalization. This finding runs contrary to the prediction of nearly all prior schools of thought about business groups, as these prior schools of thought each, for the most part, predicted that groups would become more focused/less diversified with liberalization. At the same time, we also show in this table that Indian business groups have been increasingly diversifying over time into industries related to the ones in their existing portfolios.

The finding that Indian business groups increased in size and breadth by embracing closely related affiliates cannot be explained by the conceptualizations of business groups prevailing in finance and strategy, which foresee obsolescence when market and governance institutions mature. We believe that our view of business groups is unique in its ability to explain these results. If groups exist to take advantage of scale-and-scope economies in the production of knowledge, they should grow in response to economic liberalization and the entry into the market of new investors looking for value-added projects. Given that groups are repositories of production skills, they are in a better position than stand-alones to use outside financial capital and to fund increasingly complex value-added investments.

## 3, Discussion

We believe that the literature on corporate governance needs to be reconceptualized in response to our findings. A thorough understanding of corporate governance requires analysis not merely of governance itself but also of business strategy. The conventional wisdom in the financial literature is that business groups are primarily expropriation devices for their controlling shareholders.14 We find that, in India at least, group affiliates pursue systematically different business activities than their stand-alone counterparts. Because these business activities tend to be more value-creating, business groups pay significantly higher rates of value-added excise taxes into the public coffers than stand-alones.

Moreover, if such groups take on much higher incremental operational costs to advance their strategic activities in response to positive industry shocks, they should derive a smaller incremental profit boost than stand-alones that enjoy the industry shock without the operational costs. Thus, groups should be expected to earn less from the same industry shock, on average, than their stand-alone peers. Yet we find no significant difference in the incremental profit outcome between groups and stand-alones.

It is important to point out what this study does not do. We do not aim to test whether groups are “the good guys” in India. To make such a judgment, one would need to examine not just how firms treat outside investors, but also whether they try to corrupt the political process and whether they respect the rights of workers, the environment, and society at large. We lack the data to arrive at such a judgment, and our study focuses on the relationship between controlling shareholders and outside investors. That said, firms must utilize their resources efficiently to be faithful to their outside investors. From an efficiency point of view, we show that groups appear to reflect the effects of macro-level shocks honestly and to invest more intensely in the kinds of market activities that investors tend to value (technological and marketing capabilities, shown by the strategy literature to be an important source of long-term competitive advantage, and related diversification endorsed by both the finance and strategy literatures).

Our results suggest that subsequent studies should reexamine the phenomenon of propping, whereby the controlling shareholder secretly injects his or her own funds into the firm; propping is the opposite of tunneling. Friedman, Johnson, and Mitton (2003) provide a compelling law- and finance-based argument for why propping might occur. We do not take issue with their logic, but we believe that the extent of propping could be better assessed by determining whether legitimate strategic activities are sometimes mistaken for propping. For example, if Bertrand, Mehta, and Mullainathan's (2002) results were accurate, their negative business group coefficient would mean that Indian group affiliates typically engage in propping whenever a negative industry shock occurs. But even if the business group coefficient were significantly negative (which it is not, as this study shows), there could be legitimate reasons why business groups would experience a negative shock less acutely than stand-alone firms. One such explanation is that groups can cut back on strategic-activity investments to preserve more of their steady-state profit level, whereas stand-alones that trade finished goods have fewer strategic activity costs to cut back on. Thus, different starting points could lead firms to adjust efficiently in different ways. It is important to consider those starting points and efficiency explanations when conducting corporate governance research.

Further research is also needed on the effect of country differences on business group behavior. India has reasonably good governance institutions, but business groups in other emerging economies may neither play the same value-creating role nor pay higher value-added taxes. We hope this study will spark empirical analysis of country differences and the role of business strategy in the evaluation of governance.

## 4. Conclusion

The design of this study was inspired by Williamson's (1988) assertion that efficiency-based and agency-based arguments both elucidate firm behavior, and that the two arguments should be used jointly. It is striking how few efforts to control for, and examine, alternative efficiency-based arguments appear in the corporate governance literature. We thus designed a methodology that simultaneously analyzes efficiency-based and agency-based arguments to gain insight into corporate governance.

The concept of business groups in emerging economies as expropriation devices and perpetrators of bad governance should be reformulated using methodology that incorporates analysis of strategic activities. We have shown that groups have a legitimate reason to profit less from positive industry shocks than do stand-alones: groups typically respond to increased demand and other positive industry shocks by pursuing costly value-creating recombinative activities. Yet there is no significant difference between groups and stand-alones in incremental profits from positive industry shocks, perhaps because groups tend to be governed better than stand-alones. We also find that groups have grown larger and more diversified with market liberalization, and continue to tower over stand-alones in their marketing and technological capabilities. If we are to deepen our understanding of the quality of corporate governance by taking into account these capabilities and the environments in which groups function, we will need to forge stronger links between the fields of corporate governance and strategy.

The authors are grateful for comments and criticisms from Laura Alfaro, Sharon Belenzon, Utpal Bhattacharya, Jim Dana, Kathryn Dewenter, Artyom Durnev, Ann Goodsell, Vit Henisz, Adrianna Lohnes, Jackson Nickerson, Rosalyn Reiser, Richard Siegel, Harbir Singh, Eric Van den Steen, Toni Whited, Yishay Yafeh, Sai Yayavaram, Mark Zbaracki, and seminar and conference audiences at Duke University, Washington University in St. Louis, Northeastern University, George Washington University, the National University of Singapore, and the University of Western Ontario. Research funding was provided by the Harvard Business School Division of Research.

1 We cannot discount the possibility that stand-alones more often underpay their taxes; testing for that possibility is impossible with currently available data. It is also important to note that this study does not evaluate whether Indian business groups are “good guys”; such a value judgment would require knowing how each firm gave to politicians and what social externalities are produced by each firm's business activities. This study focuses squarely on the treatment of outside investors and relative efficiency considerations.
2 In historical versions of Prowess referenced in Bertrand, Mehta, and Mullainathan (2002), this variable was abbreviated as PBDIT. The abbreviation was later changed to PBDITA by CMIE to signify that it represents firms' profits before depreciation, interest, taxes, and amortization. That the profit variable used here and in Bertrand, Mehta, and Mullainathan (2002) are one and the same was confirmed in a communication to the author from CMIE on November 12, 2009.
3 This information was attained from http://www.cbec.gov.in/faq.htm on September 3, 2009.
4 The input-output tables, available only for a few years in the past two decades, were downloaded from http://www.mospi.gov.in/cso_rept_pubn.htm in July 2010. The Indian Ministry of Statistics and Programme Implementation made the data available to the researchers.
5 In finance, operating leverage is a key form of risk. Firms with higher operating risk have higher fixed cost intensity, and gain greater incremental profit from a sudden positive shock while losing far more incrementally from a sudden negative shock. Firms that recombine inputs (generating higher fixed cost intensity during stable periods) see smaller incremental profit from a sudden positive shock because their variable costs rise. Also, few independent stand-alones have costs like plant repair bills or steep peak-time electricity spikes. Thus, stand-alones with lower fixed cost intensity during stable times see higher highs and lower lows during shocks. In the present situation, different strategies led to different cost structures; some firms lack certain cost items completely. This difference elicits differential responses to a sudden increase in demand. Reseller-oriented firms may have higher incremental fixed cost intensity during a shock—because they do not have incremental spikes in variable costs—but this differs from the classic operating leverage story such as Rushmore's (1997). To understand how a firm could have lower fixed asset intensity during a stable period but higher incremental fixed asset intensity during a sudden shock requires examining its strategic choices. The fixed asset intensity comparison between the two sets of firms flips during a shock, and one cannot make sense of a symptomatic difference (operating leverage) without delving into how these firms choose to compete (their business strategy).
6 The sales figures cited in this last sentence and shown in Figure 1 are for presentational clarity not put into 1995 inflation-adjusted values.
13 We hope to examine the relatedness of these Indian groups' portfolios of affiliates via Indian patent citation flows or occupational demography data. Such data are not yet available to academic researchers.
14Jian and Wong (2010) used a methodology partially based on Bertrand, Mehta, and Mullainathan (2002) to examine industry shocks and tunneling at Chinese state-owned firms in 1998–2002, but they did not include firm fixed effects and their sample consisted solely of state-owned firms. Other contributions to conventional wisdom on tunneling and business groups (e.g., Bae, Kang, and Kim 2002, 2008; Baek, Kang, and Lee 2006) rarely if ever included firm fixed effects or a comparison reference set of private stand-alones. It is one thing to show that a nontrivial number of business groups expropriate and quite another to show that groups routinely steal more than stand-alones. Bertrand, Mehta, and Mullainathan (2002) continued as of mid-2011 to represent the state-of-the-art in the literature on tunneling and private-sector business groups.

## References

Bae
GS
Cheon
YS
Kang
JK
Intragroup Propping: Evidence from the Stock-price Effects of Earning Announcements by Korean Business Groups
Review of Financial Studies
,
2008
, vol.
21
(pg.
2015
-
60
)
Bae
KH
Kang
JK
Kim
JM
Tunneling or Value Added? Evidence from Mergers by Korean Business Groups
Journal of Finance
,
2002
, vol.
47
(pg.
2695
-
40
)
Baek
JS
Kang
JK
Lee
IM
Business Groups and Tunneling: Evidence from Private Securities Offerings by Korean Chaebols
Journal of Finance
,
2006
, vol.
61
(pg.
2415
-
49
)
Bebchuk
LA
Weisbach
MS
The State of Corporate Governance Research
Review of Financial Studies
,
2010
, vol.
23
(pg.
939
-
61
)
Bertrand
M
Duflo
E
Mullainathan
S
How Much Should We Trust Differences-in-differences Estimates?
2002

Working Paper No. 8841, National Bureau of Economic Research
Bertrand
M
Mehta
P
Mullainathan
S
Ferreting Out Tunneling: An Application to Indian Business Groups
Quarterly Journal of Economics
,
2002
, vol.
117
(pg.
121
-
48
)
Caves
RE
Porter
ME
From Entry Barriers to Mobility Barriers: Conjectural Decisions and Contrived Deterrence to New Competition
Quarterly Journal of Economics
,
1977
, vol.
91
(pg.
241
-
61
)
Dharmapala
D
Khanna
V
Corporate Governance, Enforcement, and Firm Value: Evidence from India
2008

Working Paper 08–005, University of Michigan Law & Economics, Olin
Friedman
E
Johnson
S
Mitton
T
Propping and Tunneling
Journal of Comparative Economics
,
2003
, vol.
31
(pg.
732
-
50
)
Jian
M
Wong
TJ
Propping Through Related Party Transactions
Review of Accounting Studies
,
2010
, vol.
15
(pg.
70
-
105
)
Johnson
S
LaPorta
R
Lopez-de-Silanes
F
Shleifer
A
Tunneling
American Economics Review Papers and Proceedings
,
2000
, vol.
90
(pg.
22
-
27
)
Morck
R
Wolfenzon
D
Yeung
B
Corporate Governance, Economic Entrenchment, and Growth
Journal of Economic Literature
,
2005
, vol.
43
(pg.
657
-
722
)
Nilkamal Limited
Annual Report 2007–08
,
2007–08
Mumbai, India
Nilkamal Limited
Porter
ME
Competitive Strategy: Techniques for Analyzing Industries and Competitors
,
1980
New York
Free Press
Porter
ME
Competitive Advantage: Creating and Sustaining Superior Performance
,
1985
New York
Free Press
Porter
ME
What Is Strategy?
,
1996
, vol.
74
(pg.
61
-
78
)
Rushmore
S
The Ups and Downs of Operating Leverage
Lodging Hospitality
,
1997

January, 8
Siegel
JI
Can Foreign Firms Bond Themselves Effectively by Renting U.S. Securities Laws?
Journal of Financial Economics
,
2005
, vol.
75
(pg.
319
-
59
)
Supreme Industries
Annual Report 2007–08
,
2007–08
Mumbai, India
Supreme Industries
Williamson
OE
Corporate Finance and Corporate Governance
Journal of Finance
,
1988
, vol.
43
(pg.
567
-
91
)