![Sample our Global Development journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5929/TOC-Subject-Sample-2021-Global-Development.jpg"})
Abstract
This paper examines the importance of finance constraints for firm investment expenditures by looking at the investment-asset sales sensitivity in financially healthy Indian manufacturing firms. Voluntary asset sales is a cleaner indicator of firms' liquidity than cash flows since it is unlikely to influence firms' growth opportunities unless they are financially constrained. We take care of the endogeneity and the implicit monotonicity problems, which are much debated in the literature, by using an endogenous regime switching regression model. We find that the investment-asset sales sensitivity is significantly greater for firms that are likely to be financially constraints.
Notes
1. See Hubbard (1998) and Lensink, Bo and Sterken (2001) for a comprehensive literature survey.
2. For a detailed discussion on asymmetric information see Stiglitz and Weiss (1981), Myers and Majluf (1984), Jensen and Meckling (1976), Jensen (1986).
3. See Lensink, Bo and Sterken (2001) for a comprehensive reference to such works.
4. Besides the three criticisms mentioned above on which this paper focuses, there are two other issues which have attracted attention in recent years. The first concern relates to the errors in the measurement of growth opportunities of firms which are generally captured by accelerator or Tobin's average q. The problem with accelerator is that it is a backward looking measure. Tobin's average q, on the other hand, is used as the proxy for Tobin's marginal q which is the theoretically justified measure of growth opportunities of firms (Tobin 1969). The need for the proxy arises because Tobin's marginal q is not observable as there is no direct measurement of the price of a capital-hour of service of a given quality (Branson 2005). See Hayashi (1982) for conditions under which Tobin's average q can equal the Tobin's marginal q; Whited (1992), Gilchrist and Himmelberg (1995), Bond et al. (2003) and Cummins, Hassett and Oliner (2006) for alternatives of Tobin's q-investment specification; and Erickson and Whited (2000) for the influence of measurement error in Tobin's average q on investment cash flow sensitivities. The second concern relates to the irreversible nature of investment expenditure, i.e. the initial cost of investment is at least partially sunk. This implies that if there is uncertainty over the future returns from the investment, its timing can be tuned to get more information. Empirically, this can lead to lumps and bumps in the investment behaviour of firms. See Dixit and Pindyck (1994), Carruth, Dickerson and Henley (2000) and Bond et al. (2003) for a comprehensive discussion on irreversible investments. In this paper, however, we focus only on the three issues raised in the text. Because of the volume of the other two problems mentioned here, we keep them for future research.
5. Athey and Laumas (1994) use the sales accelerator to measure growth opportunities of firms whereas Athey and Reeser (2000) supplement Tobin's average q along with the accelerator to capture growth opportunities of firms.
6. Some of the studies in the Indian context analyse the role of financial variables without dividing the sample of firms into constrained and unconstrained groups. For example, Rajkumar (2005) investigates public listed Indian manufacturing firms in the period 1988–89 to 1998–99. Using panel data estimation, the author finds that while demand is the only significant factor in debt financed firms; investment behaviour of equity financed firms, in addition, depends on internal liquidity and investment opportunities also. Nayak and Kulshreshtha (2007) in their study of large public limited companies from 1969 to 1999 find that lagged output change, lagged capital stock, lagged change in cost of capital and current public investment has a positive impact on private corporate investment whereas current change in cost of capital has a negative impact on the gross investment of firms. Moreover, using the Chow test they also find a structural change in the investment behaviour of the firms in the 1990s. Bhattacharyya (2008) studies two industries Electronics, Electrical Equipment and Cables (EEEC) and General Engineering (GE) with the sample size of 26 (for the period 1991–92 to 1997–98) and 28 firms (for the period 1993–94 to 1998–99), respectively. Using a pooled ordinary least squares (OLS) regression the author finds that though the investment–accelerator relationship is significant, financial strength of firms is important from the point of view of their investment decision.
7. The endogenous regime switching model, in essence, presents the cross sectional differences in the investment behaviour of firms. Various other econometric tools like OLS, panel data models, dynamic panel data models and error correction models have been used in the literature to investigate the role of finance constraints in meeting investment expenditure of firms. However, unlike these models, the endogenous regime switching model, as noted above, allows for endogenous switching of firms from constrained group to unconstrained and vice versa. On contrary, in the other econometric methods, a prerequisite for such estimation is a priori sorting of firms into the two groups. This may cause biased estimates on two counts: first, it implies a restrictive assumption of monotonic relation of finance constraints with the sorting variable and second, firms are restricted to only one regime throughout the sample period. The use of time dummies also as alternatives may not suffice because they control for only those effects which are common across years. Different firms, based on their fundamentals and market strength, may get affected differently even with the same shock.
8. Here and henceforth Tobin's q means Tobin's average q, unless stated otherwise.
9. Variables are defined in the Appendix.
10. We allow up to 1000% jump if manufactured sales is up to Rs. 10 million; 500% if manufactured sales is between Rs. 10 million plus to Rs. 50 million; 300% if manufactured sales is between Rs. 50 million plus to Rs. 100 million; 200% if manufactured sales is between Rs. 100 million plus to Rs. 250 million and; 100% if manufactured sales is above Rs. 250 million. We tried various other numbers but these numbers are chosen to include maximum possible number of observations in the sample and yet putting a restriction on restructuring firms.
11. Excluding distressed observations is important because a marginal rupee from asset sales is less likely to be invested if a firm is in financial distress. Inclusion of distressed firms in the sample would thus bias the results against finding a relationship between firm investment and internal funds. For a detailed discussion on this issue see Fazzari, Hubbard and Petersen (1988, 2000) and Kaplan and Zingales (1997, 2000).
12. For example, Lang, Poulsen and Stulz (1995) considering asset sales worth $1 million and above report mean and median asset sale size of 11% and 9% of total assets, respectively.
13. This is besides the issues of tax compliance and clearance from managing board.
14. We do not interpret the coefficients with the other variables here. Our main estimation results are presented in the next section where we discuss all the results in a greater detail.
15. The framework is standard in the literature. See, for example, Maddala (1983), Hu and Schiantarelli (1998) and Hovakimian and Titman (2006).
16. If σ1s = σ2s = 0, we have a model with exogenous switching.
17. The second equality in Equation (13) uses the fact that a joint density equals the product of conditional density and the marginal density. If β1 is equal to β2 and σ1s is equal to σ2s then u 1 is equal to u 2 and the likelihood function reduces to a standard normal density.
18. The dummy specification of is overwritten in estimating endogenous regime switching model based on a numerical maximization technique.
19. To test the existence of two distinct investment regimes we rely on χ2 tests suggested by Goldfeld and Quandt (1976). The test rejects the hypothesis of similar investment behaviour by the two regimes at 0.1% significance level.
20. We perform χ2 test, as done above, which confirms rejection of similar investment behaviour by the two regimes for the sub-sample at 0.1% significance level.
21. This is standard in the literature. For example, see Fazzari, Hubbard and Petersen (1988).