Spot exchange rate volatility, uncertain policies and export investment decision of firms: a mean-variance decision approach

Abstract

This paper studies characteristics of optimal investment decisions of risk-averse firms who engage in exports under two types of risks: endogenous and background risks. While endogenous risk arises from the fluctuations in spot exchange rate and affects directly the profit of an exporting firm, background risk arises from uncertain changes in firm- and industry-specific domestic and foreign policies. We propose a mean-variance decision-theoretic model to trace out impact of perturbations in the distributions of these uncertainties on the optimal investment strategy. A testable empirical model is derived and applied to a panel of 840 exporting Indian manufacturing firms for the period 1995–2015. Our results suggest that Indian manufacturing exporters depict decreasing absolute risk aversion and that firms’ risk preferences are prone to variance vulnerability.

KEYWORDS:

• Spot exchange rate risk
• Background risk
• Investment decision
• Mark-up estimation
• Risk aversion elasticities

JEL Codes:

• D22
• D81
• F41
• L11

Acknowledgements

The authors are grateful to Chris Adcock (the Editor), an Associate Editor and five anonymous referees, whose constructive comments and suggestions have significantly improved the quality of the paper. The authors are also immensely grateful to Simon Wolfe, Larisa Yarovaya, Nina Pavcnik, Sudipta Sarangi, Amit Batabyal, Sushanta Mallick, Chandan Jha, and the participants in the 23rd Biennial Conference 2019 of the Association of Indian Economic and Financial Studies (IEFS) for valuable feedbacks on the earlier version of the paper. Remaining errors (if any) are the responsibilities of the authors themselves.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 This is the reason to why we have focused ourselves considering a developing economy case.

2 It is worth to mention about the literature on MNEs investing in emerging markets in this context, namely Ferraris (2014); Luo and Zhang (2016); Dang, Jasovska, and Rammal (2020). These researches recognise the importance of firm-level investment in capital deepening, necessary for improving the firms’ performance.

3 We have taken the sample of Indian manufacturing firms as a case study to empirically demonstrate the validity of our analytical results/propositions. In this context, Figure 2 shows that these firms tend to keep positive risk premium, once they extract positive profits from their export markets.

3 It is important to note that our paper does not investigate the aspect of whether or not to enter the export market. Rather it takes export-investment as an indispensable factor to promote intensive margin of exporting, i.e., promoting the firm to cater greater foreign market share than what had been in the previous years. Hence, the decision problem we consider here naturally becomes how much investment a risk-averse decision-maker should make, at the margin, under the presence of both spot exchange rate uncertainty and background risk. The results we have derived are categorised as comparative static responses of how does the optimal investment decision changes, at the intensive margin of exports, due to the ceteris paribus perturbation in the moments of the distributions of exchange rate uncertainty and background risk.

5 Given that we are considering firms of a small, open developing economy, each firm would take the world price of the exportable as exogenously given. Therefore, no firm has any price setting power in the world market. Also, no firm in our model can affect the industry-specific policies. However, it would be interesting to consider a scenario with dependent background risk where the firm can influence the industry-specific and firm-specific policies. Since the present paper is not intending to explore any external or internal policy impact on firm's export investment decisions, this is beyond the scope of the present paper. Nevertheless, it can be taken up as an exciting future research agenda.

6 We would like to clarify that how and why fluctuations in spot foreign exchange rate ($\tilde{e}$) is termed as an endogenous risk, while the background risk ($\tilde{Z}$) is exogenous in our model. Because the impact of export investment on the distribution of end-of-year uncertain profit, $\tilde{\pi }$, directly works through $\tilde{e}$ (given the world price of the exportable, $p$), in line with Eichner and Wagener (2009, 1143–1144), we have termed $\tilde{e}$ as the endogenous risk. On the other hand, background risk is purely based on firm-specific, industry-specific and domestic macroeconomic policy-specific exogenous shocks, on which no firm has any influence. Hence, the background risk purely enters the model as a passive random variable, weighted by a positive scale factor, $\text{β}$.

7 In our main model, we have employed static panel data regression model, using fixed-effects (FE) and Heckman's two-step estimation procedure to correct for sample-selection bias (endogeneity issues). However, for the sake of robustness of our estimated risk aversion elasticities, we have used dynamic panel data regression model, deploying the system-GMM procedure. We have explicitly talked about this in Footnote 22 and the results are available in Appendix A (Tables A.1 and A.2, see supplementary material).

8 The result remains robust for Heckman's estimation procedure.

9 One can always assume any constant returns to scale (CRS) production technology for a firm using two broad inputs: a bundle of labour and intermediate goods ($L$) and invested capital ($I$). Now, it can be easily checked that any CRS production technology, viz., Cobb-Douglas, constant elasticity of substitution or trans-log production function would yield the linear relationship between $Q$ and $I$ for a given level of $L$-$I$ ratio (say the steady-state level). We can also assume that the firm faces a pre-specified interest rate schedule, $r\left(I\right)$, with $r^{\mathrm{\prime }}(.)>0$, $r^{\mathrm{\prime }\mathrm{\prime }}(.)>0$. However, for analytical simplicity, we assume that the cost of investment is linear, viz., $rI$.

10 All random variables are denoted by a tilde, while their realisations are not.

11 As already argued in footnote 5, we are considering firms operating within a small, open, developing economy in a competitive world market, where no firm has the capacity to influence the industry-specific policies/strategies to cope up with the fluctuations in the nominal effective exchange rate.

12 See section 1, where we have already argued for this. Also see Figure 2, which also gives an a ‘priori hunch on this property for the firms in our sample.

13 Please see Broll, Wahl, and Wong (2006); Broll and Mukherjee (2017); Broll, Mukherjee, and Sensarma (2020) as only few of the many relevant contributions who have also justified the validity of this assumption in the context of exchange rate risk.

14 See Davis (1989); Broll et al. (2015b); Broll and Mukherjee (2017) in this context.

15 One very good example in this context is from Bougheas et al. (2018), who showed that exporters whose foreign sales are more competitively priced under a devalued currency responds by keeping a positive premium over cost at the margin and improve their foreign market shares.

16 This is the most robust and standard approach following the literature to estimate output elasticities by assuming a flexible trans-log production function with Hicks-neutral productivity. We would like to cite Dai and Cheng (2018) in this context. Also, please note that such trans-log production function would yield the linear relationship between $Q$ and $I$ for any given level of labour - capital (or investment) ratio.

17 We use deflated sales revenue, capital spending and different input expenditures as proxies for the physical quantities of output, capital and intermediate inputs, respectively, following the literature on productivity estimation. To get the deflated values of sales, compensation to employees, power and fuel expenditure, capital employed, raw material expenditure, we use industry- specific wholesale price indices, keeping 2004 as the base year to accord with the 1995-2017 period covered by our study. All the industry specific-wholesale price indices are obtained from the Economic Adviser, Ministry of Commerce and Industry, Government of India. http://www.eaindustry.nic.in/wpi_revision_0405.asp.

18 To estimate firm-level physical capital stocks for each year we closely follow the methodology adopted by Balakrishnan et al. (2006), which uses perpetual inventory model. At first, we obtain firm-level net investment by taking the difference between the current and lagged values of gross assets less depreciation for each year. Next, by taking the sum of investment in subsequent years for each firm, we obtain the firm-level capital stock for every time period. Moreover, using industry- specific wholesale price indices of Machinery and machine tools and keeping 2004 as the base year to accord with the 1995-2017 period, we obtain firm-level real capital stock for each year by deflating the value of capital stock obtained in the previous step. For more detail of this method see, Balakrishnan et al. (2006) (pp.71-73), and Topalova and Khandelwal (2011) (pp. 23).

19 Note that (4.5) is the differentiated (w.r.t. $m_{it}$) version of the estimated trans-log production function reported in Column 1 of Table 1.

20 We have also performed appropriate Hausman test for the model selection ex-ante, where the ${\chi }^2$. test statistic is significant at 1% level, thereby rejecting the null hypothesis and validating the choice of fixed-effects (FE) model, as opposed to the random-effects (RE) model.

21 Topalova and Khandelwal (2011) have shown that rm size and age (in its non-linear form) (measured as a proxy for experience) play an important role in determining the firm-level productivity over time. In similar line we also control for firm size, age and age² in our model while estimating firm-level Mark-up, which has been derived from firm-level Productivity.

22 We have used a novel approach to address the possible endogeneity problem in firm's decision making on firm-level Mark-up (i.e., potential positive mark-up bias), which could arise due to possible sample selection bias in keeping only exporting firms (i.e., export earnings are positive) in our dataset (See, page 160 of James J. Heckman (1979, 153–161) ‘Sample Selection Bias as a Specification Error,’ for further details). However, we have also performed the conventional Dynamic Panel System GMM estimation (Blundell and Bond 1998) to control for possible trade policy endogeneity (which arises due to reverse causality between last period's firm-level export risk and returns on current period's firm-level mark-up during our study period. The dynamic panel results remain consistent with our main results. As our main objective is to empirically estimate the risk aversion elasticities of the main variables of our theoretical model $({{\mu }_{\pi }}_{it},{v_{\pi }}_{it}and\mathrm{\partial }{v}_{\pi }(I^{\ast }\left)/\mathrm{\partial }I\right)$ rather than examining any trade policy effect on mark-up, we did not provide the results of our dynamic panel (with lags 1, 2 and 3) in the main text. See Appendix A (Tables A1 and A2, see supplementary material) for the detailed discussion on our Dynamic Panel analysis.

23 We have explored an important source of endogeneity, i.e., lagged mark-up (equation, 4.7) which may cause the self-selection behaviour of firms in terms of increasing their mark-ups (i.e., risk premium) in the subsequent period following fewer net export earnings in the previous period.

24 It should be noted that we also extend our analysis further to examine whether the firm-level mark-up is endogenous with its second lag (lag 2). However, the result suggests the absence of endogeneity of firm-level mark-ups at lag 2.

25 In this context, the reader should note that by including all possible fixed effects (Firm, Year and Industry-Year), the R-square has increased considerably. Therefore, results in column 3 are the main results of this paper.