A. An Illustrative Model of Firm Investment and the RER with Financial Frictions

Consider a tradable goods producing firm in a small open economy. The firm is a price-taker on international markets, with its foreign price given at P* normalized at 1, such that, by the law of one price, the domestic currency price of the tradable good is P^T = eP* = e, with e being the nominal exchange rate. Demand structure of the economy is Cobb-Douglas in tradable and non-tradable goods, so that the overall price index is given by:

where P^N is the price of non-tradable goods and thus the RER, or relative price of tradable goods, is given by $RER=\frac{e}{P^N}$. For simplicity, assume also that labor supply is perfectly elastic (e.g., due to linear preferences in the extensive margin of labor), so that the real consumption wage is fixed, i.e., $\frac{W}{P}=\overline{w}$.

Profits and investment of tradable goods producing firm:

The firm lives for two periods. In the first period, it is endowed with a fixed level of capital, normalized to 1, and it maximizes profits by choosing labor input, taking prices and wages as given. At the end of the first period, the firm can invest in its capital stock for the second period using its accumulated profits and/or external borrowing. In the second period, the new capital level is attained, the firm re-optimizes labor input, and pays any internal and external user cost of capital.

The firm produces using the Cobb-Douglas technology:

The firm maximizes its net present value of real profits (measured in units of tradable goods) as follows:

where w denotes the product wage $\frac{W}{e}$ and profit in the first period is given by ${\mathrm{\Pi }}_1=L_1^{\alpha }-w{L}_1$.

Capital adjustment costs are quadratic:

and the cost of external finance increases with the degree of leverage:

The opportunity cost of investing equity (accumulated profits) is assumed constant, normalized to zero, so that C′(.) effectively measures the wedge between external and internal finance. Note that the financing cost schedule is indexed by f, the degree of financial frictions, so that both C′(., f) and C ″(., f) > 0 are increasing in f. Since the cost of internal finance is lower than that of external finance, firms will always exhaust their internal funds before borrowing externally. This specification for the cost of external finance and its relation to financial frictions follows a long line of literature (see, e.g., Fazzari, Hubbard, and Petersen, 2000).

The firm’s first-order condition with respect to labor demand in both periods is given by:

Writing out the product wage in terms of the exogenous consumption wage and real exchange rate, $w=\frac{\overline{w}}{RE{R}^{1-\gamma }}$, we obtain that labor demand and profits increase with a real depreciation (i.e., an increase in the RER), and the elasticities with respect to the RER are given by:

Note that labor demand and profit are more sensitive to changes in the RER the higher is the labor share in production α and the higher is the non-tradable share in consumption (1 − γ). The former relation holds because a higher labor share implies that the wage bill is a large share of revenues, so that a given change in the product wages triggered by an RER change has a larger impact on the firm’s profits. The second relation holds because a larger non-tradable share implies a higher sensitivity of the relative price of tradable goods with respect to the RER.

The first-order condition with respect to capital level in the second period is:

where the return on capital function is given by:

and the optimal capital level is achieved when the marginal return on capital is equal to the marginal cost:

Note that F ′(.) > 0 for plausible range of parameter values and F ″(.) < 0 so that a unique solution exists. Furthermore, by the implicit function theorem, we can derive the sensitivity of capital with respect to available internal funds as:

due to convexity of external financing costs and diminishing marginal returns on investment. As in Fazzari, Hubbard, and Petersen (2000), capital investment is more sensitive to internal funds if the cost of external finance increases more steeply, and if the marginal return to investment decreases more slowly.

Proposition: A real depreciation increases investment and capital of the tradable goods producing firm. That is,

Moreover, the elasticity of investment with respect to the RER is larger the higher the labor share α, and the larger the financial frictions embodied in C″(.). That is,

Proof: See Appendix A.

While the algebraic proof is deferred to Appendix A, the intuition of these results can be sketched out as follows. The larger the labor share in production and hence the payroll share in firms’ revenues, the more will profits increase with a given real depreciation that reduces the product wage. Furthermore, for a given labor share of the firm, the steeper the external borrowing cost the firm faces, the more sensitive is its investment spending with respect to a given change in profit.

We illustrate the main theoretical predictions of the model in the three diagrams below. Figure 1 depicts the optimality condition given by equation (1) which pins down the optimal level of capital. In the absence of financial frictions, when the marginal cost of borrowing is constant and equal to the opportunity cost of internal financing, here C(.) = 0, then the optimal capital level for the firm is $\overline{K}$. If the firm does not have any cumulated profits and must finance all investment externally, the capital level chosen would be K₀, where marginal cost of capital equals its marginal return. Availability of internal funds of the amount Π allows the firm to finance a higher level of capital K₁, which is the level of finance where the optimality condition (1) holds for Π> 0.

View Full Size

Internal funds allow firms to expand investment when facing financial constraints

Citation: IMF Working Papers 2017, 183; 10.5089/9781484313749.001.A001

Notes: F’(K)=marginal return to capital schedule; B=K-Π is the external financing need; C’(B,f)=marginal cost of borrowing, expressed as spread over the marginal borrowing cost under perfect financial markets (normalized at zero); $\overline{K}$=level of capital under perfect financial markets; K₀ =level of capital under frictional financial markets and no internal finance (Π=0); K₁ =level of capital under frictional financial markets and internal financing capacity Π.

• Download Figure
• Download figure as PowerPoint slide

Figure 2 illustrates the effect of a real depreciation on optimal investment. A real depreciation increases profits from Π to Π′, and hence increases the capital level from K₁ to K₂, all else equal, as the financial constraint on investment is internally relaxed (internal financing channel). At the same time, if the depreciation is expected to be permanent (or long-lasting), it also shifts the marginal return schedule F ‘(.) outward (see equation (2)), further increasing capital from K2 to K3 (competitiveness channel). The magnitudes of these investment shifts depend on the firm’s labor intensity. As shown in the proposition above, profits increase more, and investment is boosted through the internal financing channel more, the higher is the labor share, α.

View Full Size

A depreciation boosts profits and investment through the internal financing channel

Citation: IMF Working Papers 2017, 183; 10.5089/9781484313749.001.A001

• Download Figure
• Download figure as PowerPoint slide

Figure 3 illustrates the differential effect of a given increase in profits on investment for different levels of financial frictions. A firm that increases profit from zero to a given amount Π′ in the relatively low financial friction regime f will increase its capital by ΔK_f. The same firm, with the same production technology and hence the same F′(.) schedule, will increase its capital level by more given the same increase in profits if it faces larger financial frictions f’: ΔK_f′ > ΔK_f.

View Full Size

The effect of profits on investment is larger the more financial constraints bind

Citation: IMF Working Papers 2017, 183; 10.5089/9781484313749.001.A001

• Download Figure
• Download figure as PowerPoint slide

B. Effects on Non-Tradable Sector Firms

The theoretical analysis yields unambiguous predictions for tradable sector firms. These predictions go through in a more complex environment where firms have price-setting power, if there are general equilibrium effects from the demand side, and possible reallocation effects across tradable and non-tradable sectors. By contrast, the predictions for non-tradable goods producing firms are less clear. A real depreciation reduces the relative price of non-tradable output and, all else constant, increases the product wage of non-tradable producers, leading to a contractionary supply effect. However, as non-tradable firms typically have more market power because of less competition, they can pass on some of the increase in product wages to consumers. Moreover, non-tradable outputs are used heavily as intermediate inputs in tradable production (di Giovanni and Levchenko, 2010), which means that an expansion of the tradable sector may increase demand for non-tradable goods (Chen and Dao, 2011). In addition, higher income in the tradable sector may increase demand for non-tradable goods. These general equilibrium effects can act toward mitigating the negative supply effect of a real depreciation on non-tradable firms. Overall, we expect the net effect of a real depreciation on non-tradable firm growth to depend on factors such as input-output linkages with the tradable sector, and be therefore negative or ambiguous, in contrast to the prediction for tradable sector firms.

C. Testable Hypotheses

Our model illustrates the internal financing channel by showing how real depreciation enables tradable sector firms with higher labor shares and facing greater financial frictions to growth faster than other firms. This mechanism operates through the impact of a real depreciation on producer real wages and hence labor costs, profits, and investment. Our main testable hypotheses are that, for a given real depreciation, (i) tradable sector firms with higher labor shares have higher profits and cash savings; and (ii) tradable sector firms that face higher financial constraints, as they do for instance in less financially-developed countries, have higher investment than other tradable firms. For non-tradable firms, we expect the effects to be negative as the supply effect operates symmetrically to tradable firms. However, the general equilibrium effects discussed earlier may overwhelm the negative supply effect, leading to smaller, statistically insignificant, or even positive effects in the data. Non-tradable sector firms thus constitute a useful subsample for placebo tests.

A. Firm-level Data

We use balance sheet and income statement data from Thomson Reuters Worldscope for non-financial firms over the 2000-2011 period. Worldscope is a comprehensive data source for publicly listed firms around the world. Market capitalization coverage as a share of the S&P Global Stock Markets Factbook exceeds 90 percent for advanced economies and European emerging markets and is almost 70 percent for non-European emerging markets. The choice of this data source is in part guided by our need to control for firms’ investment profitability, typically measured by Tobin’s Q, for which market value is required. In the sample of firms with non-missing asset information, there are 25,416 tradable sector firms from 66 countries and 5,774 non-tradable sector firms from 63 countries.⁵ The median tradable sector firm in the dataset has total assets of USD 114 million and capital stock of USD 47 million. The median firm has 515 employees.

The countries with largest coverage are the US (accounting for 23.2 percent of firms), Japan (12.2 percent), China (7.7 percent), and the UK (6.6 percent). (See Table 1 for geographical composition of the sample). Together these four countries cover half of the sample. Other emerging markets that contribute at least 2 percent of the sample include South Korea, India, and Malaysia. For each firm we have basic information such as balance sheet size, employment (which we use as a proxy for size), capital stock, capital expenditure, market capitalization and debt (on the basis of which we compute Tobin’s Q), and cash flow (used as a proxy for firm profits). Capital expenditure is scaled by lagged fixed assets while cash flow is scaled by lagged total assets.

Intrinsic labor intensity of firms is measured using the average payroll shares of US firms over the 2000-2011 period at the 3-digit level NAICS industry level, computed using data from Elsby, Hobijn and Sahin (2013). This measure allows us to exploit variation in labor shares across industries that are driven by technology and product characteristics rather than by firms’ hiring and investment decisions, which themselves may be driven by profitability shocks or tax regimes. As shown in Figure 4, this measure is positively correlated with median payroll share across firms and is statistically significant at the 5 percent level. The latter is computed from Worldscope using data on payroll costs (salaries and benefits) scaled by firms’ total sales. Since firm-level payroll information is only available for one third of firms and we are interested in measuring intrinsic rather than potentially endogenous current or recent labor shares, we opt for the average US firm payroll share for our analysis.

To compute the RER, we use data on nominal exchange rates (XRAT) and purchasing power parity conversion factors (PPP) from the Penn World Tables Mark 8.1. The RER is given by ln RER = ln (XRAT / PPP) and is comparable across countries and years. To measure the extent to which firms face financial frictions, we employ the private credit-to-GDP ratio from the World Development Indicators database. This is a commonly used measure of financial development as it captures the amount of credit channeled through financial intermediation to the private sector. Table 2 reports descriptive statistics for the regression sample.

B. Empirical Specifications

Using firm-level data, we closely follow the model predictions to test the effects of exchange rate depreciation on firm profits and investment by using several difference-in-difference identification strategies. This approach offers a valuable advantage relative to cross-country regressions, which are more prone to endogeneity concerns. In particular, the exchange rate can be plausibly assumed to be exogenous for the individual firm (as opposed to country or industry), while the panel structure of the data allows us to control for firm-specific unobservable factors. In the first step, we estimate the following equation for the firm’s cash flow:

where the dependent variable is cash flow scaled by lagged total assets, LaborShare_j is an industry-specific labor share (denoted by α in the model), and z_ijct refers to firm size (log-employment). We expect β₁ > 0. All regressions are estimated using Ordinary Least Squares (OLS) and standard errors are clustered at the firm level. That is, we test the hypothesis that an economy-wide depreciation boosts profits more for firms with inherently higher labor shares, as predicted by the model.

In the next step, we use a triple difference-in-difference approach to test the prediction for investment:

where the investment ratio is given by capital expenditure scaled by lagged fixed assets and is log-transformed. Credit / GDP_ct−1 is the lagged ratio of private credit to GDP, so that (−1×Credit / GDP_ct−1) becomes a country-level measure of financial frictions (equivalent to f in the model), such that larger values indicate less financial depth. We expect a given depreciation to boost investment more the larger is the firm’s labor share (i.e. β₃ >0), and the less access it has to external finance, (i.e. β₂>0). Z_ijct is a vector of firm-level control variables which comprises lagged firm size to control for initial conditions. Importantly, we control for investment profitability by using average Tobin’s Q, defined as the sum of total market value plus debt divided by total assets. But since this may be a noisy proxy for investment profitability, we supplement it with the firm’s lagged sales growth as another predictor of return to capital (see Gilchrist and Himmelberg, 1995). The coefficient on the first two interaction terms therefore measure the differential effect of depreciation on investment through the internal financing channel only, for any given degree of future profitability/competitiveness. Importantly, the (first) triple-interaction term captures how a given boost to profit is used to fund investment more in an environment with less developed financial markets.

Finally, we expect higher fixed capital investment to translate to faster firm growth. That is, we estimate the same equation for total asset growth:

where AssetGrowth_ijct is total asset growth for firm i in industry j in country c in period t.

In all specifications, we use a 25-industry classification from Worldscope that is analogous to a standard two-digit industry classification. The regressions include firm fixed effects (α_i), industry-year effects (η_jt), and country-year effects (δ_ct), with the latter subsuming the interaction term that varies at the country-year level (ln RER_ct × Credit / GDP_ct−1). Country-year fixed effects allow us to control for all macroeconomic factors affecting firms within a country-period (such as terms of trade shocks, fiscal policy, and external demand) and thus reduce the possibility of omitted variable bias. Industry-year fixed effects control for unobserved industry-level technology and demand shocks.

A. Stylized Facts: Firm Growth and the Real Exchange Rate

We start by documenting stylized facts regarding the relationship between the RER and firm growth. We explore the unconditional correlation between the RER and firm performance, measured by asset, market capitalization, and capital expenditure growth, during periods of real appreciation and depreciation. Figure 5 shows the median values for these firm outcomes in years when the currency is either depreciating or appreciating, separately for tradable sector firms from advanced economies and emerging market countries.. In Table 3 we also report non-parametric tests of equality of medians for all the variables shown in the figure.

Figure 5 and Table 3 confirm our prior that real depreciations are associated with higher investment and growth for tradable sector firms. However, this correlation does not account for other factors simultaneously affecting firm performance, nor does it pin down the precise mechanism through which the positive correlation between real depreciation and higher firm growth arises. In the next section we exploit firm heterogeneity in labor intensity and financing constraints to provide econometric evidence supporting one channel behind this relationship—the internal financing channel. Importantly, our analysis will not allow us to estimate the economy-wide effect of the RER on firm growth, but rather to identify a causal channel from real depreciation to corporate investment which holds regardless of the sign of the overall effect.

B. Baseline Results

Our main results for tradable goods producing firms are reported in Table 4. We show two variants of each specification, one in which we control for country-year fixed effects that absorb macroeconomic developments affecting all firms in a given period (including changes in the RER) and firm fixed effects; and another one that also control for industry-year fixed effects. We find that real depreciation raises balance sheet growth for firms with higher labor shares (columns 1-2). For firms with the same labor share, growth is higher in countries where firms face greater financial frictions, here proxied by lower credit-to-GDP ratios. These differential effects arise above and beyond the direct competitiveness effect of the RER on firms’ profitability, as the estimated investment response is conditional on Tobin’s Q and sales growth.

Both coefficients of interest are statistically significant at conventional levels and conform to our theoretical predictions. They are also economically significant. The estimates on the double and triple interaction terms (0.413 and 0.293 in column 2) indicate that, in countries with low financial development (10th percentile of the credit-to-GDP distribution corresponding to Turkey), a real depreciation of 10 percent is associated with an asset growth rate that is 1.8 percentage points higher for high versus low-labor share firms (90th vs. 10th percentile of the labor share distribution, corresponding to the difference between wood products and mining), whereas the same differential is close to zero in countries with high financial development (90th percentile of the credit-to-GDP distribution corresponding to the Netherlands). Alternatively, for high-labor share firms, a real depreciation of 10 percent is associated with an annual asset growth rate that is 3.7 percentage points higher in a country with low versus high financial development, whereas the same differential is only 1.6 percentage points for low-labor share firms. Given a median asset growth rate of 7.2 percent per year, these magnitudes are economically significant. Note that the cross-country difference in credit-to-GDP ratios, e.g., between the Netherlands and Turkey, could reflect differences in institutions, culture, or stage of development as deeper causes for financial depth. However, as long as such differences contribute to the wedge between external and internal finance for firms, they too are valid treatment variables for the differences-in-difference approach.

The regressions for fixed capital investment, shown in columns 3-6 of Table 4, yield similar results to those for total asset growth. Once again we find that a real depreciation increases the capital expenditure of firms with relatively higher wage bill, and even more so that of labor-intensive firms facing larger financial frictions. The coefficient signs are consistent with our theoretical predictions and hold up to controlling in a standard investment specification for Tobin’s Q, sales growth, and the firms’ capital stock.

In columns 7-8 of Table 4, we regress cash flow (scaled by lagged total assets) on firm size and the interaction of the RER with labor intensity. We find a positive and statistically significant coefficient, consistent with model prediction. The estimates indicate that, for the same RER depreciation, firms with a higher labor share have higher cash flow than other firms, which is consistent with the notion that depreciations depress real wages and increase the profits of firms that rely relatively more on labor, boosting their internal funds.

We next turn to results for non-tradable sector firms, which are a useful placebo test for our main theoretical predictions. For firms producing non-tradable goods, our conceptual framework predicts negative coefficients as the supply effect of a real depreciation operates symmetrically to tradable sector firms. The results, shown in Table 5, suggest that the effects on investment and cash flow of a real depreciation for non-tradable sector firms through the same internal financing channel are either negative or statistically insignificant. These findings confirm our prior that the internal financing channel operates in the opposite direction or is not present for non-tradable goods producing firms.

C. Implications for Firm Value and Stock Returns

An interesting and relatively understudied aspect in empirical investment literature is how shifts in firms’ internal financial resources which alter their investment behavior affect the market value of the firm. Boosts to cash flows (identified here by differential exposure to exchange rate variations) can either help firms overcome funding shortages and realize productive investment opportunities, or lead to wasteful projects associated with “empire building” and other perks (see detailed discussion in Tirole, 2006). One way to empirically discriminate between these two stories of under-investment (due to financial frictions) versus over-investment (due to agency conflict) in the context of our paper is to examine the effect of the internal financing channel on market capitalization and stock returns. If the rise in investment we estimated above is indeed driven by our internal financing channel as predicted by the model, then the internal funds should be directed towards productive projects. If so, the internal financing boost triggered by a depreciation should increase the market’s assessment of the firm’s value relatively more in labor-intensive industries and underdeveloped financial markets.

In Table 6 we present the results of this exercise. First, as shown in columns 1-2, a depreciation provides a stronger boost to sales growth in industries that are more labor intensive, and for a given labor intensity, more so in countries with less developed financial markets. This result suggests that the rise in cash flow is indeed directed toward increasing the productive capacity and performance of the firm. This positive effect on sales is evident in the current as well as subsequent year, suggesting a sustained benefit of higher productive capacity. Consistent with stronger sales, we also see that market capitalization and stock market returns increase relatively more in labor intensive industries and financially underdeveloped countries, both contemporaneously and after one year. These results strengthen the empirical relevance of our mechanism as market data provide an external validation of the baseline investment response.