Exchange rate risk and the skill composition of labor

Exchange rate changes have the potential to substantially affect the competitive situation of firms that trade internationally. Large literatures examine links between exchange rates and issues such as prices and price setting currencies [see e.g. Burstein and Gopinath (2014)], investments [see e.g. Goldberg (1993)] and stock market valuation of firms [see e.g. Dominguez and Tesar (2006)]. Not only the level of the exchange rate, but also its variability may affect a firm’s decision making. In a seminal contribution (Stigler 1939) showed that a firm that faces more variability may want to organize production in a way that sacrifices static efficiency to gain flexibility. Stigler’s intuition, together with the observation that more flexible work practices generally require more highly skilled labor [see e.g. Caroli and Van Reenen (2001), Thesmar and Thoenig (2000)] motivate our analysis of links between firm specific exchange rate variability and the skill composition of the labor force in these firms.

We use matched employer–employee data covering the universe of private sector Swedish firms for 1998–2014 inclusive. These data are matched to firm-level customs data that include observations on each firm’s imports and exports across currency areas. With this rich data, we derive a risk index that measures each firm’s exposure to exchange rate shocks, which is a function of the firm’s trading activities across currency areas, and the magnitude of the variability of the exchange rate. Firms differ in their international composition of imports and exports and this gives us variation in exchange rate risk at the firm level: the risk index varies between firms and within firms over time. We use this variation to identify the effect of risk on skill composition at the firm level.¹

As a first test of the relationship between risk and skilled labor, we study firm-level effects within narrowly defined industrial sectors. Effects are positive throughout and non-trivial in magnitude when we consider the full population of firms with at least some international trade in the sample period. We find substantial effects for the tercile of firms with the highest trade share: for these firms, the estimated elasticity of a firm’s currency risk exposure from net trade on the firm’s share of skilled labor is 0.06. For manufacturing firms, the corresponding elasticity is 0.05. Given the variation of our measure of exchange rate risk, this suggests that a firm’s net exchange rate exposure has an important effect on the share of skilled labor employed by these firms. For example, for a given level of trade intensity, a manufacturing firm switching its exports and imports from Germany to the United States faces a 130% increase in exchange rate risk and would employ 7% higher share of labor with post-secondary education.²

Identification of these first estimates relies on variation in the risk index and skill levels between firms as well as within firms over time. However, we observe that most of the variation in risk and skill is between firms, which suggests that the estimated effects are derived mainly from between-firm variation. This evidence is consistent with a view that the decision on what skill level to use, and what currency areas to trade with, are long run in nature with significant sunk costs. Even with a rich set of controls and high-quality register data a causal interpretation of the estimates partly derived from between-firm variation in risk should therefore be undertaken carefully.

We therefore proceed to a specification that relies on within-firm variation in exchange rate risk to identify effects. To account for the time that it can take for firms to adjust their skill composition, given the relatively strong labor protection laws in Sweden, we examine 4-year periods. For the top tercile of trade intensive firms the effects are marked with an estimated elasticity of the share of high-skilled labor with respect to exchange rate risk of 0.07 and the associated increase in exchange rate risk associated with fully switching exports from Germany to the U.S. this would imply an increase in the share of high-skilled by some 9%. For an average firm with around 130 employees this is equivalent to replacing three low-skilled workers with three high-skilled workers. We parse the population into subsamples and find that the effect is particularly strong on the most trade intensive firms with fewer than 50 employees with an estimated elasticity of 0.1, or by some 13% for the thought experiment of fully switching from Germany to the U.S. as a trading partner.

To the best of our knowledge we are the first to examine the links between exchange rate risk and the skill composition of labor. In a related study (Kaiser and Siegenthaler 2016) examine the effect of a first moment exchange rate shock on the composition of skilled labor at the firm. Using rich data from Switzerland they establish that an appreciation of the Swiss franc is associated with a higher share of skilled labor. In contrast, we study skill-biased organizational change under second moment shock. Nonetheless, we also identify the same effect from first moment shocks to Swedish firms.

Our contribution is to establish a distinct effect of exchange rate risk on skill composition. We relate to a large literature that examines links between labor demand and a firm’s international engagement [see e.g. Bandick and Hansson (2009), Grossman (2013), Davidson et al. (2017), Malgouyres and Mayer (2018)]. Although are focus is on the role of risk, our results fit well with this literature. For instance we see that firms with a higher export share are more skill intensive.

In the next section we outline the theoretical foundations for the predicted effects and discuss the related literature. Section 3 describes the data and the risk indices that we compute. We discuss the empirical specification in Sect. 4 and then turn to results in Sect. 5. Section 6 concludes.

In the introduction we posited that risk creates a value to being flexible, which in turn increases the value of a highly skilled/well educated work force. The reasoning is nontrivial and in the following we lay the theoretical foundation for the mechanism we study and review previous empirical evidence.

A first conceptual issue is that the terms “risk” and “uncertainty” are at times used as distinct concepts in the literature and at times used interchangeably. We will use the term risk and give it a broad interpretation, letting it capture a lack of knowledge of the realizations of future states of the world. This same broad interpretation is manifested for instance in the regulations from the SEC that govern how U.S. firms are to discuss “Risk factors” in Section 1A of their 10-K statements [see e.g. Friberg and Seiler (2017)].³

2.1 Flexibility as a way of dealing with exchange rate risk

A depreciation of the real exchange rate tends to both raise export revenue and raise the cost of imported inputs. Exchange rate variations may thus translate into profit variations for a firm. Firms may respond to expectations of such future variability in many ways.⁴ For example, investing in state dependent contracts to create an offsetting exposure (e.g. using financial deratives to hedge some exposure), modifying exposure (for instance engaging in “natural hedging”, trying to match currency inflows and outflows) or building buffers to limit consequences of adverse shocks [for instance precautionary cash holding as in e.g. Bates et al. (2009) or keeping greater inventory].⁵ These methods for managing risk are not mutually exclusive.

In addition to these examples, and in line with the focus of this paper, firms may adapt their degree of flexibility. The strategic management literature highlights the importance of flexibility in adjusting to changing conditions [see e.g. Teece et al. (1997), Sull (2009) or Collins and Hansen (2011)]. Also the economics literature has examined flexibility as a way to respond to risk and the formalization by Stigler (1939), where greater flexibility is associated with a flatter average cost curve, remains a seminal reference. In this formalization flexible firms have a flatter average cost curve allowing them to at a lower cost respond to positive demand shocks. Figure 1 illustrates, in the simplest terms, the intuition for this mechanism but graphing profit functions rather than cost functions. The left panel illustrates the situation for an exporter with only domestic costs. rfx denotes the real exchange rate in domestic to foreign currency units.

Profits increase as the domestic currency depreciates against the foreign currency. Assume that the exchange rate is stochastic and has an expected value of E(rfx). Consider first the case of a “benchmark” technology where profits increase linearly with the exchange rate. Assume that this is the production technology that maximizes profits if the exchange were equal to its expected value, E(rfx). As profits in the benchmark case are linear in the exchange rate, expected profits are unaffected by a mean preserving spread of the exchange rate.

Now consider an alternative technology that has lower profits if the exchange rate has a realization close to the expected value but that has higher profits in the tails of the exchange rate distribution. We may call this a flexible technology—it sacrifices some profits in the case of small shocks for greater profits in the case of either large negative or large positive shocks. As with any convex function its expected value increases if we take a mean preserving spread of the exchange rate.

Stigler’s realization is that if there is substantial risk, e.g. a sufficiently high probability of either high or low exchange rates, then it may optimal for the firm to invest in flexible production.

Open image in new window

The right hand panel of Fig. 1 illustrates the corresponding case for a firm that only sells domestically but uses imported inputs. Here the exchange rate exposure is negative: As the real exchange rate depreciates, imports become more expensive and profits decrease. An importer, just as an exporter, can increase expected profits by being more flexible. The two panels highlight that exchange rate risk is likely to operate both as demand and as cost shocks.⁶

Firms that face sufficiently high risk value flexibility, but what strategies can the firm deploy to operationalize flexibility? We may think of flexibility as the ability to respond quickly or at lower-cost to changes in the state of the world.⁷ Choosing a production function with higher fixed costs but lower marginal costs as suggested by Stigler (1939) is one way of being flexible but there are a variety of strategies available to the firm that operate through the different components of the profit function: price, demand and costs.

Flexibility through pricing can for example be achieved by greater price discrimination [see e.g. Friberg (2001)] or by lower price adjustment costs. On the demand side, the customer segments that a firm targets, and the degree of product differentiation, can also be related to flexibility [see e.g. Chang and Harrington Jr (1996) for an analysis]. On the cost side, flexibility can be achieved by choosing the appropriate number and location of suppliers of intermediate inputs [a large literature in supply chain management explores this strategy, see e.g. Esmaeilikia et al. (2016) and Antras et al. (2017) for evidence at the source country level] or adjusting capital structure [for instance leasing rather than owning equipment as explored in Gavazza (2011)]. The fact that a firm strives to become more flexible can also affect its demand for labor, an issue that we turn to now.

We use matched employer–employee data on the universe of Swedish firms. Both manufacturing and service firms are included. We limit attention to firms with at least 10 employees and furthermore limit attention to firms that export or import at least 1 year in the sample. The main source of the data is Sweden’s official statistics agency, Statistics Sweden. The data are yearly and our data cover the period 1997–2014 inclusive. The main regression analysis uses one yearly lag for some independent variables and thereby these regressions rely on data from 1998 to 2014. We observe firm \(\times\) country exports and imports and also use employee data and capital from Statistics Sweden. Capital is defined as the sum of machines, inventory and tangible capital. We also observe industry classification at the 5-digit level of the Swedish industry classification (SNI 2007). At the 4-digit the level the classification is analogous to NACE Rev 2. The 5-digit level is quite detailed and for instance the 4-digit NACE industry 1711 (manufacture of pulp) is further divided into three 5-digit industries, manufacture of sulphite pulp, of sulphate pulp and of mechanical pulp.

As our measure of skill level we use the percentage share of the total workforce in a firm that has at least some post-secondary education. Sweden has 9 years of mandatory schooling, followed by secondary education which consists of 2 or 3 years of vocational or academic schooling, which is followed by the majority: 78% of low-skilled workers have completed secondary education.¹² We refer to those with at most a degree from secondary education as low-skilled. Those with post-high secondary education are counted as high-skilled. This group includes not only those with a bachelor degree or higher, but also those that have studied at least one semester at university or those that have followed post-secondary vocational training. The Swedish classification of educations follows the ISCED 1997 standard from year 2000 onward and individuals with education levels 4, 5 and 6 in this nomenclature are counted as high-skilled. Due to changes in the classifications in year 2000 data from before this are not directly comparable to later data. In the benchmark regressions that use 1998 as their first year this is a limited concern because industry \(\times\) year fixed effects are likely to capture the change (robustness checks for the post-2000 period confirm that this is not important for results). The change in classifications is more of a potential concern in firm fixed regressions and in these we examine the post 2000 period only.

Employees are linked to employers via wage bills in the tax records and there is no direct distinction between permanent (on open-ended contract) and temporary (on fixed-term contract) employees and the latter are therefore included in e.g. our calculated share of high-skilled. Swedish labor laws permit temporary employment for up to 2 years, after which the employment becomes permanent unless there are specific circumstances. Statistics Sweden reports on various issues related to temporary employment in their labor force surveys, which covers a random sample of around 20,000 individuals per month. Based on these surveys around 15% of employment is temporary during our period of study. A little less than a third of temporary employees are substitutes (generous parental leave of more than 1 year per child is one important reason for substitutes) and overall the evidence suggests that temporary employment is used both for reasons of flexibility and as way for employers to screen employees, thus acting as a gateway for permanent employment, see e.g. Berglund et al. (2017) for further analysis.

We combine the data from Statistics Sweden with quarterly data on nominal exchange rates and consumer price indices from IMF’s International Financial Statistics. For each trading partner we calculate the real exchange rate, expressed as the real Swedish krona (SEK) price of foreign currency.¹³ A higher value for an exchange rate is thus associated with a real depreciation of the SEK. All real exchange rates are normalized by their mean value and the real exchange rate vis-à-vis country a in year t is denoted by \(rfx_{at}\). We use \(rfx_{at}\) to calculate a measure of volatility for each currency area against the SEK. To do so we calculate standard deviations at the quarterly level for a 3 year rolling window and then take the average in year t for currency area a as a measure of the risk, denoted by \(\sigma _{at}\).¹⁴

The key variable of interest is the measures of the firm \(\times\) year specific exchange rate volatility where both gross (\(\sigma ^{G}_{it}\)) and net volatility (\(\sigma ^{NET}_{it}\)) are explored. Where gross volatility \(\sigma ^{G}_{it}\) is examined, it is implicitly assumed that there is scope for risk from both exports and imports to increase the value of flexibility. For example, consider a firm that derives a substantial share of revenue from exports to the U.S., and likewise has a substantial share of U.S. imports in its costs. An increase in SEK-USD exchange rate volatility on its exports might drive a firm to invest in a more flexible production system that allows it to switch across product lines. Likewise, increased volatility of import prices might drive a firm to increase flexibility by investing in production systems that allows it to substitute input suppliers from other currency areas. Currency risk would then increase in both export and import exposure, or gross exposure. Gross exposure will be our benchmark specification.

Alternatively, note that exposure to first moment export and import shocks are likely to partly cancel, i.e. “natural hedging”. A firm’s exposure to an appreciation in the domestic currency is likely to be offset by exporting to, and importing from, the same currency area. Hence, it could be that export and import sources of risk cancel and that net exposure is what matters from the point of view of the firm. We do not have any prior on whether it is gross or net currency exposure that matters, but we do note that the nature of the shocks and the production technology of the firm play a role. If it is “net” exposure that matters, we would observe a significant effect of net exchange rate risk and we also explore this with a specification identical to Eq. (7) except we replace \(\sigma ^{G}_{it}\) with \(\sigma ^{NET}_{it}\).

As controls we use the natural log of the level of the firm-specific real exchange rate on exports and imports, \(RER^{X}_{it}\) and \(RER^{M}_{it}\) respectively. As all these variables are in natural logarithmic form the coefficients can be interpreted as elasticities. In addition we include year \(\times\) 2-digit sector fixed effects (\(\gamma _{jt}\)) and 5-digit sector fixed effects (\(\theta _s\)), where j and s denote the 2 and 5-digit sectors, respectively.

\(X_{it-1}\) is a vector of additional controls that are lagged 1 year: capital in logarithmic form for firm i in year \(t-1\), and to control for differences in international exposure we also include the lagged export share of revenue and import share of costs, \(S^X_{it-1}\) and \(S^M_{it-1}\), respectively. The statistical error term is denoted by \(\varepsilon _{it}\). Equation (7) is estimated using data from 1998–2014 on all Swedish private firms with at least 10 employees and at least some imports and/or exports during the sample period. Observations from 1997 are used as lagged control variables in the first year of the sample.

The results support our hypothesis: higher exchange rate risk is associated with a higher share of high-skilled labor. The main coefficient of interest on firm level exchange rate risk falls somewhat as additional controls are added but remains statistically significant at the 1% level and is non-trivial in magnitude even in column (4). The estimated elasticity is 0.018 and there is substantial variation in the index of exchange rate risk, which has a standard deviation of 0.01, which is greater than the average 0.008. For example, for a given level of trade intensity, a firm switching its trade activities from Germany to the U.S. would see an increase in exchange rate risk of 130%, and the associated increase in the share of highly skilled by 2.3% (0.0178 \(\times\) 130%). An average firm in the sample has around 29% high-skilled employees and around 130 employees. An increase of 2.3% in the share is roughly equivalent to such a firm replacing one low-skilled worker with one high-skilled worker. Henceforth, we refer to the results in Table 2 column (4) as the “benchmark” regression.

The estimated effect of a currency depreciation on skill levels is negative for exports \(\ln (RER^{X}_{it})\) but positive for imports \(\ln (RER^{M}_{it})\). The point estimates differ slightly across columns but under column (4) we note that the magnitude of the (negative) estimated coefficient on exports is greater than that on imports, indicating that a depreciation (appreciation) of SEK leads to a decrease (increase) in the share of high-skilled workers. In recent work (Kaiser and Siegenthaler 2016) claim to be the first to study this issue and, using data from Switzerland, find the same effect: the share of high-skilled workers increase as the Swiss franc appreciates, consistent with the idea that imported inputs are a closer substitute for low-skilled workers.

More capital is associated with a lower skill level, indicating that overall capital and high skilled labor are substitutes. While an off-the-shelf production function would typically have overall labor and overall capital as substitutes the result may still be surprising as much recent interest regarding labor markets has focused on complementarity between certain forms of capital, such as information technology, and the demand for high-skilled labor. Note that such effects would be possible also in our sample but masked by the use of an overall measure of capital. Indeed work by e.g. Michaels et al. (2014) across 11 developed economies indicates that capital in the form of information and communications technology (ICT) is a complement to high-skilled labor but that overall (non-ICT) capital rather is a substitute for high-skilled labor.

Column (4) also controls for export shares and import shares. Again the effect on export and import shares is statistically significant at the 1% level and economically significant. An increase in the export (import) share by 1% is associated with an increase in the share of skilled workers by 0.29% (0.26%). Both of these results square well with the previous literature. A number of articles have shown that exporting, and the composition of export markets, may matter for the firm’s skill composition, even disregarding risk.²⁰ The role of skill in exporting reported with the current data corroborate findings documented for other countries.

Column (5) presents the same specification reported in column (4) but uses net rather than gross currency risk. The estimated coefficient on net currency is positive and statistically significant at the 1% level. The point estimate for the elasticity of 0.014 is somewhat lower than the point estimated for the elasticity of gross currency risk and net volatility is also less variable (means and standard deviations are roughly around half as large). The other coefficients are essentially stable across columns (4) and (5) and explanatory power of regressions as measured by R-squared is almost identical. In the Online Appendix we report regression results that show that these benchmark results are essentially unchanged if we consider the period after the change in educational classifications (Table A.1) or only consider trade with OECD countries (Table A.2).

The evidence thus far indicates that firms that face high exchange rate risk employ a higher share of high-skilled workers. The effect is robust across specifications and the quantitative impact is strongest for firms that are trade intensive and in manufacturing sectors.

Little is known about firms’ responses to risk and how they impact labor demand. Some work has centered on volatility of labor demand: Kurz and Senses (2016) document systematic links between the volatility of employment growth and exporting/importing status; and Bergin et al. (2011) document how U.S. firms appear to use maquiladora production in Mexico as a margin of adjustment in response to demand shocks, with consequential high volatility of employment in these plants. Other work focuses on the interaction between sunk entry costs in foreign markets and uncertainty [see e.g. Handley and Limão (2017) whose findings point to large positive welfare effects of lowering trade policy uncertainty]. Related are also Hershbein and Kahn (2018) who show how large shocks (in their case the great recession) can accelerate an adjustment to a higher skill share and more generally work that documents changes in the relative demand for skilled labor over the business cycle (Modestino et al. 2016).

Given the level and variety of risks currently facing firms, further investigation of firm responses to risk are likely to be useful. This paper has examined the relationship between exchange rate risk and the skill level of employees at the firm level. Our study contributes to a growing body of research on the operational measures undertaken by firms to manage the risks they face. We find substantial effects of second moment shocks and firms that face higher exchange rate risk employ a higher share of skilled labor.