A. Data and Variables

We employ the offshoring data set by Hoberg and Moon (Reference Hoberg and Moon2017), (Reference Hoberg and Moon2019) for the main outcome variable. The authors develop text-based offshoring variables by parsing through firms’ SEC EDGAR (10-K) filings. They create two vocabulary dictionaries: the first identifies firms’ footprint in foreign countries (based on country names); the second contains a set of neighbor words closely linked with offshoring activities that co-exist with the countries’ words. The offshoring variables created from this process capture the raw counts of how often firms mention having offshoring activities in each country-year pair from 1997 to 2017.

There are four variables capturing different offshoring activities: Output (which refers to sales in foreign countries), external input (which refers to the outsourcing of inputs from suppliers in foreign countries), internal input (which refers to the production of inputs within the firm in foreign countries), and overall input (capturing both external and internal inputs).

Given that our primary interest is to study how domestic product market competition affects foreign market expansion (sales), we focus our analysis on the output offshoring variable. For instance, words such as Sell, Sales, Revenues, Markets, Consumers, Store, Export, and Distribute are all considered as words associated with output offshoring activities. The rich vocabulary describing firms’ output offshoring activities captures more elaborate means of expanding into foreign markets rather than simply via exports (which is just one word out of many in the vocabulary). Hence, we interpret output offshore as a more encompassing measure of firms’ presence and expansion into foreign markets. The logic behind this choice is that competitive pressure in firms’ home product market should affect their sales strategy and, hence, whether and how to offshore their output to foreign markets. Thus, we measure output offshoring (OUTPUT_OFFSHORE) as the aggregate raw counts of how often firms mention offshoring output in any given year. In additional tests, we also analyze how competition relates to input offshoring (INPUT_OFFSHORE) to get a more comprehensive understanding of how competitive pressure affects firms’ overall foreign activities.

To measure product market competition, we employ the text-based Herfindahl–Hirschman Index (TNIC_HHI) developed by Hoberg and Phillips (Reference Hoberg and Phillips2016). The HHI index measures industry sales concentration on a scale from 0 to 1. The HHI index is an inverse measure of competition, with higher values implying a higher concentration of sales within the industry, and thus weaker product market competition. The TNIC_HHI measure of competition has advantages compared to other typical measures (such as the HHI ratio based on SIC codes, for instance) because the text-based network industry classification (TNIC) allows firms to shift across industries depending on the similarity of their product vocabularies with respect to their competitors. Hence, TNIC_HHI displays within-firm variation, allowing for a cleaner identification of competition.

We perform several sensitivity checks on the measurement of both competition and OUTPUT_OFFSHORE. As alternative measures of competition, we use the product market fluidity variable developed by Hoberg et al. (Reference Hoberg, Phillips and Prabhala2014), which is a text-based variable that captures product market threats; the product similarity variable developed by Hoberg and Phillips (Reference Hoberg and Phillips2016), which captures competitive pressure firms face from close substitutes; and the number of firms in 3-digit SIC industries (e.g., Morellec and Zhdanov (Reference Morellec and Zhdanov2019)). Regarding foreign sales, we also measure it as a dummy variable equal to 1 if firms mention offshoring output, and equal to 0 otherwise.

We further merge the data set with Compustat and construct control variables. In our main specifications, we control for well-known structural drivers of internationalization, namely firm size (natural log of market value of equity) and maturity (natural log of firm age). Nevertheless, we also conduct robustness tests with additional controls to absorb greater variance (see Section VI). We define this extended vector of controls following Hoberg and Moon (Reference Hoberg and Moon2019), including growth opportunities (TOBIN’S_Q), OPERATING_PROFITABILITY (operating income before depreciation divided by total assets), LEVERAGE (long-term plus short-term debt divided by total assets), a dummy for dividend payers, CASH_HOLDINGS (cash and equivalents divided by total assets), TANGIBILITY (gross property, plant and equipment divided by total assets), INVESTMENT (capital expenditures divided by sales), R&D_ASSETS (R&D divided by total assets), ORGANIZATIONAL_CAPITAL (following the perpetual inventory method by Eisfeldt and Papanikolaou (Reference Eisfeldt and Papanikolaou2013)), and capital-labor intensity (the natural log of gross property, plant and equipment divided by the total number of employees). After excluding firms from the financial and utilities sectors, our main data set contains 5,759 firms and 53,732 firm-year observations. All variables are defined in the Appendix.

Table 1 shows descriptive statistics for the sample. The mean value of OUTPUT_OFFSHORE (C) is 19.9, meaning that on average, the raw count of output offshore mentions is roughly 20. Hence, the average firm in the sample mentions offshoring output 20 times per year. The mean value of OUTPUT_OFFSHORE (D) is 0.844, indicating that most firms (84%) in the sample engage in foreign sales activity. The mean value of the TNIC_HHI variable is 0.319, indicating an industry sales concentration ratio of about 32%. The descriptive statistics of the remaining variables are similar to those in related studies (see, e.g., Hoberg and Moon (Reference Hoberg and Moon2019), Loncan (Reference Loncan2023)).

TABLE 1Summary Statistics

TABLE 1 Long description

The table contains 36 rows of variables with columns for Mean, Std. Dev., P 25, P 50, P 75, and No. of Obs. Most variables have 53,732 observations. Key data points include:

* O U T P U T underscore O F F S H O R E (C): Mean 19.972, Std. Dev. 27.730.

* O U T P U T underscore O F F S H O R E (D): Mean 0.844, Std. Dev. 0.363.

* T N I C underscore H H I: Mean 0.319, Std. Dev. 0.283.

* L n (M E): Mean 5.883, Std. Dev. 2.163.

* T O B I N’ S underscore Q: Mean 2.187, Std. Dev. 2.546.

* O P E R A T I N G underscore P R O F I T A B I L I T Y: Mean 0.051, Std. Dev. 0.301.

* L E V E R A G E: Mean 0.216, Std. Dev. 0.252.

* C A S H underscore H O L D I N G S: Mean 0.201, Std. Dev. 0.214.

* T A N G I B I L I T Y: Mean 0.489, Std. Dev. 0.393.

* R and D underscore A S S E T S: Mean 0.056, Std. Dev. 0.111.

* I N P U T underscore O F F S H O R E (C): Mean 19.615, Std. Dev. 30.868.

* F L U I D I T Y: Mean 6.313, Std. Dev. 3.123 (53,310 observations).

* C O M P E T I T O R S (N): Mean 156.931, Std. Dev. 220.499.

* N E T underscore P R O F I T A B I L I T Y: Mean minus 0.055, Std. Dev. 0.422.

* P A T E N T underscore S T O C K: Mean 156.111, Std. Dev. 1337.054.

* B R A N D underscore C A P I T A L: Mean 0.097, Std. Dev. 0.167 (21,853 observations).

* L n (A S S E T S): Mean 5.854, Std. Dev. 2.074.

* L n (E M P L O Y M E N T): Mean 0.300, Std. Dev. 2.066.

* I D D R: Mean 0.363, Std. Dev. 0.481.

* L n (G D P) (State): Mean 13.305, Std. Dev. 0.873 (52,256 observations).

B. Empirical Model

We aim at explaining firms’ output offshoring activity with domestic market competition. We therefore regress the output offshoring variable on TNIC HHI as a measure of competition, control variables, and fixed effects. The baseline regression model is as followsFootnote ⁴:

(1) $$ OUTPUT\_{OFFSHORE}_{it}^C=\beta \times TNIC\_{HHI}_{it-1}+{X}_{it-1}^{\prime}\lambda +{\alpha}_i+{\alpha}_t+{\epsilon}_{it} $$

The dependent variable, OUTPUT_OFFSHORE $ {}_{it}^C $ , is the raw counts of how often firms mention offshoring output. Since our dependent variable is a count-based outcome, we follow the recent developments in the empirical literature and estimate the model via Poisson fixed-effects regressions. As noted by Cohn, Liu, and Wardlaw (Reference Cohn, Liu and Wardlaw2022), the Poisson model produces consistent and reasonably efficient estimates that are compliant with more general assumptions.

The main explanatory variable is the TNIC_HHI sales concentration ratio. $ {X}_{it-1} $ is the vector of control variables that includes firm size and age. We account for firm and time unobserved heterogeneity by including firm ( $ {\alpha}_i $ ) and year ( $ {\alpha}_t $ ) fixed effects. We also estimate models absorbing time-varying industry and regional shocks by including industry × year, state × year, or industry × state × year fixed effects. TNIC_HHI and all explanatory variables are lagged by 1 year. Standard errors are clustered by firm. The coefficient of interest is $ \beta $ . A negative coefficient is consistent with the salubrious competition hypothesis, whereas a positive coefficient supports the national champions hypothesis.

A. Intangible and Organizational Capabilities

The interpretation of IDD rejections as a labor mobility shock highlights the role of intangible and organizational capital. If greater labor mobility intensifies competition partly by making knowledge more transferable across firms, then firms’ responses to such shocks should depend on their ability to preserve rents through differentiation. Firms with stronger technological knowledge, organizational capital, and brand capital are better positioned to differentiate their products, commercialize innovations, and manage expansion across markets. Moreover, the literature on firm internationalization also highlights that firms with stronger intangible resources and organizational capabilities are more likely to expand abroad (Grubaugh (Reference Grubaugh1987), Morck and Yeung (Reference Morck and Yeung1992)).

We use three proxies for firms’ intangible and organizational capabilities. First, innovation and product quality shape the relation between domestic competition and foreign expansion. Existing research shows that multinational firms have high innovation capabilities (Markusen (Reference Markusen1995)), that knowledge assets are a core source of ownership advantage that enables successful internationalization (Dunning (Reference Dunning1980)), and that investments in innovation are positively associated with export sales (McGuinness and Little (Reference McGuinness and Little1981), Kogut and Chang (Reference Kogut and Chang1991), Ito and Pucik (Reference Ito and Pucik1993), and Salomon and Shaver (Reference Salomon and Shaver2005)). Moreover, domestic competition affects firms’ incentives to innovate and to improve product quality (e.g., Aghion, Bloom, Blundell, Griffith, and Howitt (Reference Aghion, Bloom, Blundell, Griffith and Howitt2005)). Therefore, we expect that firms with higher innovation capabilities will be able to sell more products and services abroad when faced with more intense domestic competition. To measure innovation capability, we use data on patent counts (Bena, Ferreira, Matos, and Pires (Reference Bena, Ferreira, Matos and Pires2017), Barkemeyer, Bena, and Li (Reference Barkemeyer, Bena and Li2026)) to construct a measure of the cumulative stock of patents (PATENT_STOCK). Patent stock is well-suited in the IDD setting because it captures accumulated technological knowledge that may help firms maintain differentiation when worker mobility increases the diffusion of know-how across domestic rivals.

Second, we use organizational capital as a measure of intangible capital that is embodied in the firm’s key employees (e.g., Eisfeldt and Papanikolaou (Reference Eisfeldt and Papanikolaou2013)). Existing research shows that organizational capital enables firms to access foreign markets. For example, the model by Yeaple (Reference Yeaple2023) shows that firms are heterogeneous in their ability to accumulate and deploy organizational capital abroad. Firms with greater organizational capital have an absolute advantage because they are able to support larger global operations across a wider product mix. Consistent with this, Yeaple (Reference Yeaple2023) provides evidence that firms with greater organizational capital have higher exports and foreign sales. Organizational capital may also help firms deploy knowledge and routines across markets when domestic competition intensifies, making it easier to respond to competitive pressure through foreign expansion. As such, we expect that firms with greater organizational capital are better able to respond to domestic competitive shocks by expanding abroad. We measure ORGANIZATIONAL_CAPITAL following Eisfeldt and Papanikolaou (Reference Eisfeldt and Papanikolaou2013) and construct the stock of organization capital using the perpetual inventory method.Footnote ⁷

Third, we use brand capital as another measure of intangible capital. Firms that are better able to internalize marketing capabilities, consumer goodwill and leverage such capabilities, are more likely to expand into foreign markets and succeed abroad (Grubaugh (Reference Grubaugh1987), Morck and Yeung (Reference Morck and Yeung1992), and Salomon and Jin (Reference Salomon and Jin2010)). Moreover, from the perspective of differentiation, brand capital is relevant because it allows firms to preserve customer loyalty and product distinctiveness even when competition and knowledge diffusion increase. Therefore, we expect that firms with strong brand capital will respond to domestic competitive shocks by offshoring output more compared to firms with weak brand capital. We measure BRAND_CAPITAL as the ratio of 3-year cumulative advertising expenses to total assets.

Each year, we split the sample at the median of the three measures of intangible and organizational capabilities and allocate firms into the high or low subsample. Next, we separately estimate for each subsample the DiD model, where the IDD rejection indicator is the explanatory variable and output offshore is the dependent variable. Table 6 presents the results.

TABLE 6Competition, Output Offshoring, and Organizational Capabilities

TABLE 6 Long description

The table presents difference-in-differences regression results for the variable I D D R across six columns divided into three categories: PATENT_STOCK (Columns 1 and 2), O R G_CAPITAL (Columns 3 and 4), and BRAND_CAPITAL (Columns 5 and 6). Each category is split into High (H) and Low (L) sub-samples.

* Row 1: I D D R coefficients.

- PATENT_STOCK: H = 3.882 (significant at 0.01 level); L = 0.283.

- O R G_CAPITAL: H = 4.801 (significant at 0.01 level); L = minus 1.032.

- BRAND_CAPITAL: H = 9.809 (significant at 0.01 level); L = minus 5.565.

* Row 2: Standard errors in parentheses.

- PATENT_STOCK: H = (1.380); L = (1.312).

- O R G_CAPITAL: H = (1.210); L = (1.107).

- BRAND_CAPITAL: H = (2.080); L = (4.562).

* Row 3: Difference H minus L (p-value).

- PATENT_STOCK: 0.166.

- O R G_CAPITAL: 0.013.

- BRAND_CAPITAL: 0.008.

* Row 4-6: Controls, Firm F E, and Year F E are all marked ‘Yes’ for all columns.

* Row 7: Observations (N).

- PATENT_STOCK: H = 20,751; L = 25,386.

- O R G_CAPITAL: H = 22,604; L = 22,602.

- BRAND_CAPITAL: H = 8,664; L = 8,012.

Columns 1 and 2 in Table 6 report the results for the stock of patents. We find a significantly positive effect of IDD rejections for firms with a larger patent stock (coefficient estimate of 3.882), and an insignificant effect for firms with a smaller stock of patents (coefficient estimate of 0.283), although the difference in the effects across the high and low subsamples is not statistically significant (p-value = 0.166). This finding provides suggestive evidence that firms possessing intangible technological advantages respond more strongly to IDD rejections by expanding abroad.

Columns 3 and 4 show the results for organizational capital. Consistent with the results for patent stock, we find a significantly positive effect of IDD rejections for firms with greater organizational capital (coefficient estimate of 4.801), and an insignificant effect for firms with weaker organizational capital (coefficient estimate of −1.032). The difference between the two estimated coefficients is statistically significant (p-value = 0.013), providing evidence that organizational capital strengthens firms’ ability to respond to labor mobility-induced competitive shocks through foreign expansion. Lastly, columns 5 and 6 report the results for brand capital. We find a significantly positive effect of IDD rejections for firms with high advertising intensity (coefficient estimate of 9.809), and an insignificant effect for firms with low intensity (coefficient estimate of −5.565). The difference in the two estimates is statistically significant (p-value = 0.008), suggesting that firms with stronger brand capital are in a better position to respond to IDD rejections through foreign expansion.Footnote ⁸

In sum, these findings suggest that firms with stronger intangible and organizational capabilities are better able to respond to labor mobility-induced competitive pressure by expanding abroad. The strongest responses are observed for organizational and brand capital, for which the effect of IDD rejections differs significantly between the high and low subsamples, while the evidence for patent stocks is more suggestive. Thus, the findings support the interpretation that when employee mobility increases and knowledge becomes more transferable across firms, firms with stronger organizational and brand capital are better positioned to preserve differentiation, redeploy knowledge, and compete in foreign markets.

B. Profitability

Firms need to be sufficiently profitable to overcome the fixed costs of expanding abroad and realizing foreign sales. Indeed, a key finding in the international economics literature is that more productive firms self-select into international activities (see, e.g., Melitz and Redding (Reference Melitz, Redding, Gopinath, Helpman and Rogof2014)). For example, Van den Berg, van Marrewijk, and Tamminen (Reference Van den Berg, van Marrewijk and Tamminen2018) show that firms with higher productivity have higher profit margins and are more likely to export. Bellone, Musso, Nesta, and Schiavo (Reference Bellone, Musso, Nesta and Schiavo2010) show that firms enjoying better financial health are more likely to become exporters, whereas Greenaway, Guariglia, and Kneller (Reference Greenaway, Guariglia and Kneller2007) find a positive relation between financial health and exporting that is concentrated in continuous exporters rather than in starters. Buch, Kesternich, Lipponer, and Schnitzer (Reference Buch, Kesternich, Lipponer and Schnitzer2014) further highlight that financing constraints limit firms’ ability to expand abroad.

As such, we hypothesize that financial capabilities in the form of sufficiently high profits are important drivers of foreign expansion under competitive pressure. Since domestic competition creates winners and losers, only firms that are sufficiently profitable and in good financial health will be able to expand abroad. Therefore, we expect that firms with higher profitability are more likely to respond to domestic competitive shocks by expanding abroad. To measure profitability, we employ three complementary measures to ensure robustness: NET_PROFITABILITY (net income divided by total assets), OPERATING_PROFITABILITY (operating income divided by total assets), and GROSS_PROFITABILITY (gross profit divided by total assets). As before, we split the sample at the yearly median and estimate separate DiD models for firms with high and low profitability. Table 7 reports the results.

TABLE 7Competition, Output Offshoring, and Profitability

TABLE 7 Long description

A data table with six columns of regression results. The primary independent variable is I D D R.

Columns 1 and 2: N E T underscore P R O F.

* Column 1 (High): I D D R coefficient is 3.664 with three asterisks, standard error 1.374. N is 23,300.

* Column 2 (Low): I D D R coefficient is minus 0.357, standard error 1.409. N is 21,571.

* Difference H minus L p-value is 0.171.

Columns 3 and 4: O P E R A T I N G underscore P R O F.

* Column 3 (High): I D D R coefficient is 4.352 with three asterisks, standard error 1.269. N is 23,306.

* Column 4 (Low): I D D R coefficient is minus 0.765, standard error 1.723. N is 21,627.

* Difference H minus L p-value is 0.089.

Columns 5 and 6: G R O S S underscore P R O F.

* Column 5 (High): I D D R coefficient is 5.605 with three asterisks, standard error 1.821. N is 22,824.

* Column 6 (Low): I D D R coefficient is minus 1.756, standard error 1.075. N is 22,477.

* Difference H minus L p-value is 0.046.

All columns include Controls, Firm F E, and Year F E marked as Yes.

Columns 1 and 2 show the results for samples split based on net profitability. We find a significantly positive effect of IDD rejections for firms with high profitability (coefficient estimate of 3.664), and an insignificant effect for firms with low net profitability (coefficient estimate of −0.357). We find similar results in columns 3 and 4 (operating profitability) and 5 and 6 (gross profitability). These results indicate that more profitable firms are better able to afford the investments required to support foreign expansion when exposed to IDD rejection shocks. Moreover, the comparisons between the high and low subsamples provide additional support for this interpretation. While the difference in the estimated effects is not statistically significant for net profitability (p-value = 0.171), we observe significant differences for operating profitability (p-value = 0.089) and gross profitability (p-value = 0.046). Overall, these results suggest that firms with stronger profitability are better able to absorb the fixed costs of foreign expansion when exposed to labor mobility shocks that intensify domestic competitive pressure.

C. Growth Capabilities

Firms’ ability and willingness to offshore output also depend on growth opportunities. Specifically, firms with high growth opportunities and little domestic competition may not need to expand abroad, while growth firms in more competitive domestic markets will seek to offshore output to capture growth in foreign markets. Existing empirical evidence supports the view that growth opportunities are positively correlated with foreign expansion. For example, Loncan (Reference Loncan2023) finds that under high domestic competitive pressure, firms with stronger growth opportunities are more likely to conduct greenfield FDI. Similarly, Bodnar and Weintrop (Reference Bodnar and Weintrop1997) find that growth opportunities are associated with stronger foreign sales performance. Furthermore, Doukas (Reference Doukas1995) shows that the returns to cross-border acquisitions are higher for firms with stronger growth opportunities.

Hence, we expect that firms with stronger growth opportunities will offshore output more when faced with more intense domestic competition. We measure growth opportunities in several ways. We use TOBIN’S_Q, SALES_GROWTH (growth rate of total sales), and ASSET_GROWTH (growth rate of total assets). Table 8 presents the results of the DiD models estimated for split samples based on these measures of growth capabilities.

TABLE 8Competition, Output Offshoring, and Growth Opportunities

TABLE 8 Long description

The table presents difference-in-differences regression results for the variable I D D R across six columns.

Column headers from left to right are: Variables, Tobin’s Q (Columns 1 and 2), Sales Growth (Columns 3 and 4), and Asset Growth (Columns 5 and 6). Each growth category is split into H (High) and L (Low) subsamples.

Row 1: I D D R coefficients.

- Tobin’s Q: Column 1 (H) is 3.356 with one asterisk; Column 2 (L) is 0.061.

- Sales Growth: Column 3 (H) is 3.031 with one asterisk; Column 4 (L) is 1.109.

- Asset Growth: Column 5 (H) is 3.593 with one asterisk; Column 6 (L) is 0.043.

Row 2: Standard errors in parentheses.

- Tobin’s Q: (1.873) for H; (1.049) for L.

- Sales Growth: (1.682) for H; (0.827) for L.

- Asset Growth: (2.138) for H; (1.110) for L.

Row 3: Difference H minus L (p-value).

- Tobin’s Q: 0.344.

- Sales Growth: 0.423.

- Asset Growth: 0.263.

Rows 4-6: Controls, Firm F E, and Year F E are all marked ‘Yes’ for all columns.

Row 7: N (Sample size).

- Tobin’s Q: 22,395 (H); 22,350 (L).

- Sales Growth: 22,175 (H); 22,301 (L).

- Asset Growth: 22,422 (H); 22,346 (L).

Columns 1 and 2 report the results for TOBIN’S_Q. We find a weakly significant and positive effect of IDD rejections for firms with high TOBIN’S_Q (coefficient estimate of 3.356), and an insignificant effect for firms with low TOBIN’S_Q (coefficient estimate of 0.061). Columns 3 and 4 split firms in terms of SALES_GROWTH. We find a weakly significant effect among firms with higher SALES_GROWTH, and an insignificant effect for firms with low SALES_GROWTH. Finally, in columns 5 and 6, we find a weakly significant and positive effect for firms with high ASSET_GROWTH, and an insignificant effect for firms with low ASSET_GROWTH. Yet, the comparisons between the high and low subsamples are statistically insignificant for all variables.

Overall, the growth opportunities results are positive but weaker than the evidence for intangible and organizational capabilities and profitability. Although firms with higher Tobin’s Q, sales growth, and asset growth exhibit larger estimated responses to IDD rejections, the differences are not statistically significant in the split sample tests. We therefore interpret these findings as suggestive evidence that growth opportunities may increase firms’ incentives to expand into foreign markets following IDD rejection shocks.

D. Firm Scale

Finally, we explore the effects of competitive shocks considering two potential structural drivers of foreign expansion: firm size and maturity (age). Regarding firm size, the literature suggests that larger firms are more likely to overcome the elevated fixed costs of internationalization (e.g., Markusen (Reference Markusen1995)). For example, Tomiura (Reference Tomiura2007) shows that the gap between globalization and domestic orientation widens with firm size. Given their resourcefulness and ability to deploy scale economies, we expect that larger firms will respond more strongly to domestic competitive shocks by offshoring output.

The role of firm age is less clear. On the one hand, older firms are more resourceful, and this enables them to overcome the high fixed costs associated with expanding abroad (Li, Zhang, and Shi (Reference Li, Zhang and Shi2020)). On the other hand, Carr et al. (Reference Carr, Haggard, Hmieleski and Zahra2010) argue that firm age can manifest both in terms of a liability of newness and a liability of aging. While older firms may be more resourceful, younger firms are more agile. Corroborating the latter view, the findings from Hoberg and Moon (Reference Hoberg and Moon2017) point toward a negative association between firm age and output offshoring. Given such mixed views, we remain agnostic about the role of age for firms’ response in output offshoring to competitive shocks.

We measure size with the log of market equity (Ln(ME)) and divide the sample into large (above the median) and small (below the median) firms. For age (Ln(AGE)), we count the years since a firm appears in Compustat and split the sample at the yearly median. Table 9 reports the results.

TABLE 9Competition, Output Offshoring, and Firm Scale

TABLE 9 Long description

The table presents difference-in-differences regression results for the variable I D D R across four columns.

Columns 1 and 2 split the sample by L n (M E) into Large and Small firms.

- Column 1 (Large): Coefficient is 2.689 with a standard error of 1.581, significant at the 0.10 level.

- Column 2 (Small): Coefficient is 1.396 with a standard error of 0.996.

- The p-value for the difference between Large and Small is 0.622.

Columns 3 and 4 split the sample by L n (A G E) into Old and Young firms.

- Column 3 (Old): Coefficient is 1.467 with a standard error of 1.428.

- Column 4 (Young): Coefficient is 3.321 with a standard error of 1.912, significant at the 0.10 level.

- The p-value for the difference between Old and Young is 0.588.

All columns include Controls, Firm F E, and Year F E. The sample size N is 23,435 for Large, 22,067 for Small, 23,868 for Old, and 22,120 for Young.

Columns 1 and 2 show the results for firm size. We find a weakly significant and positive effect of IDD rejections among large firms (coefficient estimate of 2.689), and an insignificant effect among small firms (coefficient estimate of 1.396). Columns 3 and 4 present the results for firm age. We find a weakly significant and positive effect of IDD rejections among younger firms (coefficient estimate of 3.321), and an insignificant effect for older firms (coefficient estimate of 1.467). For both the size and age variables, the comparisons between subsamples return insignificant results. We therefore interpret the scale and age results as suggestive at most. While larger and younger firms exhibit positive estimated responses to IDD rejections, the evidence does not indicate that these responses are statistically different from those of smaller or older firms.

Overall, the results in this section suggest that firms’ foreign sales response to labor mobility-induced competitive shocks is concentrated among firms with stronger pre-existing capabilities. The evidence is strongest for intangible and organizational capabilities, especially organizational capital and brand capital, and for profitability, particularly operating and gross profitability. The evidence for growth opportunities and scale is weaker and should be interpreted as suggestive. The patterns are consistent with the idea that IDD rejections increase competitive pressure by making workers and knowledge more mobile across firms, but that only some firms are able to respond by expanding abroad. Specifically, organizational capital and brand capital help firms preserve differentiation, redeploy knowledge across markets, and protect rents when labor mobility increases. Profitability, in turn, provides the internal resources needed to absorb the fixed costs and risks of foreign expansion. These findings suggest that labor mobility-induced competitive pressure stimulates output offshoring primarily among firms with the capabilities needed to translate greater domestic pressure into foreign sales expansion.

A. Evidence on Input Offshoring

Our analysis shows that domestic competition has a positive and significant effect on the revenue side of firms’ internationalization. Domestic competition, however, could also affect firms’ foreign activities through a cost side channel. Specifically, competitive shocks can induce firms to reorganize production and sourcing activities internationally to reduce costs. We therefore explore if and how domestic competition affects firms’ input offshoring activity.

Output and input offshoring are correlated because firms often combine foreign sales with foreign sourcing or production. For example, Hoberg and Moon (Reference Hoberg and Moon2017) show that firms use offshore input purchases partly as an operational hedge against foreign currency exposure arising from foreign output sales. However, while correlated, output and input offshoring are conceptually distinct, and the economic channels that affect firms’ decisions to offshore output versus input are different. While output offshoring captures foreign market expansion, input offshoring captures firms’ decisions to procure, produce, or source inputs abroad. As such, input offshoring is tightly linked to cost reduction motives and production reorganization.

The idea that input offshoring reduces costs is consistent with the international trade literature on offshoring. For example, Feenstra and Hanson (Reference Feenstra and Hanson1996), (Reference Feenstra and Hanson1999) emphasize that foreign outsourcing allows firms to relocate parts of production to locations with different factor costs, thereby changing firms’ effective input costs and domestic labor demand. Grossman and Rossi-Hansberg (Reference Grossman and Rossi-Hansberg2008) develop a model in which offshoring reduces the cost of performing particular tasks and generates cost savings for firms. Moreover, Egger, Kreickemeier, and Wrona (Reference Egger, Kreickemeier and Wrona2015) and Braconier, Norbäck, and Urban (Reference Braconier, Norbäck and Urban2005) show that offshoring input allows firms from high-income countries to hire workers for routine tasks at a lower wage abroad. Because competition puts pressure on firms’ margins and thus incentivizes firms to save costs (see, e.g., Raith (Reference Raith2003)), we expect that higher domestic competition leads to more input offshoring because it allows firms to reduce labor costs.

We test whether domestic competition affects input offshoring in two steps. First, we re-estimate the baseline model (1) and replace the dependent variable with input offshoring. Second, we use the IDD rejection framework to analyze whether labor mobility-induced competitive shocks affect input offshoring more strongly among firms for which labor cost savings are likely to be more valuable. Specifically, we estimate the model separately for firms with high and low employment levels, high and low employment growth, and firms operating in industries with high and low labor cost growth.Footnote ⁹ As in the previous tests, we split the sample at the yearly median in the pre-treatment period. The three variables capture different dimensions of firms’ exposure to labor costs. Firms with a larger workforce have a larger domestic labor cost base. Firms with faster employment growth face rising labor demand and potentially greater pressure to contain labor costs. Firms in industries with rising labor costs have stronger incentives to substitute toward lower cost foreign inputs when competition intensifies. Panel A of Table 10 presents the baseline regression results, and Panel B shows the results of the IDD rejection tests.

TABLE 10Competition and Input Offshoring

TABLE 10 Long description

Panel A. Baseline Model. This panel presents Poisson regressions with T N I C underscore H H I as the primary independent variable across five columns.

* Column 1: Coefficient -0.131 (standard error 0.038), significant at 0.01 level. Includes Firm and Year F E. N = 50,701.

* Column 2: Coefficient -0.123 (standard error 0.038), significant at 0.01 level. Adds Controls. N = 50,701.

* Column 3: Coefficient -0.128 (standard error 0.039), significant at 0.01 level. Includes Industry cross Year F E. N = 49,558.

* Column 4: Coefficient -0.133 (standard error 0.039), significant at 0.01 level. Includes State cross Year F E. N = 48,773.

* Column 5: Coefficient -0.082 (standard error 0.054). Includes Industry cross State cross Year F E. N = 31,071.

Panel B. I D D R Difference-in-Differences Model. This panel reports the effect of I D D R on input offshoring across seven columns.

* Column 1 (Full Sample): Coefficient 1.553 (standard error 0.927), significant at 0.10 level. N = 47,033.

* Columns 2 and 3 (Ln EMPLOY): High (H) coefficient 3.232 (significant at 0.10 level) vs Low (L) coefficient -0.103. Difference p-value is 0.213.

* Columns 4 and 5 (E M P underscore G R O W T H): High (H) coefficient 7.090 (significant at 0.01 level) vs Low (L) coefficient -0.567. Difference p-value is 0.013.

* Columns 6 and 7 (L A B underscore C O S T underscore G R O W T H): High (H) coefficient 5.422 (significant at 0.01 level) vs Low (L) coefficient -0.315. Difference p-value is 0.038.

All models in Panel B include Controls, Firm F E, and Year F E.

The results in Table 10 broadly support the cost minimization interpretation. In Panel A, the TNIC_HHI variable is negatively and significantly associated with INPUT_OFFSHORE in all specifications except in the specification that includes industry×state×year fixed effects. The marginal effect associated with the estimated coefficient of −0.123 (column 2) is −2.55. This implies a 2.5 decrease in counts of INPUT_OFFSHORE for a one-unit increase in TNIC_HHI. With a sample average of 19.6, this effect implies a 13% relative decline in input counts.

Panel B presents evidence using IDD rejections as competitive shocks. Column 1 shows that IDD rejections have a weakly significant and positive effect on input offshoring for the full sample (coefficient estimate of 1.55).Footnote ¹⁰ Furthermore, the findings in columns 2 to 7 indicate that firms’ input offshoring response to IDD rejection shocks is concentrated among firms with high labor cost exposure. Firms with high employment in levels and growth, as well as firms in industries experiencing high labor cost growth, increase input offshoring significantly following IDD rejections. We do not observe such a reaction for firms with low employment in levels and growth, nor in industries with low labor cost growth.Footnote ¹¹ Moreover, we find statistically significant differences in the estimates across the high and low subsamples for employment growth (p-value = 0.013) and for labor cost growth (p-value = 0.038). These findings suggest that domestic competition affects input offshoring through a distinct cost efficiency channel. When domestic competitive pressure increases, firms with greater labor cost exposure are more likely to shift sourcing or production activities abroad.

These results complement our main findings on output offshoring. The results on output offshoring indicate that competition encourages firms to expand into foreign markets, especially when they possess the intangible and financial capabilities needed to compete abroad. The results on input offshoring point to a different margin of adjustment. Competition also induces firms to reduce production costs by reorganizing input sourcing internationally, particularly when labor cost savings are likely to be economically important. Overall, the evidence suggests that labor mobility shocks that increase domestic competition shape firms’ internationalization strategies on both the revenue and cost sides, but through economically distinct channels.

B. Robustness Tests

In this section, we discuss further tests to account for potential endogeneity, reverse causality, and simultaneity between competition and foreign market expansion, as well as measurement and estimation issues (these additional results are reported in the Supplementary Material).

First, we re-estimate the baseline specification (1) and expand the vector of firm control variables to absorb greater variation. Following Hoberg and Moon (Reference Hoberg and Moon2019), we control for growth opportunities, operating profitability, leverage, a dummy for dividend payers, cash holdings, asset tangibility, investment, R&D intensity, organizational capital, and capital-labor intensity. The results are reported in Table OA1 in the Supplementary Material. We continue to observe a significantly negative (positive) association between concentration (competition) and output offshoring.

Second, we employ an instrumental variable approach. We follow Gutiérrez and Philippon (Reference Gutiérrez and Philippon2017) and construct a measure of excess entry in industries in the 1990s as an instrumental variable for concentration in the following decade (2000–2009). They propose a model to study the effect of industry concentration on investment and suggest that an instrument that randomly changes the opportunity cost of entry across industries is suitable to instrument for industry concentration. They argue that the peculiar dynamics of entry in the 1990s, in particular regarding abnormally high equity valuations, can satisfy such requirements.

We calculate the instrumental variable EXCESS_ENTRY $ {}_j^{90s} $ as the residual of a cross-sectional industry-level (3-digit SIC) regression in which entry (the growth rate in the number of firms in each industry between 1990 and 1999) is predicted by a set of industry fundamentals (corresponding to the same period) obtained from Compustat, including Tobin’s Q, sales growth, cash flow to assets, and firms’ mean assets and age at the beginning of the decade (as of 1989). The residual of this predictive regression arguably captures entry of competitors that is orthogonal to fundamentals, thus capturing excess entry and hence excess competition.

The results of the IV models are reported in Table OA2 in the Supplementary Material. The first-stage results show that EXCESS_ENTRY $ {}_j^{90s} $ in the 1990s significantly predicts weaker concentration (TNIC_HHI) and stronger competition (FLUIDITY) in the next decade (2000–2009).Footnote ¹² In the second-stage regressions, akin to the IV setting, we use GMM and, hence, measure output offshoring as a dummy (thus fitting a pooled linear probability model). The second-stage results indicate that both TNIC_HHI and FLUIDITY (instrumented by EXCESS_ENTRY $ {}_j^{90s} $ ) have significantly negative and positive effects, respectively, on the propensity of foreign market expansion.

Third, we analyze whether the effect of competition remains robust when we introduce a longer temporal distance between TNIC_HHI and OUTPUT_OFFSHORE. To do so, we lag TNIC_HHI by 2, 3, 4, and 5 years. The results are reported in Table OA3 in the Supplementary Material. While the relation between TNIC_HHI and OUTPUT_OFFSHORE remains statistically significant, it becomes economically weaker for longer lags, fading out when concentration is lagged by 5 years or more. Overall, these results suggest that our findings are unlikely to be affected by reverse causality, and that the effect of concentration (competition) on output offshoring is persistent over time, though eventually decreasing in the long run.

Fourth, we perform additional tests on measurement and estimation, reported in Table OA4 in the Supplementary Material. We i) estimate the baseline model via OLS (instead of Poisson), ii) estimate the effect of changes in concentration on changes in OUTPUT_OFFSHORE via first-difference models, iii) use the ABNORMAL_OUTPUT variable following the orthogonal transformation suggested in Hoberg and Moon (Reference Hoberg and Moon2019), measured as the residual of a regression in which the dependent variable is (1 + log) output counts and the explanatory variable is (1 + log) Input counts (also implementing the orthogonal transformation for positive entries of Output and Input, hence without applying the 1 + log operator), iv) estimate the effect of concentration (competition) on the propensity to offshore output with an output dummy as dependent variable, via both logit and probit regressions. The main result remains robust to all these additional analyses.

Fifth, we conduct robustness tests on the estimation of the DiD model for the IDD rejections. Specifically, we estimate stacked DiD models (see, e.g., Gormley and Matsa (Reference Gormley and Matsa2011), (Reference Gormley and Matsa2016)) as an alternative possibility to account for heterogeneous treatment effects. Moreover, the stacked DiD models allow us to test for the channels (organizational capabilities, profitability, growth options, and firm scale) using interactions between the IDD rejection indicator and the variables capturing such characteristics. Table OA5 in the Supplementary Material presents the results. The coefficient estimates on the interaction terms all have the expected positive sign and are statistically significant in 8 out of 11 specifications. Only for organizational capital, Tobin’s Q, and age, the interaction coefficient with the IDD rejection indicator variable is not statistically significant. Overall, these results support our main finding of a positive effect of IDD rejection shocks on firms’ foreign sales activity and our interpretation that firms’ intangible, organizational, and financial capabilities are potential economic channels.