Import, Productivity, and Export Performances

1 Introduction

It is widely acknowledged that global trade has significantly contributed to economic growth, particularly in developed countries. Firms of all sizes and sectors have contributed to and benefited from this development. For instance, in Sweden, over 70% of manufacturing firms with a minimum of 10 employees export their products to international markets. Concurrently, it is well-documented that most of the firms, particularly those in small open economies like Sweden, heavily rely on global sourcing, as the production process nowadays is more fragmented and spread around the world. The increasing role of global value chains (GVCs), which enables firms’ involvement in an international network of production, has significantly enhanced the link between the origin of imports and the destination of exports. To effectively engage in the export market, firms need to carefully assess the sources of their intermediate inputs. Participation in GVCs is therefore an important factor in shaping firms’ sourcing decisions, which, in turn, may impact their overall export strategies. Additionally, previous research has highlighted the crucial role of productivity in enabling firms to enter the export market and identify potential export destinations.

Given these circumstances, the present article focus on the link between import, productivity, and export performances in terms of export sales and the number of export markets. The article seeks to address two key questions. First, it aims to empirically analyze how firms’ ex-ante characteristics influence their import decision and their choice between low-wage or high-wage countries or both as import sources. Second, the article aims to analyze whether productivity and export performances differ between firms engaged in importing and those that are not, taking into account various firm-specific characteristics that may affect the outcomes and the geographical origin of the imports.

The data used in this article are from Statistics Sweden (SCB) covering the entire manufacturing firms with at least 10 employees over the period 2007–2020. The register-based information at the firm level is linked to the firms’ foreign trade, disaggregated at the six-digit product level for imports and exports.

Before we start to answer the above questions analytically, however, it is important to provide a comprehensive overview of how firms’ involvement in the international market has developed in recent years. This includes examining the presence of firms active in the international market, whether as importers, exporters or both, and the number of products and different countries from which the intermediate inputs are imported.

Although several international literatures have studied the role of imports in firm productivity and export opportunities, there are still knowledge gaps. The literature, up to date, has focused on the characteristics important for firms to become importers (e.g., Kedia & Mukherjee, 2009; Lewin et al., 2009; Mol et al., 2005), the location choice of the import (e.g., Doh et al., 2009; Rasciute & Downward, 2017), the import effect on different labor market outcomes (e.g., Abramovsky et al., 2017; Bandick, 2016; Feenstra & Hanson, 1999; Head & Ries, 2002; Hummels et al., 2014), or how imports affect firm productivity and other performance (e.g., Aristei et al., 2013; Bas & Strauss-Kahn, 2014; Bertrand, 2011; Elliott et al., 2016; Hijzen et al., 2010; Jabbour, 2010; Smeets & Warzynski, 2013). Nevertheless, as argued by Antràs and Chor (2022) and Antràs et al. (2022), there remains a knowledge gap in understanding how these issues are interconnected. It is reasonable to believe that factors influencing firms’ import decisions and sourcing location may also correlate with firm performance in terms of productivity and export opportunities. Still, even if the decision to import and sourcing location are endogenously driven, firms may also experience improved productivity and export performance ex-post due to learning by doing (Ethier, 1982; Grossman & Helpman, 1991; Markusen, 1989; Young, 1991). However, most of the relatively small number of empirical firm-level studies (e.g., Amiti & Konings, 2007; Aristei et al., 2013; Bas, 2012; Castellani et al., 2010; Kasahara & Rodrigue, 2008; Kugler & Verhoogen, 2009; Muuls & Pisu, 2009; Smeets & Warzynski, 2013) do not explore if there are differences in the learning process between different sourcing locations, and many of them even ignore investigating the direction of the causality, whether it is self-selection, learning, or both.

The learning effect partly originates from the perspective that imports may lead to productivity enhancement by introducing higher quality and more diversified varieties of inputs that are otherwise unavailable in the domestic market. Additionally, importing firms may encounter lower sunk-entry costs for exporting, as they possess a deeper understanding of the export destination, including factors such as language, culture, distribution channels, and networks. Furthermore, the impact of the learning effect on productivity and export depends on the sourcing location. The larger the technology gap between the firm’s home country and the origin of imports, the greater the learning effect, provided that the firms have the capacity to assimilate the advanced technology inherent in the imported inputs (Lööf & Andersson, 2010).

However, empirical findings regarding the presence of the learning-by-importing effect exhibit significant variation across previous studies, as reviewed by Wagner (2012) and further discussed by Fernández and Gavilanes (2017). For instance, Pane and Patunru (2023) find that Indonesian firms utilizing imported inputs in their production process experience enhanced productivity, with a more pronounced effect for inputs originating from developed countries. Elliott et al. (2016) find greater productivity gains among Chinese firms that import from high-wage countries, in comparing to importing from low-wage countries. Feng et al. (2016) show that intermediate inputs from G7 countries prove more advantageous for Chinese firms initiating exports to the same destinations. Additionally, for the Spanish firms, Máñez et al. (2020) show that the import of intermediate inputs increases the likelihood of becoming an exporter, while Díaz-Mora et al. (2015) find positive export survival rates for small firms that source inputs from abroad. Requena et al. (2023), also investigating Spanish firms, found modest export increase for those firms utilizing imported intermediates from non-EUEFTA countries, while sourcing high-quality inputs, regardless of the income level of the country, contributed to enhancing Spanish firms’ exports to the EU15. Bas and Strauss-Kahn (2014) examine French firms and find that imported inputs from developed countries augment productivity levels and exports, whereas inputs from developing countries only increase productivity levels. Conversely, Jabbour (2010) who also studies French firms, reveals that inputs from developing countries increase firm productivity, while inputs from developed countries have no significant effect on productivity.

Hence, the objective of this article is to empirically investigate the relevant firm-level characteristics necessary for firms to engage in global sourcing and determine the sources of such imports. Another crucial gap that this article tries to address is how import origin affects firm productivity and the link to export sales and the number of export markets. As discussed earlier, the impact of imported intermediate inputs on productivity and export performance, and whether these effects differ based on imports from low-wage or high-wage countries or both, remains unclear – both theoretically and empirically. Furthermore, few studies have examined these questions from a Swedish perspective, many of which are outdated due to the availability of more recent data and significant methodological advancements in recent years. In this article, we utilize the most available data ranging from 2007–2020 and implement an event study approach to capture heterogeneous treatment effects (de Chaisemartin & D’Haultfoeuille, 2022; Clarke & Schythe, 2020; Sun & Abraham, 2021) and a difference-in-differences (DiD) approach with a propensity score matching estimator (PSM).

This article underscores several important findings. First, almost 80% of Swedish manufacturing firms with at least 10 employees participate in the international market. The smallest firms predominantly engage as exporters, middle-size firms focus on imports, and the largest firms engage in both import and export activities, so-called two-way traders. Second, irrespective of firm size, most imports originate from high-wage countries. However, over the years, a gradual shift from high-wage to low-wage country imports has been observed. Third, becoming an importer requires possessing high levels of firm-specific characteristics. The most productive, largest, and have the highest average wage costs, and exporters are more inclined to import from all countries. Firms with slightly lower characteristics tend to import from either low- or high-wage countries, while firms with the lowest characteristics are non-importers. Among the small importing firms, the level of capital stock is a relatively more important determinant if imports are concentrated in either low- or high-wage countries, as compared to the case if firms import form all countries. Fourth, controlling for the self-selection mechanisms using an event study approach and difference-in-differences matching estimator, we find, for larger firms, no differences in productivity and export sales growth between importing and non-importing firms. However, we do find evidence suggesting that small importing firms experience higher productivity growth, primarily driven by imports from high-wage countries. Lastly, both small and large firms importing from both high- and low-wage countries have, on average, more high-wage export markets as compared to non-importers.

The next section describes the data and summary statistics. Section 3 describes the estimation strategies. Section 4 reports the results, and Section 5 concludes.

2 Data and Summary Statistics

The data utilized in this article were collected by Statistic Sweden over the period 2007–2020, covering all manufacturing firms with at least 10 employees^[1]. The data comprise two databases, the Firm Statistics Register (FirmStat) and the Swedish Foreign Trade Register (TradeStat). The information extracted from FirmStat includes general firm accounting data such as sales, number of employees, wage costs, value-added, net investment value, capital stock, intermediate inputs, purchased final goods, and industry identifier^[2]. Utilizing the number of employees, firms are categorized into small firms with fewer than 50 employees and large firms with 50 employees or more. Moreover, the information from FirmStat enables us to calculate capital intensity, defined as capital stock over sales, average wage costs, labor productivity, defined as sales per employee^[3], and estimates total factor productivity by using the Levinsohn and Petrin (2003) methodology.

The information from TradeStat contains firm-level export and import values disaggregated by destination/origin and products, measured at the six-digit Combined Nomenclature (CN6). These data enable us to calculate, for each firm, the aggregate value of exports and imports to/from low- and high-wage countries. The definition of low- and high-wage countries is from The World Bank as of 2010 fiscal year.^[4]

FirmStat and TradeStat can be merged by using the unique firm identifier which means that we have access to general accounting data and aggregate export and import values for both low- and high-wage countries for each firm. Unfortunately, however, there are some limitations to our dataset.

First, TradeStat does not provide information regarding whether traded products are raw materials, semi-manufactured goods, or intermediaries. Consequently, for imports, we cannot distinguish between transactions involving intermediate inputs or final goods. This implies that importing firms might include retailers that are not engaged in product manufacturing. Nevertheless, focusing only on the manufacturing sector, we observe that a very small fraction of firms report positive values for purchased final goods while simultaneously reporting zero values for intermediate inputs. This suggests that most of the firms in our data can be classified as manufacturers, with imported transactions likely serving as intermediate inputs. Second, TradeStat lacks information regarding whether imported transactions originate from affiliated providers (intra-firm sourcing) or unaffiliated providers (offshore outsourcing). Hence, we follow previous literature (e.g., Bandick, 2020; Crinò, 2009; Grossman & Rossi-Hansberg, 2006; Hijzen et al., 2010; Olsen, 2006) by using the value of imports as a proxy for global sourcing, without distinguishing the affiliation of the provider with the importing firm^[5].

Given the available data, we are able to classify firms as exporters, importers, or both (referred to as two-way traders). Exporting firms are identified by assigning an indicator (Exporter) equal to one if the export value is positive, and zero otherwise. Additionally, we calculate the export intensity for each firm (EX_intensity_sales), defined as the ratio of export sales to total sales. Similarly, for importing firms, we assign an indicator (Importer) equal to one if the import value is positive, and zero otherwise. Import intensity is measured in two ways: one as the ratio of import value to total sales (IM_intensity_sales), and the other as the ratio of import value to the total value of intermediate inputs (IM_intensity_intermediate). Lastly, an additional indicator (two-way) is assigned to two-way traders, equaling one if both export and import values are positive, and zero otherwise.

With the panel data structure defined in terms of content and limitations, as well as the main variables, we proceed to present key facts and recent developments concerning the role of global sourcing in the Swedish manufacturing sector.

Table 1 presents summary statistics for the number of firms, mean employment and the share of importer and exporter over the period 2007–2020. Although the total number of firms declined over the period, the proportion of firms engaged in international trade, whether as importers, exporters, or both, remains relatively stable around 80%. On average, 7% of the firms exclusively use the international market for imports, while almost 13% exclusively engage in exporting, and 58% are two-way traders. Furthermore, Table 1 reveals clear differences in firm size, where the largest firms are two-way traders with an average of 124 employees, followed by exclusive importers with 33 employees, and the smallest firms are exclusive exporters with an average of 26 employees.

Table 2 shows the distribution of importing firms based on their source countries, classifying them as exclusively importing from low-wage countries, exclusively from high-wage countries, or from both. At the beginning of the analysis period in 2007, nearly 60% of importing firms exclusively imported from high-wage countries. However, by the end of the study period in 2020, this share had decreased to less than 40%. Importantly, compared to all manufacturing firms, the share of firms exclusively importing from high-wage countries declined by almost 12% points during the 2007–2020 period. Meanwhile, firms importing from both low- and high-wage countries increased by 16% points relative to all importing firms and by 12% points relative to all firms. Firms exclusively importing from low-wage countries also exhibited a positive trend, with their share among total importing firms nearly quadrupling, from 1.6 to 5.4%, and among all firms increasing from 1.0 to 3.7%.

As shown in Table 3, the shifts in import origin observed in Table 2 are evident across nearly all the two-digit (NACE Rev.2) manufacturing industries. However, these patterns exhibit significant variation among industries. For instance, in the Manufacturing industry of leather and related products (SNI07_15), the share of firms exclusively importing from high-wage countries decreased by about 2% points from 2007 to 2020, whereas in the Manufacturing industry of coke and refined petroleum products (SNI07_19), this share dropped by 33% points. Similarly, while some industries had no change in the share of firms exclusively importing from low-wage countries, others experienced an increase of nearly 5% points over the same period. These differences highlight industry-specific trends in import sourcing that we need to control for in the empirical model below.

When examining the development of import intensity and the number of imported products over the years, it becomes evident that differences exist between small and large firms. This distinction is highlighted in Table 4. For small firms (those with fewer than 50 employees), the average share of import value over sales is 6%, and, comparatively, the share of import value over total intermediate inputs is nearly 11%. In contrast, large firms exhibit significantly higher ratios of 17 and 31%, respectively. Moreover, regarding the number of imported products (as shown in the last column of Table 4), small firms import an average of 3.6 products, while large firms import an average of 13 products.

In Table 5, we present the average number of unique countries from which firms import. The categories of countries that imports originate from are divided into EU-15, EU-12 (countries that joint the EU during the 2004, 2007, and 2013 enlargement), other high-wage countries, and lastly, low-wage countries. It is evident that regardless of firm size, imports primarily originate from EU15 and other high-wage countries. The dominance of EU15 countries as a source of Swedish imports, besides the membership of the European Union, can be explained by firms’ participation in GVCs, especially within strategic sectors such for example the automotive and pharmaceutical sectors, where the production chains mainly are clustered regionally in Europe. Over the years, small firms appear to have imported from an average of 1.3 EU15 countries, while large firms imported from more than one-third of EU15 countries. As it seems, the number of countries within EU12 and other high-wage countries as import partners are relatively consistent over the years for small firms.

However, the significance of low-wage countries as import partners has increased over time. For instance, the average number of low-wage countries from which small firms import has doubled, from an average of 0.4 countries in 2007 to 0.8 countries in 2020. During the same period, small firms reduced the number of countries within the EU15 as import partners by an average of 0.3 countries. Among large firms, the number of import countries within EU15 and other high-wage counties remains relatively constant, while they appear to import from more EU12 and low-wage countries. Notably, the number of EU12 import countries increased by more than 70%, while the number of low-wage import countries increased by almost 40%, respectively^[6].

Moving to Table 6, we present differences in various firm-specific characteristics between importers and non-importers, stratified by firm size and import origin. A notable difference emerges between small firms that are importers and their non-importing counterparts, irrespective of whether they import exclusively from low- or high-wage countries or both. Small importing firms exhibit higher productivity, larger export sales, engagement with more export destinations, larger employment and capital stock, and higher average wage costs. Interestingly, these differences are even more pronounced when comparing small firms that import from all countries to those that do not engage in importing. Conversely, no statistically significant differences are observed between large firms exclusively importing from low-wage countries and their non-importing counterparts. However, large firms importing from either high-wage countries or all countries display significantly better firm-specific characteristics compared to non-importing large firms.

In conclusion, this section highlights the substantial engagement of Swedish manufacturing firms in the international market, as importers, exporters or, as most of the firms are, two-way traders. There is a clear ranking of the firms, with exclusive exporters being the smallest, exclusive importers being the middle, and two-way traders are the largest group in terms of the number of employees. Irrespective of firm size, a considerable share of imports originates from high-wage countries. Nonetheless, changes in import distribution are observed over the years, including shifts from high- to low-wage countries and an increase in the number of low-wage import partners. Notable, importing firms display superior firm-specific characteristics compared to non-importing counterparts, with this gap being more pronounced among those importing from both low- and high-wage countries.

Given the significance development and shift of import origin, as well as the disparities between importers and non-importers, we proceed to analyze in subsequent sections how these trends have impacted firms’ productivity and export performances.

3 Econometric Considerations

The empirical strategy in this paper is to investigate the differences in outcomes for firms that import with the outcomes they would have had if they had not imported. However, this hypothetical “non-importer” event is not observable for firms already involved in importing. Following Heckman et al. (1997), we can define the average effect of importing on the outcome variable (in our case productivity and export performances) as:

Here, IMP it is a binary variable equal to 1 if firm i imports at time t. The first term, E { y t + s 1 | IMP it = 1 } is the average effect on the outcome variable y, after s years of the firm initiating importing. The last term, E { y t + s 0 | IMP it = 1 } , is the average effect on the outcome variable had the firm not started importing. Since the latter term is unobservable, we need to construct a counterfactual for this hypothetical event, namely, the outcome variable y that importing firms would have experienced had they not begun importing. This is estimated by the average outcome y for firms that did not import, denoted as E { y t + s 0 | IMP it = 0 } .

However, this approximation is only valid if there are no contemporaneous effects that are correlated with the decision to import. Such effects could introduce endogeneity bias into the empirical analysis. Unfortunately, this is not unlikely to be the case since, as discussed earlier, and demonstrated in Table 6, there is empirical evidence that firms that import tend to possess better firm-specific characteristics compared to non-importing firms. This makes it difficult to distinguish whether post-import productivity and export performances are influenced by the decision to import, or by the firms’ ex-ante characteristics. Therefore, when constructing the counterfactual event, it is crucial to select a valid control group. To address this issue of endogeneity bias, we employ two common methods, namely Event Study approach and DiD with a PSM.

Starting with the Event Study approach, the initial step involves examining whether there are any differences in pre- and post- trends between importing and non-importing firms. The challenge at hand involves two-sided heterogeneity: one across firms to be treated (becoming an importer) and another with respect to the timing of the treatment. As previously discussed, importing firms often possess superior characteristics that may influence their decision to engage in importing. This selection bias can obscure the true impact of importing on productivity and export performances, given that importers tend to be inherently more productive and possess advantageous attributes compared to their counterparts. Moreover, firms may opt to import from various sourcing countries due to distinct strategic motives that may correlate with their pre-existing firm-specific attributes. Furthermore, the effect of imports on productivity and export performances varies depending on when firms begin importing since not all firms initiate imports simultaneously.

To address these complexities, we adopt the panel Event Study framework proposed by Clarke and Schythe (2020) and de Chaisemartin and D’Haultfoeuille (2022). More precisely, we estimate the following dynamic two-way fixed model:

In equation (2), y igt represents the outcome variable, in our case productivity (measured as sales per employee) or export sales, for firm i in industry g at time t. Here, θ i , θ g , and θ t denote firm, industry, and time fixed effect, respectively. X ig base captures control variables measured at the base year, which is the first year the firm enters the sample. The variables included in X i are; log average wage costs to account for labor costs, growth (in terms of sales) relative to the industry to consider workload volatility, and firm size to control for economies of scale. Proxy to measure firm size will be the log level of employment and capital stock. ε igt stands for an unobserved error term.

The binary variables j and k indicate the number of years before and after the first event in which the import indicator switches from zero to one. For instance, j = k = 1 refers to one year before and one year after the firm´s initial import occurrence. It is important to note that including longer periods of leads ( J ) and lags ( K ) in the model imposes more constraints, as firms must have a longer survival spell in the panel. This concern arises particularly if firm survivability relies on pre-trend characteristics (Roth, 2022). In the subsequent analysis, we incorporate three leads and three lags such that J = K = 3 . The binary variables j_3 and k_3 are equal to one for years in which the “Time to Event” surpasses 3 leads and lags for firms that start importing, respectively.

Firms that never started to import (never treated) are used as control firms. For these firms, leads and lags for the corresponding years are set to zero. The first lead in equation (2) serves as the reference base year^[7]. Consequently, leads and lags estimate the differences between treated and control firms in comparison to the baseline year differences. The assumption underlying this approach is that in the absence of treatment, the difference in the outcome variable between treated and control groups would have remained similar to the baseline. This is commonly referred to as the parallel trends assumption. As an additional robustness measure for selecting appropriate control firms, we will consider those that became importers last year in the panel (last-treated) as proposed by Callaway and Sant’Annac (2021) and Sun and Abraham (2021).

As mentioned earlier, the other common method to assess the impact of a treatment is to use DiD with a propensity score matching. The objective of the matching approach is to mitigate dissimilarities among firms by identifying, for each importing firm, a comparable firm that did not engage in importing (to approximate the non-observed counterfactual event). Conversely, the concept of DiD aims to eliminate the influence of unobservable firm-specific effects. The DID-PSM proceeds through the following steps: Conditional on a set of firm characteristics, the firm’s probability (referred to as propensity score) to initiate importing is estimated using the probit model as described by equation (A1) in the Appendix.

Subsequently, once the propensity scores are computed, the nearest control firms for which the propensity score falls within a pre-defined radius can be selected as matches for the importing firm.^[8] Moreover, it is essential to verify the balancing condition – ensuring that each independent variable does not significantly differ between importing and non-importing firms – and the so-called common support condition^[9], which ensures that firms with identical X values possess a positive probability of being both importing and non-importing firms.

The difference-in-differences matching estimator, β 1 , can be formulated as follows:

Here, p i and p j , generated by using equation (A1), represent the predicted probability for firm i, which belongs to the treated group (importing firms), to be an importer, and the predicted probability for firm j in the control group (non-importing firms) to be an importer, respectively.

Δ y is the log difference between the level of the outcome y before and after the import status. The function g(.) assigns weights to the comparison firm j during the construction of the counterfactual for importing firm i. In the case of nearest neighbor matching, as applied in this analysis, g(.) equals one for the pair with the smallest difference between p _i and p _j , while it is equal to zero for all other pairs. Additionally, w _i represents the weight used in forming the outcome distribution for the treated sample (1/N in the case of nearest neighbor matching).

4 Results

Before reporting the main results, we first summarize the results analyzing the role of ex-ante firm-specific characteristics in the decision for a firm to source inputs from abroad. Here, we refer to the Appendix for the empirical setup and the corresponding tables to save space. As revealed in Tables A1 and A2, there is compelling evidence that superior firm-specific characteristics are necessary for firms to become importers. The results clearly demonstrate a ranking order where firms that are most productive, largest, and have highest average wage costs are more likely to import from all countries, followed by firms exclusively import from either low- or high-wage countries, and finally, firms that do not import. The findings also suggest that among importers, particularly for small firms, the level of capital stock is a relatively more important determinant for sourcing from either low- or high-wage countries. Given these results, it is imperative to account for the self-selection mechanism when evaluating the effects of imports. Without such consideration, the estimate of the causal effect could potentially be biased, as discussed in Section 3.

Moving now to the results from Event study and DiD-PSM. In Figure 1, we illustrate the effect of imports on productivity levels using equation (2). As discussed earlier, our analysis includes three lags and three leads representing time before and after the firm started importing. Therefore, the binary variables t−3, t−2, and t−1, on the x-axis are set to one for three, two, and one years before the firm´s import initiation at time t0 (base year), while t+1, t+2, and t+3 are set to one for the years after the import. The differences in productivity levels between importing and non-importing firms, displayed on the y-axis, are presented by point estimates and 95% confidence intervals for each of the three lags and leads. As discussed before, we also include base year firm-specific variables as included in the probit model, equation (A1), and control for firm and time-fixed effects.

Figure 1

Productivity differences between importing and non-importing firms. Notes: Source: Statistics Sweden. The sample include all Swedish manufacturing firms with at least 10 employees. Firms fewer than 50 employees are defined as small firms while firms with 50 employees or more are defined as large firms. The upper-left corner presents the results of estimating equation (2) for all firms, the upper-middle solely for small firms and upper-right corner solely for large firms. The lower plots present the results when import originates exclusively from high-wage countries in lower-left corner, exclusively from low-wage countries in the lower-middle, and from all countries in lower-right corner. On the x-axis, t−3, t−2 and t−1 are binary variables equals to one for three, two and one year before the firm´s import initiation at time t0 (base year), and t+1, t+2 and t+3 are set to one for the years after the import. On the y-axis, the differences in the level of the outcome between importing and non-importing firms are presented by point estimates and 95% confidence intervals for each of the three lags and leads. The model includes base year firm-specific variables (as in Table A1) and control for firm and time fixed effects.

The results of estimating equation (2) for all firms without distinguishing their size or import source are shown in the upper-left corner of Figure 1. It can be observed that the pre-trends in productivity levels, measured as log sales per employees, do not appear to differ between importing and non-importing firms^[10]. This suggests that the key assumption for the validity of the parallel trends is satisfied. Additionally, the point estimates for periods following the base year t0 are not significantly different from zero, indicating that, on average, importing and non-importing firms experience similar productivity levels over the years.

In the subsequent step, we proceed to analyze whether the impact of imports on productivity depends on firm size and import source. The upper plots of Figure 1 present the results of estimating equation (2) solely for small firms in the upper-middle and solely for large firms in the upper-right corner. The lower plots of Figure 1 present the results when import originates exclusively from high-wage countries in the lower-left corner, exclusively from low-wage countries in the lower-middle, and from all countries in the lower-right corner. In all cases, the productivity differences between pre- and post-periods for importing firms are compared to productivity differences for non-importing firms. It can be observed that, regardless of firm size and import origin, there are no significant differences in the productivity levels between importing and non-importing firms in the periods before or after the base year t0.

Thus far, the results from the event study suggest that sourcing from abroad does not influence the differences between importing and non-importing firms, at least in terms of productivity levels. Consequently, we continue our analysis to determine whether sourcing from abroad impacts another crucial outcome for firms, namely export sales. Figure 2 presents the re-estimated equation (2), with log export sales as the dependent variable. As before, we include similar base year firm-specific variables and control for firm and time-fixed effects. The plot arrangement in Figure 2 mirrors that of Figure 1, with upper plots dividing the results by different firm sizes, and lower plots by different import origins. The results once again suggest that, regardless of firm size or import source, importing firms experience similar levels of export sales as non-importing firms.

Figure 2

Export sales differences between importing and non-importing firms. Notes: see Figure 1.

In summary, the results from the Event Study indicate that import does not have a statistically significant impact on firms’ productivity and export sales, whether we control for firm size or import origin. However, is this result robust when simultaneously controlling for firm size and import origin? In other words, how do productivity and export sales change for small and large firms, respectively, while accounting for import origin? Moreover, how does the number of export markets these firms engage in change based on the origin of their imports? To answer these questions, we proceed to implement one of the most commonly used evaluation methods in the field, the DiD propensity score-matching by using equation (3). As mentioned before, the matched non-importing firms are used to approximate unobserved counterfactual events, that is, what would have happened to the productivity and export performance of importing firms on average if they had not initiated imports.

Table 7 presents the DiD matching estimator, β 1 from equation (3). In specification (1) the matched sample consists solely of small firms, with small non-importers as the reference group. Specification (2) includes only large firms, with large non-importers as the reference group. The matched sample in each column is generated by dividing the binary indicator in equation (A1) into three separate indicators. Therefore, in column (1), the dependent variable in equation (1) is set to one for firms importing exclusively from high-wage countries, column (2) it is set to one for firms importing exclusively from low-wage countries, and column (3) it is equal to one for firms importing from all countries.

The result in Table 7, Specification (1) and column (1), indicates that small firms importing exclusively from high-wage countries experience, on average, a 0.9% points higher growth in productivity compared to small non-importing firms. In column (3), this difference increases to about 1.5% points for small firms importing from all countries, in comparison to small non-importers. However, small firms that exclusively import from low-wage countries do not appear to exhibit different productivity growth compared to small non-importers, as shown in specification (1); column (2). The results in specification (2) for larger firms also indicate that importing firms, regardless of import origin, experience similar productivity growth as non-importing large firms.

To summarize the findings of the DiD-PSM thus far, we can conclude that importing firms tend to have higher ex-ante productivity levels as compared to non-importing firms. Upon accounting for this, it appears that larger importing firms exhibit similar post-import productivity development to their non-importing counterparts. For small firms, however, there seems to be a slight learning effect from importing in terms of higher productivity, particularly when importing from high-wage countries, albeit marginally more so than from low-wage countries.

We now proceed to present the results of how imports affect export performances, as shown in Table 8. This analysis excludes non-exporting firms. The first three columns depict the results for export sales. Columns (4)–(6) and (7)–(9) present the results for how imports impact the total number of export countries and the total number of high-wage export counties, respectively. The structures of the two specifications and the division of the import indicators in the respective columns mirror those in Table 7. Controlling for firm size and import origin, the export sales do not seem to differ between importing and non-importing firms. Moreover, relative to small non-importing firms, small firms concentrating their import from low-wage countries appear to leave high-wage export markets to a greater extent. Simultaneously, small firms importing from both low- and high-wage countries export to on average 8 more high-wage countries than their non-importing counterparts. Similar trends are observed for large firms importing from all countries, which export to an average of 3 more high-wage countries compared to large non-importing firms. However, large firms that primarily import from high-wage countries seem to reduce their participation in high-wage export markets by over 1 market on average, in contrast to large non-importers.

5 Summary and Conclusions

There is substantial evidence underscoring the significance of global trade in driving economic growth, particularly benefiting firms in small open economies by providing access to larger markets. However, there is also evidence that only the most productive firms enter the export market, and they rely heavily on global sourcing due to the increasing fragmentation of production processes. This article aims to explore the relationship between imports, productivity, and export performance. Our initial empirical exploration delves into the relevant firm-level characteristics required for firms to engage in foreign sourcing, as well as the origins of their sourcing. Subsequently, we examine how the source of imports impacts firm productivity and its connection to export sales and the number of export markets.

The data utilized in this article cover the entire Swedish manufacturing firms with at least 10 employees over the period 2007–2020. We adopt the most recently developed event study approach to account for heterogeneous treatment effects and combine a DiD approach with a PSM.

This article underscores several key findings. First, nearly 80% of the sampled firms engage in international trade, with smaller firms predominantly functioning as exporters, mid-sized firms as importers, and larger firms engaging in two-way trading. Second, while imports predominantly originate from high-wage countries, there is a gradual shift toward low-wage countries over time. Third, a clear hierarchy emerges regarding firms’ import behaviors based on productivity levels, with high-productive firms importing from all countries, followed by firms sourcing from low- or high-wage countries, and non-importing firms. Among small importing firms, the level of capital stock appears as a crucial determinant, particularly when imports are concentrated in low- or high-wage countries. Fourth, accounting for self-selection mechanisms, we find no significant differences in productivity and export sales growth between importing and non-importing larger firms. However, small importing firms experience higher productivity growth, primarily driven by imports from high-wage countries. Additionally, both small and large firms importing from both high- and low-wage countries tend to access more high-wage export markets.

As a final remark, the findings of this article provide significant policy implications, suggesting the importance of promoting access to imported intermediate inputs, especially from technologically advanced countries, to enhance domestic productivity and export potential. Policies should target small firms not currently engaged in international markets and those exclusively importing or exporting, aiming to improve their competitiveness. Encouraging diversification of import sources and aiming for favorable terms for the import of high-quality inputs during trade negotiations can further enable firms to exploit the benefits of global sourcing.