Industry comparative advantage and support for redistribution: A cross-country cross-industry analysis of the political economy of trade

Abstract

A fundamental insight of various trade theories is that trade does not have a universally negative effect on different business activities in different countries. Rather, trade’s impact varies concomitantly with the specific country and activity considered. This empirical note expands prior work linking trade to redistribution preferences by using sectoral comparative advantage to incorporate the notion that trade may hurt the prospects of a specific group in one country (e.g. workers in a highly tradeable or offshorable industry) but will simultaneously benefit this same group in another country. We expect that individuals in industries with a weaker (stronger) comparative advantage suffer (benefit) more from trade and are therefore more (less) in favour of redistribution. Empirical results confirm this expected effect of comparative advantage on redistribution preferences. We conclude that considering countries’ comparative (dis)advantage in certain activities provides a deeper and more general understanding of the political consequences of trade.

Keywords

Preferences for redistribution cross-country cross-industry analysis international competitiveness trade exposure gains from trade

Introduction

Individuals’ (perceived) economic prospects, positive or negative, affect their stance on redistribution and other key political issues and policies. Economic prospects and perceptions, in turn, are affected by a variety of factors, not least globalization (Kriesi et al., 2006). Much research has considered the political consequences of trade-induced economic anxieties and vulnerabilities, seeking to identify risk factors that link international trade to individuals’ support for redistribution (and to the demand-side of politics more generally). Some prototypical empirical results in this literature are that individuals are more supportive of government redistribution when working in a sector or occupation that is tradeable and thus exposed to foreign competition (Rueda and Stegmueller, 2019).

However, so far, this literature is asymmetrically focused on downside risks associated with increased competition from foreign imports and does not incorporate one of the core insights of standard trade theories, which is that trade may hurt the prospects of a specific group of workers in one country (e.g. workers in a highly offshorable industry or occupation) but that trade will simultaneously benefit this same group in another country. Both old and modern theories find that trade can worsen the prospects of some types of activities – for example, a specific industry (Industry 1) – in a given country (Country X). However, when an industry in a given country (Industry 1 in country X) suffers from foreign competition there will always be another country (Country Y) in which the same industry (Industry 1) flourishes because trade enables it to sell and compete internationally. The degree to which a specific industry’s exposure to trade, tradeability or offshorability poses a downside risk is thus not a characteristic that is inherent to an industry and that holds universally, regardless of where the industry is located. Trade entails both downside and upside risks that vary concomitantly with the specific country and the specific industry considered.¹

This paper seeks to expand prior literature linking globalization to the demand-side of politics to incorporate the notion that there is no industry (or other such units – for example, occupations, tasks or firms) in which all workers suffer from trade, irrespective of their location. Because of robust theoretical insights from standard trade theories and data availability, I focus on the effect of industries’ comparative advantage on individuals’ support for redistribution. Industry comparative advantage indicates how well a given industry in a country is performing compared to the same industry in other countries and compared to other industries in the same country. Hence, considering industry comparative advantage seems a very suitable way to capture the idea that trade can be a threat for some types of business activities in some locations but an opportunity for other types or in other locations. The specific prediction that I test is that, ceteris paribus, individuals in an industry in which their country has a stronger (weaker) comparative advantage are less (more) supportive of redistribution. Because comparative advantage varies with country and industry concomitantly, the main unit of analysis is country–industry combinations.

As it is, only a few empirical studies have acknowledged that comparative advantage likely plays a role in shaping redistribution preferences. Rehm (2009) considers whether an industry is a net exporter or net importer and Walter (2017) considers whether an industry is export-oriented (yes/no) and whether it is import-competing (yes/no). However, these two papers do not identify country–industry combinations as the unit of analysis and remain unclear on the relevant number of observations. In addition, their measures are crude dummy variables that do not capture the full extent of industries’ comparative (dis)advantage. Finally, and most problematically, these papers do not control for industry fixed effects (FEs) and therefore do not distinguish between effects due to unobservable industry-specific factors, including such traits as offshorability but also health and safety risks, and effects genuinely due to industries’ comparative advantage.²

For the main empirical analyses, I consider the redistribution preferences of 31,790 individuals in 503 unique country–industry combinations. The relevant variation is the cross-sectional variation in support for redistribution that occurs between these country–industry combinations, which is about 12.2% of total variation in individuals’ redistribution support. The results support my prediction, indicating that individuals in sectors with a stronger (weaker) comparative advantage are significantly less (more) supportive of redistribution. Importantly, this negative empirical relationship exists completely independent from direct effects due to industries’ or occupations’ exposure to trade or any other fixed sectoral or occupational characteristic that is extensively considered in prior research. Of course, I do not claim or even attempt to show that direct effects of industries’ and occupations’ tradeability or other such industry- and occupation-specific influences are unimportant. However, the found added explanatory power of industries’ comparative advantage contributes a useful empirical insight to the large literature linking trade to the demand-side of politics. I thus conclude that understanding of the political economy of international trade can improve from explicit consideration of countries’ comparative advantage in specific types of business activities.

Background

Economic status and prospects and demand for government redistribution

Government policies often involve redistributing material wealth and demand for redistribution is extensively studied (Iversen and Goplerud, 2018). Considering demand for redistribution, political economy studies often refer to theoretical models that emphasize the role of individuals’ current economic status, specifically their income relative to the median (e.g. Romer, 1975). However, scholars are increasingly extending this narrow view with a more nuanced view that adds a role for individuals’ expected economic status as well as anxiety about possible future economic hardship. Individuals’ economic prospects are subject to external shocks such as globalization, and redistribution can work as a social insurance against the downside risks of such shocks (Kato and Takesue, 2023; Moene and Wallerstein, 2001). Similarly, individual expectations of upward instead of downward economic mobility (e.g. due to a positive rather than a negative trade shock) reduce rather than increase demand for government redistribution (Kim, 2023).

Industry comparative advantage and redistribution preferences

As indicated, trade theories – old and new – agree that trade’s costs and benefits for workers depend simultaneously on the specific business activity and on the specific country considered. Since Ricardo’s classic study of industry comparative advantage is so well known, I use that work as the theoretical backdrop to my analysis. Having a comparative advantage means that an entity (e.g. a firm, a region, a country or an industry) can perform a certain activity at relatively lower cost than another entity. However, a comparative advantage in one activity coincides with having a comparative disadvantage in another activity. Comparative advantage affects the prospects of different industries in different countries under trade. When trade occurs, industries (or other such groups or types of activities) in which countries have a comparative disadvantage dwindle, whereas the industries (or other such types of activities) in which countries have a comparative advantage expand. In the real world, economies comprise multiple sectors and economies have comparatively weaker or stronger comparative advantages in each type of sectoral activity. Germany, for example, has a much stronger comparative advantage in the manufacture of motor vehicles, trailers and semi-trailers than in fishing. Trade openness matters because under trade the former industry would have much potential for growth as it can export to other countries. Conversely, the latter industry’s future would be bleaker as it struggles to compete with foreign imports. These latter imports may come from Greece, which means that its fishing industry benefits from trade openness, while, at the same time, Greece’s motor vehicles industry might be struggling to compete with German imports. In short, whether a country has a comparative (dis)advantage in a particular industry links to the economic prospects of this industry in the country and, hence, the economic prospects of citizens in the industry. Other trade theories make the same prediction on the country-specificity of the costs and benefits of trade for specific groups of workers but may focus attention on units other than industries; for example, occupations, tasks or firms.

Following the above argument, I expect that weaker/stronger industry comparative advantage increases/decreases support for government redistribution by individuals in the industry. On average, individuals in industries with a strong comparative advantage likely have a better labour market position and better economic prospects. Hence, these individuals likely have a higher expected lifetime income, which reduces the net benefits that they receive and can expect to receive from government redistribution. In addition, being aware of how the comparative advantage of their sector of employment affects their economic prospects and the corresponding expected net benefits of redistribution, self-interest suggests that individuals are less supportive of redistribution the stronger the comparative advantage of their industries is.³ At the same time, I recognize that if labour were perfectly mobile across sectors, workers in industries that dwindle under trade are re-employed in other industries instantly and without any cost to these workers (Owen, 2015). Empirical evidence, however, shows that labour is remarkably immobile and that adjusting to trade shocks takes at least a decade (see Autor et al., 2016 for a survey). Hence, I expect that their industries’ comparative advantage does indeed affect individuals’ economic prospects, including the future cost of labour market adjustment, and hence their redistribution preferences. Ukrainian workers, for instance, may benefit from German firms offshoring manufacturing of basic metals to Ukraine. Conversely, German workers benefit from the export of medical, precision and optical instruments to Ukraine, which puts pressure on the economic prospects of Ukrainian firms and workers active in this industry. Still, I should note that, as emphasized by ‘new new trade theory’, sectoral comparative advantage is hardly the only factor affecting trade’s cost and benefits for workers employed in the sector. In fact, I readily acknowledge that, within industries, trade likely hurts/fosters the economic prospects of some groups more/less. Similarly, my focus on cross-country cross-industry variation is not meant to suggest that countries cannot have a comparative advantage in specific tasks or that having a comparative advantage in certain tasks has a weaker effect on redistribution preferences than having a comparative advantage in certain industries does.

Data and method

The main data source is the European Social Survey (ESS; 2020), which is often used to study individuals’ preferences for various policies (e.g. Arikan and Ben-Nun Bloom, 2013; Stadelmann-Steffen and Eder, 2021). I complement these data with cross-country cross-industry data on so-called revealed comparative advantage (RCA) from the Organisation for Economic Co-operation and Development (OECD)’s (2010) STAN Structural Analysis database. The ESS is a biennial cross-national survey collecting individual-level data for representative samples from 32 European/Eurasian countries. The data are repeated cross-sections, and questions asked during the interviews concern personal attitudes, occupation and sector of employment, among others. The OECD STAN database contains data on economic characteristics of different industries across OECD member countries.

RCA data are available for the period 1999–2008. Although ESS data are available for the period after 2008, because of comparability of the industry classifications used in the ESS and the OECD data, I only consider data from the first four waves of the ESS (2002–2008).⁴ Following prior research (e.g. Van Hoorn, 2022), I further exclude respondents with missing data on the control variables that I consider. Matching the RCA data to the ESS data leaves a main sample of 31,790 respondents nested in 503 country–industry combinations. Table 1 presents descriptive statistics for and details on selected variables. Tables S1 and S2 in the online supplemental material list the 21 countries and 25 industries in the sample. Table S3 lists the industries in which the countries in the sample have the strongest/weakest RCA. Although data have been collected over multiple years, the data are not an individual panel and I do not consider time-series variation in RCA. Figure S1 presents the structure of the data, emphasizing how individuals are nested in country–industry combinations, which are the main unit of analysis. Importantly, the relevant variation is not variation between countries or between industries but variation in redistribution preferences that is between country–industry combinations. As mentioned in the introduction, this country–industry variation is 12.2% (95% confidence interval (CI): 10.5, 14.0%; n = 503; n = 31,790) of total variation in individuals’ support for redistribution.

Table 1.

Descriptive statistics for and details on selected variables.

Variable	Mean (SD)	Measurement
Support for redistribution	3.81 (1.06)	See main text. Measured using the item that asks respondents about the extent to which they agree or disagree with the statement that ‘the government should take measures to reduce differences in income levels.’ Possible answers are given by the following Likert-type scale: 1 = Agree strongly; 2 = Agree; 3 = Neither agree nor disagree; 4 = Disagree; or 5 = Disagree strongly. We recode these answers to reflect support for redistribution. This item is a standard measure of redistribution support and widely used. Following prior work, we analyse this measure as though it is a cardinal or continuous measure rather than an ordinal measure. This assumption of cardinality is supported by an extensive literature (e.g. Norman 2010; Robitzsch 2020). However, to be complete, we also analysed the item using ordered logit and ordered probit techniques. Results, presented in Table S6 in the supplemental material, confirm the results of the main model, indicating that the cardinality assumption does not cause substantial biases.
Industry	Categorical	Based on the item asking ‘What does/did the firm/organisation you work/worked for mainly make or do?’, ESS coded responses into two-digit NACE/ISIC codes. As mentioned in the text, Waves 1–4 (2002–2008) of the ESS, which are the waves in the main analyses, code respondents’ industries using NACE revision 1/1.1. For the robustness checks that involve samples that include data from Waves 5–9 (2010–2018) we convert respondents’ industries coded using NACE revision 2 to NACE revision 1.1. The crosswalk that we develop newly for this purpose is presented in Table S12 in the supplemental material.
Industry comparative advantage (natural log)	4.46 (0.983)	See main text. Measured using Balassa’s (1965) famous index of revealed comparative advantage (RCA). Data are from the OECD STAN Structural Analysis database. Whereas comparative advantage is known to be stable or ‘sticky’ (e.g. Dalum et al., 1998; Frantzen, 2008), the Balassa index is known to have low temporal stability (e.g. De Benedictis and Tamberi, 2004; Hinloopen and Van Marrewijk, 2001; Leromain and Orefice, 2014). Because the RCA data cover the period 1999–2008, we are able to address this instability and reduce measurement error due to idiosyncratic shocks by calculating average RCA scores over this entre period (see also Van Hoorn, 2017). The 1999–2008 period also matches the period of data collection of the ESS data used for the main analyses.
Education years	11.3 (4.24)	From the item asking ‘About how many years of education have you completed, whether full-time or part-time?’
Percentile rank of years of education	45.3 (28.5)	Constructed using the above item on years of education and recoding individual answers into a percentile score that ranks respondents’ years of education relative to the years of education of respondents from the same country surveyed in the same year. The higher the percentile score, the more individuals in the country have fewer years of education than the respondent has and, hence, the scarcer the individual’s level of education is in the country.
Religiosity	4.95 (2.97)	From the item asking ‘Regardless of whether you belong to a particular religion, how religious would you say you are?’ (0 = Not at all religious – 10 = Very religious).
Female (1 = yes)	44.6% (49.7%)	From the item coding the sex of the respondent.
Educational degree	Categorical	Harmonized measure constructed by the ESS from country-specific answers on item asking about educational degree. categories: Not possible to harmonize; I = less than lower secondary; II = lower secondary; IIIb = lower tier upper secondary; IIIa = upper tier upper secondary; IV = advanced vocational, sub-degree; V1 = lower tertiary education, BA level; V2 = higher tertiary education, >= MA level.
Percentile rank of tertiary education	48.8 (13.6)	Constructed using the above item on educational degree to create a dummy variable that has value 1 if the respondent has lower or higher tertiary education and value 0 if not. This dummy variable is then recoded into a percentile score that ranks respondents’ having of tertiary education relative to respondents from the same country surveyed in the same year. The higher the percentile score, the more individuals in the country do not have any tertiary education and, hence, the scarcer the individual’s tertiary education is in the country.
Union membership	Categorical	Based on the item asking ‘Are you or have you ever been a member of a trade union or similar organization?’ Categories: Yes, currently; Yes, previously; No.
Domicile	Categorical	Based on the item asking ‘Which phrase on this card best describes the area where you live?’ Categories: A big city; Suburbs or outskirts of big city; Town or small city; Country village; Farm or home in countryside.
Occupation	Categorical	Based on the item asking respondents to describe key features of their jobs. ESS coded responses into 4-digit 1988 ISCO codes for a total of 469 different occupations. Waves 6–9 of the ESS, which we consider for robustness checks involving extended samples, code respondents’ occupation using 2008 ISCO codes. For these robustness checks, we recode ISCO08 occupations into ISCO88 occupations using the crosswalk developed by Ganzeboom and Treiman (2015).
Country	Categorical	Based on administrative data provided by the ESS.
Wave	Categorical	Based on administrative data provided by the ESS.

ESS: European Social Survey; OECD: Organisation for Economic Co-operation and Development;

NACE: Nomenclature of Economic Activities; ISIC: International Standard Industry Classification; ISCO: International Standard Classification of Occupations.

Variables and measures

I measure support for government redistribution using the ESS item asking respondents to indicate their agreement with the statement that ‘the government should take measures to reduce differences in income levels’ (1 = Agree strongly; 2 = Agree; 3 = Neither agree nor disagree; 4 = Disagree; 5 = Disagree strongly.) This and similar single-item scales have been used in many studies of the demand-side of government redistribution (e.g. Yang and Kwon, 2021). For the empirical analysis I reverse the scale to measure support for redistribution. To facilitate interpretation of the empirical results, I analyse this variable as a continuous variable. I obtain similar results when I estimate ordered logit or probit models (Table S6).

I measure the independent variable, comparative advantage, using the famous and widely used Balassa (1965) index. This index concerns the RCA of a particular country (c) in a particular industry (i):

{RCA}_{c i} = 100 \times \frac{X_{c i} / X_{c}}{X_{i} / X} .

(1)

Here X_ci thereby denotes exports of country c in industry i, X_c denotes total exports of country c across all industries, X_i denotes total exports of industry i across all other countries, and X denotes total exports across all other countries. RCA is not a measure of trade or trade flows and is unaffected by the size of countries or industries. Mathematically, an industry’s RCA can vary between 0 and infinity and it is impossible for all industries in a country to have RCA > 100. My sample comprises only European countries. Nevertheless, in all countries there are both a number of industries that have a comparative disadvantage (RCA < 100) and a number of industries that have a comparative advantage (RCA > 100) (Table S4). More generally, several economies in the sample, notably Germany, export more in absolute monetary amounts than they import.

Although a country’s comparative advantage in a specific industry is a structural feature of its economy, external shocks can cause random, non-structural changes in the Balassa index scores across years (e.g. geopolitical events affecting oil prices). In general, a well-known limitation of the Balassa index is its low temporal stability (Hinloopen and Van Marrewijk, 2001). I therefore do not use yearly RCA scores but reduce measurement error by averaging scores across all the years for which there are RCA data (1999–2008) (see e.g. Van Hoorn, 2017). This way, I ensure that I am capturing RCA as a relatively ‘sticky’ feature of an industry in a country (Dalum et al., 1998).

I match the RCA data to the individual-level data from the ESS using country identifiers and two-digit NACE (Nomenclature of Economic Activities)/ISIC (International Standard Industry Classification) industry codes. Waves 1–3 of the ESS (2002–2006) have used revision 1 of the NACE classification, which is equal to NACE revision 1.1, which was used in Wave 4 (2008). NACE revision 1/1.1 is further equal to revision 3 of the ISIC, which is the classification used by the OECD STAN database. The 2-digit NACE/ISIC classification distinguishes 88 industries. Because RCA data are lacking for industries with few or no exports (mostly services-oriented industries), the main sample comprises 25 industries. I obtain similar results if I assign these non-exporting industries an RCA score of 100 for all countries and add them to the sample (Table S7).

The analysis considers various control variables. Following standard cross-country cross-industry analyses of RCA (Nunn and Trefler, 2014) I include both country and industry FEs as a means to focus on cross-country and cross-industry variation simultaneously and control for confounding country- and industry-specific effects. These FEs control for the potential confounding effect of country-specific factors such as welfare state regimes and industry-specific factors such as working conditions or tradeability as well as unit-specific measurement error. Because prior work linking trade to redistribution preferences emphasizes the role of occupation, all models also include occupation FEs. This way I address any potentially confounding effects of occupational traits such as tradeability, routine-task-intensity, automatability, etc. I also include wave FEs to control for – for example – year-specific measurement error as well as individual-level control variables commonly considered in studies of redistribution preferences. These individual controls involve standard demographics (age, age squared, gender and domicile) but also individuals’ religiosity and union membership. I further use both individuals’ years of education and dummies indicating education level to control for differences in skills and formal training. Finally, I add two more variables that speak to an individual’s formal and informal skills and competences relative to other people in their country. The first of these variables is the percentile rank of individuals’ years of education relative to fellow citizens. The rationale is that this percentile rank captures how scarce the individual’s skills and competences are in the country. The second one is the percentile rank of having tertiary education or not, which captures how scarce the individual’s tertiary education, if any, is. For one of the robustness checks, I further follow Van Hoorn (2022) and use variables that are also partly outcomes of RCA to control for potential confounders that can affect individuals’ redistribution preferences and also correlate with working in a certain industry, particularly individual ability differences (see note 3). Some key outcome variables that I add as controls are individuals’ employment status and the rank of their household income (both are proxies for individual ability and effectuations of the labour market implications of RCA), and left/right political preferences. Statisticians agree that such endogenous controls are bad controls and that they have an important disadvantage, which is that they cause a so-called included variable bias (Cinelli et al., 2022). I therefore do not include these controls in the main models and only consider them as powerful means for getting rid of omitted variable bias and a strict check of the robustness of the main results.

Method

I test the relationship between RCA and support for redistribution using regression analysis of individual-level survey data, which is standard in the literature linking macro-level phenomena such as trade to individuals’ preferences. Because data are structured hierarchically (individuals nested in unique country–industry combinations), I use cluster-robust standard errors. I obtain similar results when estimating (cross-classified) multilevel models that separate variation between and within country–industry combinations (Table S8; see also Figure S1). The basic empirical model concerns the support for redistribution (S_hci) of individual h that lives in country c and works in industry i:

S_{h c i} = β_{0} + β_{1} {RCA}_{c i} + β_{2} X_{h} + D_{c} + δ_{i} + θ_{w} + ε_{h c i}

(2)

Here RCA_ci denotes the comparative advantage of country c in industry i, X_h refers to individual characteristics such as occupation and education, and ε_hci is a random error term. We use D_c, δ_i and θ_w to denote country (c), industry (i) and wave (w) FEs. The relevant variation is the variation in redistribution preferences that occurs between country–industry combinations (ci), where each country–industry unit comprises about 63 individuals on average (see also Figure S1).

Results

Table 2 lists the main results. The simplest model that I estimate controls for country and industry FEs as well as occupation and wave FEs as potential confounders, thus focusing attention on variation in redistribution preferences among individuals from different countries working in different industries. Consistent with my expectation, results indicate a significant negative relationship between an industry’s RCA, which is country-specific, and support for redistribution (Model 1). Because I control for both industry and occupation FEs, this apparent effect exists in addition to the direct effect of industry- and occupation-specific trade exposure, import competition, tradeability and other such influences emphasized by prior work. Adding years of education, education level, union membership and other individual controls renders a closely similar estimated coefficient (Model 2). Hence, it seems that the found relationship between RCA and redistribution preferences is not driven by educational differences, among others. Comparing coefficients, which are standardized, the effect size of RCA appears to be in the same league as that of other important determinants of redistribution preferences identified in the literature.

Table 2.

Industry comparative advantage and the demand-side of redistribution.

Dependent = Support for government redistribution	Model 1	Model 2	Models that include potential outcomes of comparative advantage as controls (endogenous controls)
Dependent = Support for government redistribution	Model 1	Model 2	Model 3	Model 4
Industry comparative advantage (natural log)	-0.039 (0.007) [p = .000]	-0.038 (0.007) [p = .000]	-0.031 (0.008) [p = .000]	-0.030 (0.008) [p = .000]
Education years	-	0.069 (0.021) [p = .001]	0.046 (0.026) [p = .075]	0.046 (0.025) [p = .073]
Percentile rank of years of education	-	-0.140 (0.018) [p = .000]	-0.117 (0.023) [p = .000]	-0.113 (0.023) [p = .000]
Percentile rank of tertiary education	-	-0.035 (0.101) [p = .727]	0.002 (0.125) [p = .989]	-0.001 (0.121) [p = .992]
Female (1 = yes)	-	0.124 (0.016) [p = .000]	0.118 (0.016) [p = .000]	-0.010 (0.008) [p = .204]
Union membership (reference = yes)
Previously	-	-0.015 (0.020)[p = .457]	-0.047 (0.022)[p = .034]	-0.037 (0.021)[p = .073]
No	-	-0.136 (0.020)[p = .000]	-0.176 (0.022)[p = .000]	-0.133 (0.020)[p = .000]
Religiosity	-	-0.031 (0.007)[p = .000]	-0.037 (0.008)[p = .000]	0.094 (0.016)[p = .000]
Income rank	-	-	-0.104 (0.009)[p = .000]	-0.087 (0.008)[p = .000]
Left/right political preferences	-	-	-	-0.170 (0.012)[p = .000]
Unemployment experience (1 = yes)	-	-	-	0.083 (0.014)[p = .000]
Hours contracted	-	-	-	0.033 (0.012)[p = .006]
Hours worked	-	-	-	-0.038 (0.012)[p = .001]
Employment status (dummies)	No	No	Yes	Yes
Age and age squared	No	Yes	Yes	Yes
Educational degree (dummies)	No	Yes	Yes	Yes
Domicile (dummies)	No	Yes	Yes	Yes
Occupation FEs (4-digit)	Yes	Yes	Yes	Yes
Industry FEs (2-digit)	Yes	Yes	Yes	Yes
Country FEs	Yes	Yes	Yes	Yes
Wave FEs	Yes	Yes	Yes	Yes
No. of country–industry combinations	503	503	493	493
No. of individuals	31,790	31,790	20,621	20,621
R ²	0.1685	0.1878	0.2055	0.2330

Standard errors (in parentheses) and p-values are clustered at the level of country–industry combinations. The dependent and continuous independent variables are standardized with mean = 0 and SD = 1.

FE: fixed effects.

country–industryI have assessed the robustness of the baseline results in different ways (e.g. Tables S6–S11). Here I highlight three main robustness checks. First, I have included potential outcome variables as added controls. As indicated, I have done this because it provides further means to rule out that the apparent relationship between industry RCA and support for redistribution is spurious. Model 3 in Table 2 adds income rank and employment status as control variables that speak to individuals’ actual or realized economic status. These variables proxy for ability differences but are also partly an effectuation of the impact that their industry’s RCA has on individuals’ economic prospects. Nevertheless, results indicate a significant negative relationship between RCA and support for redistribution. Model 4 further controls for left/right political preferences, among others. Left/right political preferences are closely related to preferences for redistribution (Thewissen and Rueda, 2019). Again, however, results confirm the main results. Second, I have broadened the analysis to allow the impact of RCA on support for redistribution to vary across the countries in the sample. Models 5 and 6 in Table 3 presents the results. Results are again similar to the baseline results. Hence, it seems that any bias due to non-homogeneity among the countries studied is minor. Third, I have extended the main sample, which includes ESS data covering the years 2002–2008, to include data covering the years 2010–2018 as well. Downside of this extension is that data are subject to more measurement error and noise. RCA data do not go beyond 2008 and later waves of the ESS use a different classification system for coding respondents’ industries (see Table 1 and above), which means that I need to develop and apply a crosswalk (see Table S12).⁵ Hence, the extended samples are not my preferred samples. Results (Models 7 and 8 in Table 3) reveal the same negative relationship between RCA and support for redistribution as before. However, model fit (R²) is lower than before (Table 2), which is as expected given the period mismatch between the RCA data and later waves of the ESS and the imperfect recoding of respondents’ industries. Similarly, estimated coefficients, though still statistically significant at usual levels, are lower, and this is likely caused by attenuation bias due to measurement error.

Table 3.

Results for selected robustness checks.

Dependent = Support for government redistribution	Country heterogeneity in the effect of revealed comparative advantage (RCA) on redistribution preferences		Extended samples that include converted data from later waves of the European Social Survey (ESS)
	Model 5: Multilevel model of individuals nested in countries with slopes random over countries	Model 6: Multilevel model of individuals nested in countries with slopes random over countries and controls that are partly outcomes of comparative advantage	Model 7: Baseline model	Model 8: Model with added controls that are partly outcomes of comparative advantage
Industry comparative advantage (natural log)	-0.038 (0.007) [p = .000]	-0.030 (0.008) [p = .000]	-0.017 (0.006) [p = .002]	-0.013 (0.005) [p = .015]
Control variables from main model (Model 2)	Yes	No	Yes	No
Control variables from largest model with endogenous controls (Model 4)	No	Yes	No	Yes
No. of country–industry combinations	503	493	539	534
No. of countries	21	21	22	22
No. of individuals	31,790	20,621	70,237	47,385
Log likelihood	-41801.8	-26523.9	-	-
R ²	-	-	0.1543	0.2030

See Table 2 and Table S8 in the supplemental material. Standard errors are in parentheses; p-values are in square brackets. The dependent variable and continuous independent variables are standardized to have a mean of 0 and a standard deviation of 1. Models 7 and 8 pertain to samples that include data from Waves 5–9 (2010–2018) of the ESS and not just Waves 1–4 (2002–2008) (as in Models 1–6). Variables that the ESS codes differently across Waves 1–4 and Waves 5–9 have been recoded in the manner described in Table 1 and Table S12 in the supplemental material. As stated earlier, models with endogenous outcome variables as added controls are not our preferred models. The advantage of these models is that they go further in controlling for omitted variable bias, specifically an unobserved factor that affects both the comparative advantage of individuals’ industry and their support for redistribution. The disadvantage is that potential outcome variables are bad controls that cause an included variable bias.

Discussion and conclusion

The idea that international trade has distributional consequences and affects the demand-side of politics is widely accepted. Many studies have thereby considered how workers’ exposure to trade, operationalized by such measures as sectoral or occupational tradeability or offshorability, affects individuals’ policy preferences.⁶ However, old and modern theories agree that whereas trade can be a threat to a given type of activity in one country, the ability to sell abroad also offers opportunities, either for other types of activities in the country or for the same type of activity in another country.

This paper’s empirical results underscore the value of incorporating the fundamental theoretical insight that trade does not have a universally negative effect on different business activities in different countries when analysing the distributional consequences of trade and their impact on policy preferences. Focusing on cross-country cross-industry variation, I find a strong negative relationship between individuals’ support for redistribution and their industries’ comparative advantage (measured by their revealed comparative advantage or RCA). Because I control for any possible sectoral and occupational confounders (as well as many other potentially confounding factors), this relationship exists on top of the direct effect of such factors as industries’ and occupations’ tradeability or offshorability. Based on the added explanatory power of RCA, I conclude that considering how trade may hurt the prospects of a specific group in one country but will simultaneously benefit this same group in another country provides a deeper understanding of globalization shaping the demand-side of politics than heretofore.

A limitation of the paper is that I cannot consider the process through which individuals take note of their industry’s comparative advantage and use this information to assess their labour market position and economic prospects and ultimately the net benefits that they receive and expect to receive from government redistribution. Another limitation is that countries’ comparative advantages could only be measured at the two-digit industry level. Industry comparative advantage is, however, hardly the only form of comparative advantage likely to affect individuals’ support for redistribution. Workers may be employed by firms in the country that are internationally competitive, for instance. An interesting extension to the present analysis therefore is to consider comparative advantage or international competitiveness measured for different units, for example, for the firms that individuals work for. A further interesting extension is to link (industry) comparative advantage to support for a specific redistribution policy, namely one that explicitly aims to help workers who are hurt by trade. Finally, an interesting empirical extension would be to track individuals over time and link selection into and out of certain industries to support for redistribution. However, as it is, the individual-level panel data required for such a study are lacking.

Overall, the essential insight is that a more complete understanding of the political economy of trade comes from recognizing that the costs and benefits of trade for specific groups of workers are not universal but country-specific and that countries’ comparative advantage in certain types of activities is an important determinant of the political consequences of trade.

Supplemental Material

sj-docx-1-ips-10.1177_01925121241242440 – Supplemental material for Industry comparative advantage and support for redistribution: A cross-country cross-industry analysis of the political economy of trade

Supplemental material, sj-docx-1-ips-10.1177_01925121241242440 for Industry comparative advantage and support for redistribution: A cross-country cross-industry analysis of the political economy of trade by André van Hoorn in International Political Science Review

Supplemental Material

sj-zip-2-ips-10.1177_01925121241242440 – Supplemental material for Industry comparative advantage and support for redistribution: A cross-country cross-industry analysis of the political economy of trade

Supplemental material, sj-zip-2-ips-10.1177_01925121241242440 for Industry comparative advantage and support for redistribution: A cross-country cross-industry analysis of the political economy of trade by André van Hoorn in International Political Science Review

Footnotes

Data accessibility statement

The data that support the findings of this study and the statistical files needed to replicate the results are available as an online Appendix to this study.

Declaration of conflicting interests

The author declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.

Funding

The author received no financial support for the research, authorship and/or publication of this article.

ORCID iD

André van Hoorn

Supplemental material

Supplemental material for this article is available online.

Notes

Author biography

André van Hoorn is professor of economics at Radboud University, Nijmegen, the Netherlands. His research interests concern (dis)similarities and socioeconomic interactions between members of different social groups. Recent research topics include the effects of selection and socialization on finance professionals’ values and the effect of occupational automatability on political preferences.

References

Arikan

Gizem

Ben-Nun Bloom

Pazit

(2013) The Influence of Societal Values on Attitudes Towards Immigration. International Political Science Review 34(2): 210–226.

Autor

David H

Dorn

David

Hanson

Gordon H

(2016) The China Shock: Learning from labor-market adjustment to large changes in trade. Annual Review of Economics 8: 205–240.

Balassa

Bela

(1965) Trade Liberalisation and ‘Revealed’ Comparative Advantage. The Manchester School 33: 99–123.

Cinelli

Carlos

Forney

Andrew

Pearl

Judea

(2022) A Crash Course in Good and Bad Controls. Sociological Methods & Research. Epub ahead of print 20 May 2022. DOI: 10.1177/00491241221099552.

Dalum

Bent

Laursen

Keld

Villumsen

Gert

(1998) Structural Change in OECD Export Specialisation Patterns: De-specialisation and ‘stickiness’. International Review of Applied Economics 12(3): 423–443.

De Benedictis

Luca

Tamberi

Massimo

(2004) Overall Specialization Empirics: Techniques and applications. Open Economies Review 15: 323–346.

European Social Survey (2020) Cumulative File, ESS 1–9. Data file edition 1.0. Sikt – Norwegian Agency for Shared Services in Education and Research, Norway.

Frantzen

Dirk

(2008) Technology, Competitiveness and Specialisation in OECD Manufacturing. Journal of Economic Studies 35(1): 44–68.

Frieden

Jeffry A

(1991) Invested Interests: The politics of national economic policies in a world of global finance. International Organization 45: 425–451.

10.

Ganzeboom

Harry BG

Treiman

Donald J

(2015) International Stratification and Mobility File: Conversion tools. http://www.harryganzeboom.nl/ismf/index.htm

11.

Hinloopen

Jeroen

Van Marrewijk

Charles

(2001) On the Empirical Distribution of the Balassa Index. Weltwirtschaftliches Archiv 137(1): 1–35.

12.

Iversen

Torben

Goplerud

Max

(2018) Redistribution Without a Median Voter: Models of multidimensional politics. Annual Review of Political Science 21: 295–317.

13.

Kato

Junko

Takesue

Hiroki

(2023) The Presence of a Social Context Increases Support for Redistribution: Inequality aversion and risk aversion. International Political Science Review. Epub ahead of print 21 May 2023. DOI: 10.1177/01925121231168833.

14.

Kim

Eunji

(2023) Entertaining Beliefs in Economic Mobility. American Journal of Political Science 67(1): 39–54.

15.

Kriesi

Hanspeter

Grande

Edgar

Lachat

Romain

Dolezal

Martin

Bornschier

Simon

Frey

Timotheis

(2006) Globalization and the Transformation of the National Political Space: Six European countries compared. European Journal of Political Research 45: 921–956.

16.

Lake

James

Millimet

Daniel L

(2016) An Empirical Analysis of Trade-Related Redistribution and the Political Viability of Free Trade. Journal of International Economics 99: 156–178.

17.

Leromain

Elsa

Orefice

Gianluca

(2014) New Revealed Comparative Advantage Index: Dataset and empirical distribution. International Economics 139: 48–70.

18.

Magee

Christopher

(2001) Administered Protection for Workers: An analysis of the trade adjustment assistance program. Journal of International Economics 53(1): 105–125.

19.

Moene

Karl Ove

Wallerstein

Michael

(2001) Inequality, Social Insurance, and Redistribution. American Political Science Review 95: 859–874.

20.

Norman

Geoff

(2010) Likert Scales, Levels of Measurement and the ‘Laws’ of Statistics. Advances in Health Sciences Education 15: 625–632.

21.

Nunn

Nathan

Trefler

Daniel

(2014) Domestic Institutions as a Source of Comparative Advantage. Handbook of International Economics 4: 263–315.

22.

Organisation for Economic Co-operation and Development (2010) The OECD STAN Database for Industrial Analysis. Date of release: 26 January 2010. https://stats.oecd.org/Index.aspx?DataSetCode=STAN_INDICATORS_2009

23.

Owen

Erica

(2015) Labor and Protectionist Sentiment. In Martin

Lisa L

(ed.) The Oxford Handbook of the Political Economy of International Trade. New York, NY: Oxford University Press, 119–137.

24.

Owen

Erica

Johnston

Noel P

(2017) Occupation and the Political Economy of Trade: Job routineness, offshorability, and protectionist sentiment. International Organization 71: 665–699.

25.

Rehm

Philipp

(2009) Risks and Redistribution: An individual-level analysis. Comparative Political Studies 42: 855–881.

26.

Robitzsch

Alexander

(2020) Why Ordinal Variables Can (Almost) Always Be Treated as Continuous Variables: Clarifying assumptions of robust continuous and ordinal factor analysis estimation methods. Frontiers in Education 5: 589965.

27.

Romer

Thomas

(1975) Individual Welfare, Majority Voting, and the Properties of a Linear Income Tax. Journal of Public Economics 4: 163–185.

28.

Rueda

David

Stegmueller

Daniel

(2019) Who Wants What? Redistribution Preferences in Comparative Perspective. Cambridge: Cambridge University Press.

29.

Stadelmann-Steffen

Isabelle

Eder

Christina

(2021) Public Opinion in Policy Contexts. A comparative analysis of domestic energy policies and individual policy preferences in Europe. International Political Science Review 42: 78–94.

30.

Thewissen

Stefan

Rueda

David

(2019) Automation and the Welfare State: Technological change as a determinant of redistribution preferences. Comparative Political Studies 52: 171–208.

31.

Van Hoorn

Andre

(2017) Social Trust, Workplace Organization, and the Comparative Advantage of Nations. Oxford Economic Papers 69(4): 897–917.

32.

Van Hoorn

Andre

(2022) Automatability of Work and Preferences for Redistribution. Oxford Bulletin of Economics and Statistics 84: 130–157.

33.

Walter

Stefanie

(2017) Globalization and the Demand-Side of Politics: How globalization shapes labor market risk perceptions and policy preferences. Political Science Research and Methods 5: 55–80.

34.

Yang

Jae-Jae

Kwon

Hyeok Yong

(2021) Union Structure, Bounded Solidarity and Support for Redistribution: Implications for building a welfare state. International Political Science Review 42: 277–293.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

31.92 MB

0.00 MB

0.09 MB