Centre party, district magnitude, and wage inequality in service economy

The study addresses why wage inequality within the service sector varies across advanced democracies and proposes a political-institutional explanation. As economies transition to post-industrial structures, service employment has expanded and is associated with pronounced intra-sector wage disparities due to heterogeneous skills, job types, and fragmented labour organisation. The authors argue that protecting the diverse and diffuse interests of service-sector workers depends heavily on political intermediaries and institutions. They hypothesise that wage inequality within the service sector will be lower when centrist parties participate in government within high-district-magnitude electoral systems. The rationale is twofold: (1) centrist parties, less tied to organised labour or business, have incentives to balance flexibility and security (flexicurity) for diverse service-sector workers; and (2) larger district magnitudes encourage parties to form broad coalitions and represent diffuse interests. The key hypothesis: service-sector wage inequality decreases as the centrist party’s cabinet share increases in countries with large district magnitudes.

The paper situates its argument within three strands. First, economic and structural accounts suggest service-sector expansion increases inequality due to lower productivity growth and labour intensity (Baumol 1967) and weakening union power in fragmented service occupations. Evidence shows pronounced heterogeneity across service subsectors (dynamic, non-dynamic, welfare), contributing to wage dispersion. Second, partisan politics literature offers conflicting views: left parties may protect outsiders in PR systems (Pontusson 2011; Iversen & Soskice 2015), but others argue social democratic parties prioritise insiders over outsiders (Rueda 2005; Lindvall & Rueda 2012). Third, electoral-institutional research highlights district magnitude as shaping incentives to represent diffuse interests (Carey & Shugart 1995; Cox 1990; Myerson 1993; Lijphart 1990). The authors synthesise these by proposing an interaction: high district magnitude creates incentives to appeal to unorganised, heterogeneous service workers, but real policy traction requires centrist parties in cabinet positions. Illustrative cases, notably the Netherlands, show centrist-led flexicurity balancing flexibility with security and reducing inequality. The literature also notes potential non-linearities in district magnitude effects and the importance of party systems.

Design and sample: Time-series cross-national panel covering 14 OECD countries, primarily 1980–2007 (EU-KLEMS ISIC Rev. 3), with main hypothesis tests focused on 10 proportional representation (PR) systems where centrist parties are institutionally relevant. Dependent variable: Intra-service-sector wage inequality measured as the Gini coefficient across three service clusters (following Wren 2013): (1) dynamic (Finance; Business Services; Transport and Communication), (2) non-dynamic (Hotels/Restaurants; Wholesale/Retail; Other Community, Social, Personal Services), and (3) welfare (Public Administration; Education; Health and Social Work). The Gini computed across these three clusters is strongly correlated (r ≈ 0.82, SE 0.03) with a 13-industry-cluster Gini from EU-KLEMS. Key explanatory variables:

• District magnitude (DM): Mean District Magnitude from the Inter-American Development Bank, computed as total seats/constituencies; natural log used in figures; larger values reflect larger multi-member districts typically associated with PR systems.
• Government partisanship: Cabinet portfolio shares by ideological family (centrist, left, right) from Swank (2018), based on Castles and Mair (1984) expert classification. The focus is the centrist party cabinet share (CP). Controls:
• Centralisation of wage bargaining (Visser 2015 ICTWSS).
• Deindustrialisation: service workers as a share of total employment (EU-KLEMS).
• GDP per capita (World Bank).
• Age dependency ratio (young + old relative to working age). Model: Panel regression with interaction between DM and CP and panel-robust SEs. Estimators include: country fixed effects (one-way FE), between-effects (BE), two-way FE (country and year), and random effects (RE) with quasi-demeaning. Generic specification: y_it = γ·DM_it·(1 + CP_it) + δ·CP_it + X_itβ + α_i + μ_t + ε_it; primary interest is the conditional effect of CP at varying levels of DM (and vice versa). Computations were conducted in R using the plm package (Croissant & Millo 2008). Scope and robustness: Primary tests in 10 PR countries; comparative results for the full 14-country sample including 4 majoritarian systems. Robustness checks include two-year lags of partisanship variables, exclusion of the Netherlands, and alternative inequality measures (Ricci-Schutz, Atkinson, Theil, generalised entropy) showing correlations > 0.96 with the Gini and consistent results.

• Descriptive patterns: Across 14 countries (1980–2007), district magnitude is negatively correlated with service-sector wage inequality (Fig. 2). Within-country Gini series show substantial temporal variation (Fig. 3).
• PR sample (10 countries): The interaction term DM×CP is negative and statistically significant across country FE, two-way FE, and RE models, indicating that as district magnitude increases, higher centrist cabinet shares reduce service-sector wage inequality. Example coefficients (Table 1):
  • Country FE: DM×CP ≈ −0.018 (SE 0.007).
  • Two-way FE: DM×CP ≈ −0.017 (SE 0.006).
  • RE: DM×CP ≈ −0.016 (SE 0.007). Main effects are positive: DM ≈ +0.008 to +0.009 and CP ≈ +0.049 to +0.055; the net effect turns negative at higher DM levels, as visualised in Fig. 5, where the centrist share’s effect crosses zero around average PR district magnitude (~8 seats).
• Full sample (14 countries including majoritarian): Evidence for a negative interaction is weaker (significant at ~10% in FE specifications; Table 2), consistent with the scope condition that centrist parties have limited policy leverage in majoritarian systems. Between-effects shows DM negative on inequality on average.
• Controls: In PR models, greater centralisation of wage bargaining is associated with higher service-sector wage inequality (e.g., FE ~ +0.093), and higher GDP per capita correlates with higher inequality, consistent with faster service growth in affluent countries.
• Robustness checks:
  • Lagged partisanship (two-year): DM×lag(CP) remains negative and significant in FE and RE models (e.g., ≈ −0.014 to −0.013).
  • Excluding the Netherlands: The negative interaction strengthens (e.g., FE DM×CP ≈ −0.038; two-way FE ≈ −0.036; Table 4), confirming results are not driven by the Dutch case.
  • Alternative inequality indices highly correlated with the Gini (>0.96) reproduce the main findings. Overall, neither higher DM nor higher CP alone consistently reduces inequality; the reduction emerges from their interaction.

Findings support the core hypothesis that institutional and partisan factors jointly shape intra-service-sector wage inequality. High district magnitude encourages broader, more encompassing electoral coalitions, elevating the salience of diffuse, heterogeneous service-worker interests. Centrist parties, relatively untethered to organised labour or business, have incentives to craft flexicurity-style policies that trade off flexibility and security to serve these constituencies. The significant negative interaction implies that the inequality-reducing influence of centrist cabinet participation materialises when electoral districts are sufficiently large to reward outreach to unorganised service workers. In contrast, in systems with small districts or limited centrist cabinet presence, service-sector wage inequality is higher or unaffected. The results reconcile mixed findings in prior literature by showing that partisan effects depend on electoral context. Policy-wise, the evidence suggests that fostering inclusive representation (via larger district magnitudes) and enabling centrist participation in government can create conditions conducive to reducing wage disparities within increasingly dominant service economies.

The paper advances a political-institutional account of wage inequality within the service sector. It demonstrates that the interaction between electoral institutions (district magnitude) and partisan composition (centrist cabinet share) is crucial: in PR systems with large districts, greater centrist representation is associated with lower intra-service wage inequality, consistent with flexicurity-oriented policymaking. This contributes to debates on deindustrialisation and labour market dualisation by highlighting the role of centrist parties under conducive electoral rules. Future research could extend the temporal and geographic scope beyond the EU-KLEMS ISIC Rev. 3 period, examine post-2007 dynamics under ISIC Rev. 4, test for non-linear and party-system–conditional effects of district magnitude, and link macro-level findings with micro-level wage distributions and employment trajectories within service subsectors.

• Data scope: Dependent variable constructed from EU-KLEMS ISIC Rev. 3 (1980–2007), limiting temporal coverage; newer data (ISIC Rev. 4, 1995–2020) are not directly comparable.
• Sample size and external validity: 14 OECD countries, with primary inference in 10 PR systems; limited generalisability to majoritarian systems where centrist parties have marginal roles.
• Measurement: Wage inequality measured across three service clusters; although correlated with finer 13-cluster measures and alternative inequality indices, aggregation may mask within-cluster dispersion.
• Model uncertainty: Potential non-linearities in the effect of district magnitude and interactions with party system fragmentation are acknowledged but not fully modelled.
• Identification limits: Observational design with country and time effects cannot fully rule out omitted variable bias or policy endogeneity; lagged specifications mitigate but do not eliminate concerns.
• Descriptive inconsistencies: Some pooled between-effects estimates behave differently (e.g., interaction in 14-country BE), reflecting heterogeneity and scope conditions.