Collective Bargaining Union Effects

What the evidence says

Across countries, the strength of collective bargaining coverage tracks wage inequality closely: coverage alone accounts for 50 percent of the cross-national variance in the P1/P10 earnings ratio across 32 OECD states, and the relationship runs negative — more coverage, less inequality. The German case gives the same relationship a time dimension: as bargaining coverage fell from over 80 percent to under 60 percent over two decades, the decline contributed more to rising wage inequality than international trade did, and a separate decomposition study calls the contribution of falling coverage to inequality "sizeable." The mechanism shows up at the level of individual workers, too — between 1999 and 2010 the wage gap between covered and uncovered German workers rose from 8 to 19 percent (1 to 10 percent once firm size is controlled), while bargaining coverage itself fell 16.5 percentage points for men and 19.1 points for women between 2001 and 2006. Two natural experiments point to the same lever operating locally: abolishing administrative extension of agreements in Portugal reversed wage compression at the bottom of the distribution (and also reduced employment), while Denmark's shift from centralized to decentralized bargaining reduced wage compression. Union members and covered workers also carry a wage markup over uncovered counterparts — around 15 percent in the United States, 5 to 10 percent in other industrial countries, and of unquantified but variable size in developing countries.

Employment effects run in the opposite direction from wage-inequality effects, but the evidence for them is weaker and split by identification strategy. A cross-sectional comparison of 20 countries from 1983–88 found the unemployment rate rises with bargaining coverage, and centralization of wage-setting shows a hump-shaped relationship with unemployment: both highly decentralized and highly centralized bargaining associate with low unemployment, while intermediate centralization associates with higher unemployment. Employer-side coordination appears to matter more than employee-side coordination for keeping unemployment low, and high union density and coverage do not produce poor unemployment outcomes when paired with high bargaining coordination. But when a German study moved from cross-sectional correlation to panel and difference-in-differences estimation, the negative association between collective bargaining and employment growth disappeared entirely — the cross-sectional result reflected negative selection of declining firms into collective bargaining, not a causal effect.

Two US quasi-experimental studies isolate causal effects of specific bargaining rules. Public-sector duty-to-bargain laws depress male labor market outcomes the longer workers are exposed to them: at 10 years of exposure, annual earnings fall $2,134 (3.93 percent), weekly hours fall 0.42 (1.09 percent), employment likelihood falls 1 percentage point, labor force participation falls 0.8 points (a 6.74 percent relative decline), years of education fall 0.051, and occupational skill and noncognitive skills also decline. These effects are substantially larger for Black and Hispanic men — earnings fall 9.43 percent, hours 2.18 percent, employment 1.3 points — and are not found, in any consistent form, for women. Right-to-work laws, studied through county-level panel data, produce political rather than labor-market effects: Democratic presidential vote share falls 3.5 percentage points, turnout falls 2 to 3 points, the union share of campaign contributions falls 1.25 points, and Democratic control of state legislative seats falls 5 to 11 points, consistent with unions redirecting resources from politics into membership retention as dues become optional.

At the firm level, unionization associates with lower profitability across profit measures in the United States and lower R&D spending (though not slower technology adoption), alongside lower voluntary turnover, longer job tenure, more frequent layoffs, fewer hours worked, more overtime pay, more fringe benefits, and more training. Average productivity effects in the United States are close to zero, though a hypothetical union plant combining high-performance work practices with unionization shows 13.5 percent higher productivity than a non-union plant without those practices, versus only 4.5 percent for a non-union plant that adopts the same practices alone. These firm-level effects are not fixed: in Britain, as union density fell from 53 percent in 1979 to 28 percent in 1999, the measured effects of unions on wages, financial performance, and productivity diminished, and by the end of the 1990s there was no longer a negative association between union presence and financial performance.

Cross-nationally, institutional and economic openness variables predict violations of freedom-of-association and collective-bargaining (FACB) rights: trade openness decreases violations by 0.78 to 1.28 points on a 0–10 violation scale per standard-deviation increase in the trade/GDP ratio, and democracy and indices of economic freedom are both associated with fewer violations. Foreign direct investment penetration and ratification of ILO Conventions 87 and 98, by contrast, show no measured effect on actual rights violations — the source attributes the latter to weak enforcement mechanisms.

Some expected outcomes are covered only thinly or not at all. Job satisfaction shows a negative association with union membership only in specifications that omit working-conditions and industrial-relations-climate controls; once those are added, the effect disappears. Collective bargaining's effect on healthcare delivery during strikes is limited to one meta-analysis (17 studies, 10 countries) finding no measured difference in hospital mortality between strike and non-strike periods, alongside a decrease in elective surgeries and an increase in outpatient cancellations, neither quantified. Firm-level effects outside the US, UK, Germany, and a handful of other OECD countries are largely absent from these sources.

Where the evidence splits

Collective bargaining's effect on employment growth splits by identification strategy, not just by country. Cross-sectional and case-study evidence (Canada, Jamaica, Malaysia, the UK, the US) suggests unionized firms show slower employment growth, and a meta-analysis puts the magnitude at roughly 3 percent a year in Anglo-Saxon countries. But when a German study used panel and difference-in-differences estimation instead of cross-sectional comparison, the negative effect vanished — the sources state explicitly that the cross-sectional result reflected negative selection of firms into collective bargaining, not a causal effect of bargaining itself.

Collective bargaining compliance laws in Germany raised wages in the east (2.9 to 4.6 percent within five years of implementation) but had no measured effect in the west. The sources attribute this to a design difference: the West German estimation, unlike the East German one, showed evidence of anticipation effects that violated the parallel-trends assumption, and the sources further suggest the difference "could possibly be attributed to" West Germany's generally higher baseline collective bargaining coverage and the presence of sectoral agreements already covering public transport there.

Unionized firms' financial performance splits over time within the same country. British studies through the 1980s and 1990s found a negative association between union presence and financial performance, but that association disappeared by the end of the 1990s. The sources tie this explicitly to union density itself falling — from 53 percent in 1979 to 28 percent in 1999 — rather than to a change in study design, making it a genuine trend rather than a methodological artifact.

Thresholds and dose-response

The relationship between bargaining centralization and unemployment is hump-shaped: both extreme decentralization and extreme centralization of wage bargaining associate with low unemployment, while intermediate levels of centralization associate with higher unemployment (measured across the period from the mid-1970s to the mid-1980s).

Duty-to-bargain law effects on male labor market outcomes scale with exposure duration: effects grow with time since the law's passage, and the quantified magnitudes reported — a 3.93 percent earnings decline, a 1.09 percent decline in weekly hours, a 1-percentage-point decline in employment likelihood — are all measured specifically at the 10-year exposure mark.

A qualitative index of workers' legal bargaining power shows a marginal dose-response on labor share: a 0.1-unit increase in bargaining power (measured on a 1–6 scale covering provisions like agreement duration, amendment, and termination rules) is associated with a 3.7 percent increase in labor share in non-transition economies.

Open empirical questions

• Whether the wage-inequality effect of bargaining coverage holds under causal (panel/quasi-experimental) identification outside Germany, given that the one panel-based test available here (employment growth) reversed the cross-sectional result once selection was controlled.
• What explains the post-1981 disappearance of strikers' wage gains — the source reports the break but not a mechanism.
• What the magnitude is of the elective-surgery decrease and outpatient-cancellation increase during healthcare strikes; the direction is established but not the size.
• Whether firm-level union effects on profitability, productivity, and R&D found in the US, UK, and a handful of OECD countries generalize to developing-country labor markets, where the wage markup is noted as "variable" but never quantified.