Who Pays for Unions?

The paper asks how firms respond along every available margin — employment, prices, productivity, and profits — when a union-driven increase in labor costs hits them, and identifies this using a national policy change in Norway that quadrupled the maximum tax-deductible amount for union dues between 2002 and 2010. Because the deduction cap applied uniformly nationally, workers whose union dues were already above the old cap (and so could deduct more once it was raised) experienced a larger effective subsidy to union membership than workers whose dues were already below the cap and unaffected by the change. This creates firm-level variation in the “dose” of the subsidy shock depending on a firm’s pre-reform union dues level, which the authors exploit in an instrumented dose-response difference-in-differences design, instrumenting firm-level union density with the firm’s baseline-imputed net-of-subsidy union dues interacted with the policy timing, with firm and industry-group-by-year fixed effects throughout.

The identification relies on parallel pre-trends across differentially exposed firms and the absence of any channel other than union density through which the dues subsidy could affect firm outcomes, both of which the paper supports empirically; the first stage shows that a 1,000 NOK larger reduction in a firm’s net-of-subsidy union dues produces a 16 percentage point larger increase in firm-level union density in manufacturing (8.6 percentage points for the full private sector). The authors show no differential pre-treatment trends in union density, earnings, value added, or profits between high- and low-exposure firms over 1998–2002 (before the subsidy expansion began), and find no evidence that a firm’s exposure to other local firms’ instrument values contaminates its own estimated effects, arguing against local spillovers. The first-stage magnitude also closely matches independent survey evidence: over 5,000 Norwegian workers’ self-reported responsiveness to a hypothetical change in union dues implies a similar 7–10 percentage point change in the likelihood of union membership per 1,000 NOK.

A 1 percentage point increase in firm-level union density raises average worker earnings by about 1 log point and non-wage personnel costs by about 0.9 log points, and — rather than shedding workers in response to this higher unit labor cost — the average manufacturing firm expands employment by about 1 log point, while scaling up capital (0.9 log points) and materials (0.4 log points) roughly proportionately. This expansion in response to a cost increase is the signature of monopsony power: a monopsonistic employer that is forced to pay a higher wage moves up its labor supply curve and increases hiring, unlike a firm in a competitive labor market, which would shed workers when its wage is pushed above the marginal revenue product of labor. The proportionate scaling of other inputs alongside labor implies limited capital-labor substitution over the sample’s time horizon, and the expansion is partly debt-financed, as firms’ long-term debt rises alongside their assets.

The increase in nominal value added per worker (1.1 log points) and TFP (0.5 log points) is, for the average manufacturing firm, driven almost entirely by higher prices rather than by more units produced per worker — product-level export prices rise by 1.6–1.9 log points, a magnitude close enough to the value-added-per-worker increase in the exporter subsample (1.7 log points) to account for essentially all of it, and formal mediation analysis confirms the value-added effect is fully explained by price/markup changes rather than physical quantity. The paper uses granular Norwegian customs data on export prices (available for roughly 50% of manufacturing firm-years) to disentangle price from quantity effects that would otherwise be conflated in a standard value-added-per-worker measure. Several supplementary checks support that domestic price increases are at least as large as export price increases: prices rise more for exports to Scandinavia (treated as an extension of the domestic market) than to the rest of the world, the price effect is concentrated among firms with low export shares, and markups for non-exporters (3.5 log points) rise by far more than for exporters (0.6 log points, imprecisely estimated). This matters directly for interpreting prior literature: Barth et al. (2020), studying the same policy variation, attribute a similar value-added-per-worker increase to union-driven productivity gains, but this paper’s price-pass-through evidence suggests most of that effect is actually a price increase, not a quantity/productivity gain.

Firm profits do not decline — and rise marginally (2.25 log points per percentage point of union density, roughly a 0.05 percentage point increase in profit margin on a 5% baseline) — because the reallocation of labor, capital, and industry revenue share from smaller, less-unionized firms toward larger, more-unionized firms increases the product-market power (and pricing ability) of the expanding firms at the same time they are growing output, offsetting the direct cost of higher wages. A baseline model combining only labor monopsony and product monopoly power cannot explain simultaneously rising prices and output (since monopoly pricing implies prices should fall as output rises along a fixed demand curve) or the absence of a profit decline; the paper resolves this by allowing union density to directly shift product demand, either by reducing the firm’s price elasticity of demand as its market share grows (calibrated evidence suggests the estimated increase in market share is large enough to explain the estimated markup increase under standard elasticities), or by improving product quality via a union productivity effect (e.g., through lower turnover, improved worker voice, or efficiency-wage-driven reductions in absenteeism).

Firm size differentiates the mechanism sharply: the smallest quartile of manufacturing firms contract employment by 2.4 log points in response to a 1 percentage point union density increase (with no significant profit decline, driven by productivity gains from worker-quality improvements, labor intensification, and reduced absenteeism), while the largest quartile expands employment by 2.2 log points, sales by 2.2 log points, and profits by 3 log points, gaining 2.6 percentage points of within-industry market share. Coefficients scale roughly monotonically with firm size across the distribution. For the largest firms, the mechanism is market-power reallocation: they gain market share and pricing power as they expand, consistent with substantial pre-existing monopsony power (a 2.5 percentage point decrease in the labor markdown) that allows the firm to keep expanding without exhausting profitable hires. For the smallest firms, which have little monopsony power, the resilience of profits despite contracting output must instead come from productivity: value-added per worker rises by 2.4 log points (more than for large firms) and worker-quality proxies (average earnings fixed effects of both retained and newly hired workers, work hours, sick-leave reductions) all improve more at small firms than large ones.

Outside manufacturing — where firms generally hold much less product- and labor-market power (reflected in lower labor- and product-market concentration, HHI) — a 1 percentage point increase in union density raises compensation by 0.8 log points but reduces employment by 1.6 log points, cuts capital and materials roughly proportionately, reduces total sales by 1.4 log points, generates no increase in markups (little to no price pass-through), and weakly reduces profits. This is the textbook competitive-labor-market response to a union wage premium: when the firm lacks monopsony power, a wage increase pushed above the marginal revenue product of labor forces the firm to shed employment rather than expand, and the absence of meaningful product-market power prevents any price-based offset to the cost increase. The paper notes that markup and markdown estimates outside manufacturing should be interpreted cautiously, since the production-function approach used to separate product-market and labor-market power requires a competitively supplied input (raw materials), an assumption that holds well for manufacturing but less reliably for retail or services.

A single stylized partial-equilibrium model — combining firm-level labor monopsony (an upward-sloping labor supply curve), product-market monopoly (a downward-sloping demand curve), and Nash-bargained wages between the firm and union — explains both patterns depending on a single parameter: how large the firm’s ex ante monopsony power is relative to the union’s desired wage increment. When monopsony power is large enough that the union’s optimal wage stays below the labor-demand-constrained threshold, the bargained wage increase moves the firm up its labor supply curve, raising both wages and employment (the manufacturing pattern) — but the baseline model predicts this comes with falling prices and profits, which the data contradict. Adding either a market-share-driven reduction in the firm’s price elasticity of demand, or a union-driven product-quality/productivity improvement, allows the demand curve to shift outward enough that prices and profits rise alongside output, matching the empirical manufacturing results. When monopsony power is small (or absent, as with the smallest firms and the broader private sector), the same bargaining framework instead predicts falling employment and rising prices as the firm is bound by its (downward-sloping) demand curve rather than its labor supply curve — matching those settings’ empirical pattern once a productivity offset is added to explain resilient profits.

Back-of-envelope calculations suggest that the subsidy-driven roughly 12 percentage point rise in manufacturing union density by 2014 explains approximately 12.8% of Norway’s total 2001–2014 increase in the aggregate price level (a 5.9% counterfactual price-level reduction absent the subsidy expansion) and approximately 15.2% of the period’s total manufacturing wage growth (a 12% counterfactual wage increase), implying the net effect of unionization in this setting is a transfer from a diffuse set of consumers to a smaller population of union-affected workers and firms, rather than a transfer from shareholders to workers. Consistent with this, the labor share of value added does not rise with union density in manufacturing — firms maintain their profit share by passing costs through to prices rather than absorbing them — which the authors note stands in contrast to findings elsewhere in the literature that unionization raises the labor share of income. The reallocation toward larger, more unionized firms also has ambiguous welfare implications: it may accelerate reallocation toward more productive firms, but simultaneously concentrates market power, with unclear net effects on long-run competitive dynamics and on whether unions’ apparent efficiency-enhancing role (offsetting monopsony-driven underemployment) persists once firms gain enough market share to exercise durable pricing power of their own.