Workers in factory
VoxEU Column Labour Markets

Union ‘effects’ on hourly and weekly wages: A half-century perspective

Union membership across the developed world has been falling for decades. This column uses data on wages and hours worked in the US over the last 50 years to examine whether this has led to a fall in the ‘union wage premium’. The authors find that while the hourly wage premium for union members has fallen notably since the 1970s, the differential in weekly wages has remained large, driven in part by union members working longer hours. This underexplored role of unions is important for the welfare of workers whose consumption is dependent not only on a decent hourly wage, but the offer of sufficient paid hours of work.

Across the developed world the proportion of workers who are union members has been declining for decades (Garnero et al. 2017).  Today the rate of membership in the US stands at 33% in the public sector.  In the private sector it is 6%, down from 24% 50 years ago. 

This decline is perceived by some to be indicative of a shift in bargaining power between employers and workers which has resulted in a decline in labour’s share of income (Summers and Stansbury 2020).  Since the root of a union’s ability to bid up wages above the market rate is its ability to call on its members to support its bargaining position and, if necessary, withdraw its labour through strike action, one might expect this decline in union density to have resulted in a secular decline in the union wage premium – the mark up unions achieve over the wage similar workers would get in the absence of the union.  But has it? The short answer is no.

To address this question, we examined data on employees’ wages and hours over the last 50 years in the United States’ Census Bureau’s Current Population Survey (CPS) (Blanchflower and Bryson 2024).  We follow in a long tradition in labour economics going all the way back to Lewis (1963) in the early 1960s, who compared the unadjusted gap in mean log hourly earnings of members and non-members with the adjusted gap having adjusted for potential confounders such as education.

In 1973, union members were earning around two-fifths more than non-members in hourly earnings (40%), but this had halved to 20% by 2023. This ‘unadjusted gap’ was at or above 35% for all but two years through to the end of the 20th century, but has remained below that subsequently, with a decline apparent from around 2013-2014 which gathered pace with the advent of COVID.

The adjusted hourly union wage premium is considerably smaller than the unadjusted ‘gap’ throughout the course of the 50 years, rarely exceeding half the unadjusted gap. For much of the time it has varied between 10% and 20%.  Although it dipped below 10% during COVID, it remains sizeable and significant (see the blue line in Figure 1). 

What has gone largely unnoticed in the literature on the union wage premium is the much larger gap between union members and non-members in their mean weekly earnings. By 2023 the unadjusted weekly wage differential between union members and non-members was 31%, nearly half what it was 50 years previously. However, the adjusted weekly wage differential has been more stable, varying between 20% and 26% between the early 1990s and just prior to COVID.  During COVD it declined, like the hourly adjusted premium, but was still 16% in 2023 (see the orange line in Figure 1).

Figure 1 Adjusted union wage premia, 1973-2023, US whole economy, CPS (%)

Figure 1 Adjusted union wage premia, 1973-2023, US whole economy

The larger weekly earnings gap between union members and non-members is due to the hourly wage premium coupled with the fact that unionised workers work about 5% longer each week than their non-member counterparts – equivalent to 1 to 2 hours per week.  This is a little-known empirical regularity. Indeed, early studies wrongly concluded that non-union workers worked longer hours than union workers (Perloff and Sickles 1982, Earle and Pencavel 1990). The working hours differential partly reflects unions’ ability to tackle under-employment, such that union workers work closer to the hours they desire than their non-union counterparts.

The stability in the regression-adjusted weekly and hourly earnings differentials across a half century is quite striking. It is true that the weekly union wage premium has dropped since COVID, and that the hourly union wage premium was falling even before the pandemic. Even so, both remain large and substantial in 2023.

The trends are not consistent with a world in which trade unions have lost all bargaining power.  However, these premia are not driven exclusively by unions’ collective bargaining prowess. Other factors may be at play, including a possible ‘batting average’ (Metcalf 1989) effect arising from unions’ ability to maintain their presence in workplaces with larger rents to share. What is perhaps most striking is the role unions play in raising hours. It is a role that has not emerged clearly from the earlier literature but is important for the welfare of workers whose consumption is dependent not only on a decent hourly wage, but the offer of sufficient paid hours of work.


Blanchflower, D G and  A Bryson (2024), “Unions, Wages and Hours”, NBER Working Paper 32471.

Earle J S and J Pencavel (1990), "Hours of work and trade unionism”, Journal of Labor Economics 8(1), Part 2: Essays in Honor of Albert Rees: S150-S174

Garnero, A, S Martin and S Cazes (2017), “The state of trade unions, employer organisations and collective bargaining in OECD countries”,, 10 July.

Lewis H G (1963), Unionism and Relative Wages in the United States, University of Chicago Press.

Perloff J M and R C Sickles (1987), “Union wage, hours, and earnings differentials in the construction industry”, Journal of Labor Economics 5(2): 174-210.

Summers, L H and A Stansbury (2020), “Declining worker power versus rising monopoly power: explaining recent macro trends”,, 2 June.

Metcalf D (1989), ‘Water notes dry up: the impact of the Donovan reform proposals and Thatcherism at work on labour productivity in British manufacturing industry’, British Journal of Industrial Relations 271(1): 1-31.