Development of the Public-Private Wage Differential in the Netherlands 1979-2009

The public premium has disappeared during the 1980s when public-private wage linkages have been abandoned. Although the balance was partly restored recently when wages in Education improved, the public premium was still -2% in 2009. On average, public sector wages are unfavourable for men, people over 35 and higher educated. Most public-sector women also earn a premium over private-sector women, but not in all cases: not in 1989, nor higher educated women in recent years.
This discussion paper documents in much detail the long-run evolution of the wage gap between the private and the public sector in the Netherlands.1 It estimates and decomposes wage equations for six different years stretching over a long period (1979, 1989, 1996, 2002, 2004 and 2009). The raw public wage advantage fell from 15 per cent in 1979 to 8 per cent in 1989, subsequently doubled to 16-17 per cent in 1996 and 2002, and then fell back again to 11-12 per cent in 2004 and 2009.
This raw pay gap reflects both a composition effect (the two sectors are made up of different types of workers) and a reward effect (the sectors pay different rewards to the same type of workers). This reward effect is the prime focus of this paper. A positive reward effect means that public-sector workers are comparatively better paid, while a negative outcome implies a wage penalty for working in the public sector. Oaxaca-Blinder decomposition is used to split between the two above effects. The reward effect appears relatively high in 1979 (4 per cent) but falls considerably in the eighties and is actually negative in 1989 (minus 7 per cent), turning a premium into a considerable penalty in ten years’ time. In the 1990s the balance is restored as the public premium sticks around zero in 1996 and 2002. But then the public premium rapidly becomes negative again, resulting in a 3 per cent penalty in 2004. A slight improvement since then brings the public penalty back to around 2 per cent in 2009. A robustness check using other methods gives the same results.