Is the Swedish central government a wage leader?

Abstract

Is the Swedish central government a wage leader? This question is studied empirically in a vector error-correction model using a unique, high quality data set. It is first shown that salaries of white-collar workers in the private sector and central government are cointegrated. It is then found that private sector salaries are weakly exogenous to the system of equations. This means that the private sector is the wage leader in the long-run model. It is also found that changes in central government salaries do not Granger cause changes in private sector salaries. Together, these findings clearly demonstrate that the central government is not placing undue pressure on salaries in the private sector. The central government is not acting as a wage leader.

Acknowledgements

We would like to thank seminar participants at Stockholm University, the Swedish Institute for Social Research and Uppsala University for their comments and suggestions. Matthew Lindquist is grateful to the Jan Wallander and Tom Hedelius Foundation for financial support.

Notes

¹ If exchange rates are flexible, then upward pressure on wages in the competitive sector may result in currency depreciations. These automatic depreciations will increase exchange rate volatility. One could argue that there may be costs to doing business with a volatile exchange rate. Furthermore, total consumer welfare may go down by more than total producer welfare goes up when the exchange rate falls.

² See also Jacobsson and Lindbeck (1971) for related work on inter-sectoral wage linkages.

³ Wage leadership can also be derived from wage bargaining models, efficienc0y wage models (see e.g. Bemmels and Zaidi, 1990), models of impeded migration (Latreille and Manning, 2000) and market models with productivity leaders (see Section II). The phrase wage leadership has come to dominate the Scandinavian debate. Its intended use is similar to the term wage spillovers often found in the international literature (see e.g. McGuire and Rapping, 1968; Addison and Burton, 1979; Breitung and Meyer, 1994; Budd, 1997; Latreille and Manning, 2000). The definition of wage leadership is also related to Budd's (1997) concept of pattern bargaining and McGuire and Rapping's (1968) notion of a key bargain.

⁴ It is not the goal of this study to analyse the validity of the normative conclusions drawn from the EFO-model. These are taken as praxis. The goal is to put this praxis to a rigorous statistical test.

⁵ Central government employment peaked at 11.5% in 1979. Since then, it has been dropping steadily due to the privatization of a number of state owned companies, to the separation between church and state, and to the transfer of grade school and high school teachers from the central government to the local government. Source: Statistics Sweden (Statistiska Centralbyrån).

⁶ See Lindquist and Vilhelmsson (2004) for a more thorough description of the relevant wage setting institutions as well as a short history of their development.

⁷ Mizala and Romaguera (1995) test for public sector wage leadership in Chile. They find that after the deregulation of the Chilean labour market (between 1979 and 1982), the public sector lost its wage leading position.

⁸ Andersson and Isaksson (1997) also make this distinction between white-collar and blue-collar workers, but these data set only goes up to 1995. The present updated data set allows one to consider the impact of the new Framework Appropriations System implemented in 1994 as well as the full impact of the move towards individual wage setting stipulated by the Framework Agreement (Ramavtal) which was put into place in 1990.

⁹ The first difference of x is denoted by Δ x.

¹⁰ This is the most commonly used definition of wage leadership found in the literature, see, e.g., Bemmels and Zaidi (1990), Holmlund and Ohlsson (1992), Jacobson and Ohlsson (1994), Mizala and Romaguera (1995), Andersson and Isaksson (1997), Tägtström (2000), Friberg (2003).

¹¹ See Latreille and Manning (2000) for more on inter-occupational wage spillovers.

¹² For more detailed information concerning these tests and their results see Lindquist and Vilhelmsson (2004).

¹³ Since only lagged values of the endogenous variables appear on the right-hand side of each equation, there is no issue of simultaneity. Furthermore, the VAR model is ‘balanced’. That is, the same regressors appear in each equation. Thus, OLS is an appropriate estimation technique. It can be applied to each equation in the system separately.

¹⁴ This also means that one can estimate each equation separately using OLS.