Abstract
The wage led aggregate demand hypothesis is examined for the United Kingdom over the period 1971–2007. Existing studies disagree on the aggregate demand regime for the UK, and this appears to be due to differing empirical approaches. Studies relying on equation-by-equation estimation procedures tend to find support for wage led aggregate demand in the UK, while the single study using a multiple time series estimation procedure finds no support for the hypothesis. We test the wage led aggregate demand hypothesis in the UK using VAR models estimated on quarterly data employing an alternative identification strategy based on shocks to real earnings. The results provide support for the wage led aggregate demand hypothesis during the period of study. However, the expansionary effects of higher earnings seem to be limited and relatively short-lived.
Acknowledgements
We would like to thank Matthieu Charpe, Vince Daly, Makram El-Shagi, Hans-Martin Krolzig, Miguel Leon-Ledesma, Özlem Onaran, Mathan Satchi, Ron Smith, Chris Stewart, Engelbert Stockhammer, and seminar participants at the 2015 IMK conference in Berlin, Kingston PERG, Leeds University Business School, and the Post Keynesian Study Group for invaluable suggestions for improvement. We would also like to thank two anonymous referees and the editor for their constructive comments. Any remaining errors are the responsibility of the authors.
Notes
No potential conflict of interest was reported by the authors.
1 Positive wage shocks in New Keynesian models usually reduce output on impact. Interestingly, however, a weak wage led aggregate demand mechanism is incorporated into the Bank of England’s COMPASS model, where a wage mark-up shock ‘temporarily increases labour income and consumption of [liquidity] constrained households, which is sufficient to increase total consumption in the near term. This effect means that GDP does not fall immediately’ (Burgess et al. Citation2013a,Citation2013b, B17). Also see Charpe and Kühn (Citation2015).
2 There appears to be no non-seasonal unit root in the log-level of the GDP deflated earnings since the null hypothesis that is rejected in this case. The slight ambiguities in the unit root tests should not create too much doubt in the inference, however, as all series are expected to be non-stationary.
3 Our results are robust to replacing the weekly earnings series with an estimated hourly earnings series constructed using the ONS hours data, although this reduces our ability to infer labour share movements from the results.
4 We added dummy variables in previous specifications to deal with this problem, which was more successful in the models with NSA data than with SA data. The results were not materially affected, however, so the final specifications do not include dummy variables.
5 All confidence bands shown in Figures – were computed using the bootstrap procedure in EViews with 1000 repetitions.
6 Note the labels in Figures – are the EViews workfile labels; we have provided a Table in Appendix 3 to facilitate comparison of the graphs and use of the workfile, which is available from the authors on request.