7. The Labor Market Outcomes of Austerity. Evidence for Europe

For Spain, consider the Plan Estatal de Vivienda y Suelo 1999–2001, 2002–2005, 2006–2008, and the Plan de Infrastructuras, Transporte y Vivienda 2012–2024. For Germany see Krebs and Scheffel (2013) and Dustmann et al. (2014).

In the case of Greece, the estimation sample remains forcibly fixed at 2000:Q1-2012:Q4 always. For Spain and Portugal, the starting date of the window is kept fixed at 1995:Q1 for the first four runs.

As argued in Sect. 7.2, these VARs are likely to be affected by many shortcomings given our set-up. For this reason, the emphasis here is not on inference, but instead on the insights that can be gained in terms of model specification. The implicit assumption is that, while inference changes depending on the methodology, the dynamic relationships between variables captured by the system of VAR equations remain stable. The additional advantage is to work with a computationally lighter version of the model.

We do not allow for feedback from selected endogenous variables to the debt level. Research based on US data did not find remarkable differences for the dynamics of macroeconomic variables other than the debt itself (Favero and Giavazzi 2007). Using annual data from 1970, Nickel and Tudyka (2013) show that the level of debt matters for the effects of fiscal stimulus policies in a group of European countries. Our sample is, however, much shorter and throughout this period no country experienced large swings in its stock of debt. Also, interest rates have been converging along a downward trend and the monetization of debt through inflation has not been an option during at least half of our sample period as the creation of money was centralized and separated from national government authorities. In this scenario, it is unlikely that the preferences of domestic policymakers were greatly concerned with debt stabilization. Some of these conditions changed from 2008 onward. So, as a robustness exercise, we identify shocks also imposing a mean reverting process on the dynamics of debt: differences are negligible.

Due to the short sample available for Greece, the large VAR runs into stability issues when estimated using OLS. To circumvent this problem, instead of adding a third round of series, we substitute the latter to the set of labor market variables used in the medium sized VAR model.

The major ones have been Plan Estatal de Vivienda y Suelo 1999–2001, 2002–2005, 2006–2008, and the Plan de Infrastructuras, Transporte y Vivienda 2012–2024.

See Krebs and Scheffel (2013) and Dustmann et al. (2014) for a discussion.

On this, see Campolmi et al. (2011). An alternative explanation has to do with expectations of future expansion in activity following adjustments in public spending, see Giavazzi and Pagano (1990).

The employment to population ratio (EP) can be written as one minus the unemployment rate (UR) multiplied by the labor force participation rate (LFPR), that is \(EP = (1-UR)*LFPR\). When we construct a decomposition of the labor market changes by taking differences of logs, we obtain \(\varDelta log(EP) = - \varDelta UR + \varDelta log(LFPR)\), making use of the approximation that the log of \((1-UR)\) is equal to the negative of UR.

This finding is consistent with most studies that employ VAR techniques and consumption - instead of product - wages definition.

From the numbers in the tables, one cannot distinguish whether the employment rate IRFs are positive or negative, but can only see if the sectoral IRFs are comparatively bigger or smaller than the aggregate IRFs. Positive numbers mean that the sectoral IRFs react more, either more positively or more negatively. Negative numbers mean that the response of the sectoral IRFs is more muted.

Spending reversals can apply not only to selected policy initiatives, but more generally to positive government spending innovations. Fitting a VAR model to US data, Corsetti et al. (2012) show that the trajectory of government spending after a shock falls below zero from the twelfth quarter onward and interpret this pattern as evidence of debt-stabilizing spending reversals. We do not observe any undershooting in our VAR, which may suggest that European countries did not systematically react to their debt levels, or that they did not do so sufficiently strongly over the estimation period.