Does Austerity Go Along with Internal Devaluations?

Abstract

We empirically show that the austerity packages implemented in some euro area countries during 2010–2014 were only partially successful in generating internal devaluations. Countries that cut spending indeed experienced a decline in nominal wages, a real exchange rate depreciation, a fall in the relative price of non-tradables and a shift of consumption toward non-tradables, whereas we find no such evidence for countries raising consumption taxes. We show that this asymmetric response is in line with a small open economy model with GHH preferences. Moreover, the output costs of correcting current accounts were higher than anticipated because neither policy was successful in raising exports through lower prices. Instead, current account improvements were solely driven by lower imports stemming from faltering domestic demand. We provide evidence that exporters absorbed lower wages through higher markups, and show in a model with pricing to market that, had firms kept their markups constant, output costs of correcting current account imbalances would have been cut by almost one half.

Appendices

Appendix 1: Empirical Section

1.1 Forecasting Specification for Government Purchases

This section briefly presents the forecasting specification for government purchases. It is the same as outlined in House et al. (2019). The forecast for government spending in country i at time t is

$$\begin{aligned} \ln {\widehat{G}}_{i,t}= {\left\{ \begin{array}{ll} \ln G_{i,t-1} + \Delta {\hat{Y}} + {\hat{\gamma }} \left( \ln {\hat{Y}}_{t-1} - \ln Y_{i,t-1} \right) + {\hat{\theta }} \left( \Delta \ln Y_{i,t} - \Delta \ln {\hat{Y}}_{i,t} \right) \qquad \forall t \le \text {2010} \\ \ln {\widehat{G}}_{i,t-1} + \Delta {\hat{Y}} + {\hat{\gamma }} \left( \ln {\hat{Y}}_{t-1} - \ln Y_{i,t-1} \right) + {\hat{\theta }} \left( \Delta \ln Y_{i,t} - \Delta \ln {\hat{Y}}_{i,t} \right) \qquad \forall t > \text {2010}. \end{array}\right. }, \end{aligned}$$

where $\ln G_{i,t}$ is the log of real government spending per capita in country i at time t (deflated by the GDP deflator), $\ln Y_{i,t}$ is the log of real GDP per capita in country i, $\Delta \ln Y_{i,t} = \ln Y_{i,t} - \ln Y_{i,t-1}$ is its log-difference and $\ln Y_t$ is the log of real GDP per capita averaged across all countries. The “hat” indicates a predicted value of the variable. This forecast specification accounts for both average sample-wide growth in GDP (through the parameter $\Delta Y$) and growth convergence dynamics (through the parameter $\gamma$), as well as a cyclical relationship (through the parameter $\theta$). We take the predicted values for $\ln {\hat{Y}}_{i,t}$ and the estimated parameter values from House et al. (2019) who estimate the average growth rate, $\Delta Y$, to be 1.8%, $\gamma$ to be 2.4% and $\theta = 0.38$. The forecasting specification for GDP is

$$\begin{aligned} \ln {\widehat{Y}}_{i,t}= {\left\{ \begin{array}{ll} \ln Y_{i,t-1} +\Delta {\hat{Y}} +{\hat{\gamma }} \left( \ln {\widehat{Y}}_{t-1}-\ln Y_{i,t-1}\right) \qquad \forall \quad t \le \text {2010} \\ \ln {\widehat{Y}}_{i,t-1} + \Delta {\hat{Y}} +{\hat{\gamma }} \left( \ln {\widehat{Y}}_{t-1}-\ln {\widehat{Y}}_{i,t-1}\right) \qquad \forall \quad t > \text {2010}. \end{array}\right. } \end{aligned}$$

(31)

Here, $Y_{i,t}$ is country i’s GDP at time t and ${\widehat{Y}}_{i,t}$ is its forecast. The specification takes last period’s value of (the log of) $Y_{i,t}$ and adds a country- and time-specific growth rate, which is composed of two parts: a common term capturing the average rate of growth of the core European countries, $\Delta {\hat{Y}}$, and a catch-up term that raises this growth rate for poorer countries and lowers it for richer countries, $\gamma \left( \ln {\widehat{Y}}_{t-1}-\ln Y_{i,t-1}\right)$.

1.2 Actual and Forecasted Time Series

Figure 6 displays actual and forecasted time series of the main variables of interest for Greece. “Online Appendix” contains similar figures for the remaining countries in our sample.

Fig. 6

Greece. Notes: Actual data and forecasts. Values are indexed to 0 in 2009. A value of 0.1 means that the variable is 10% above its level in 2009. For tax rates, a value of 0.1 means that the tax rate is 10 percentage points higher than in 2009. For government purchases, a value of 0.1 means that government purchases were 10 log points above their value in 2009. For price indices and tax rates, values refer to the end of the year. Net exports are the difference between exports and imports, expressed in percent of 2003–2009 average GDP. Net exports are also normalized to 0 in 2009

1.3 Other Austerity Measures

Here, we discuss labor taxes and capital taxes. To measure changes in these tax rates, we rely on effective tax rates. This approach builds on early work by Mendoza et al. (1994), which was further refined in Eurostat and European Commission (2014). The principal idea is to classify tax revenue by economic function using data from the National Tax Lists and then approximate the base with data from the national sector accounts. Compared to statutory tax rates, the advantages of these rates are that they take into account the net effect of existing rules regarding exemptions and deductions and also incorporate social security contributions in labor taxes. In calculating forecast errors for labor and capital taxes, we proceed as with consumption taxes. We pre-multiply the forecast errors in the tax rates by a country’s (pre-crisis) tax base in percent of GDP and then divide it by the gross tax rate. Table 5 lists the average forecast errors of government purchases, consumption taxes, labor taxes and capital taxes for the period 2010–2014. On average, the cuts in government spending were twice as large as the overall increases in tax revenues. Among the three tax measures, consumption tax rates were raised the most and also display the strongest variation in the cross section.

Table 5 Austerity measures in percent of GDP

Appendix 2: Model

“Online Appendix” provides derivations of the non-stochastic steady state as well as the log-linearized equilibrium conditions of the model used to solve the model. In addition, “Online Appendix” provides more details on the non-CES demand function.

1.1 Real Exchange Rate and Nominal Wages

Here, we show that the real exchange rate relates to the nominal wage according to:

$$\begin{aligned} -q^{\mathrm{ct}}_t = ( 1 - M) w_t \end{aligned}$$

(32)

and that this relationship is not affected by pricing to market. The real exchange rate is calculated as the inverse of the retail price at constant tax rates. Non-traded consumption goods are priced at W, whereas the price for the traded consumption good is a weighted average of the price of the wholesale good and the price of distribution services (which cost W). The price of the wholesale good is a weighted average of the domestic traded good and the imported traded good:

$$\begin{aligned} -C q^{\mathrm{ct}}_t&= \left( C^N + D \right) w_t + H p_{H,t} + M p_{M,t}. \end{aligned}$$

Note that $C^N + D + H + M = C$.

No pricing to market Without pricing to market ($\varGamma =0$), the price of the domestic traded good is $p_{H,t} = W_t$ and the import price remains unchanged, so that (32) follows.

Pricing to market With pricing to market ($\varGamma >0$), the prices of the domestic and imported traded goods are given by:

$$\begin{aligned} p_{H,t} = \frac{1}{1+\varGamma } \left( w_t + \varGamma p_{V,t} \right) \qquad \text {and} \qquad p_{M,t} = \frac{\varGamma }{1+\varGamma } p_{V,t}. \end{aligned}$$

The price of the wholesale good is

$$\begin{aligned}&V p_{V,t} = H p_{H,t} + M p_{M,t} \\&V p_{V,t} (1+\varGamma ) = H w_t + H \varGamma p_{V,t} + M \varGamma p_{V,t} \\&p_{V,t} = \left( 1 - \frac{M}{V} \right) w_t. \end{aligned}$$

Then, prices of the domestic and imported traded goods are

$$\begin{aligned} p_{H,t} = \frac{1 + \left( 1 - \frac{M}{V} \right) \varGamma }{1+\varGamma } w_t \qquad \text {and} \qquad p_{M,t} = \frac{\left( 1 - \frac{M}{V} \right) \varGamma }{1+\varGamma } w_t. \end{aligned}$$

The price of the imported good reacts little to wage movements, especially if the share of imports, $\frac{M}{V}$, is large. Intuitively, if a country produces little locally, exporters are mostly competing among themselves and prices stay constant. Notice that

$$\begin{aligned} H p_{H,t} + M p_{M,t}&= \frac{H + V\left( 1 - \frac{M}{V} \right) \varGamma }{1+\varGamma } w = \frac{H+H\varGamma }{1+\varGamma } w_t = H w_t. \end{aligned}$$

and the relationship of the real exchange rate and wages remains unaffected and given by (32).