Can interest rate spreads stabilize the euro area?

Abstract

Since the onset of the financial crisis significant interest rate spreads have arisen between euro area countries, both for public and private debt. We check whether these spreads could be made to work towards the goal of providing more stability to the euro area. In particular, we focus on reducing the imbalances that arose between the core and peripheral members of the euro area in the first decade of its existence. The idea is that stable positive spreads in peripheral countries could have decreased domestic demand, preventing the boom–bust cycles that plagued these economies. They could also prevent such developments in the future. We construct a panel model for euro area countries and estimate the relationship between real interest rates and the current account balance. Next, we use the estimated parameters to perform simulations. We find that spreads on real interest rates of 0.6–5.5 percentage points would have been necessary to stabilize external positions of the four peripheral euro area member countries.

Keywords:

• euro area
• imbalances
• current account
• panel estimation

JEL Classification:

• E32
• E43
• E52

Acknowledgements

We would like to thank Alistair Dieppe, Michele Ca’ Zorzi and Enchuan Shao as well as the participants of the seminars at Narodowy Bank Polski, the Third ISCEF Conference in Paris, the IFABS Conference in Lisbon, the WIEM conference in Warsaw and the ECOMOD conference for valuable comments. The views expressed herein are those of the authors and not necessarily those of Narodowy Bank Polski or the Warsaw School of Economics.

Notes

¹ As a matter of fact, the problem also affected (and sometimes even more) countries that adopted hard pegs against the euro, Estonia, Latvia and Lithuania (Brixiova et al., 2009; Kuodis and Ramanauskas, 2009).

² As a cross-check, we ran our regressions also in a static form with panel-corrected SEs. Obtained estimates of an immediate impact of a change in the real interest rate on the CA were similar to those in our preferred dynamic model. However, the high estimated coefficient of autocorrelation of residuals from the static model as well as high and robust significance of the lagged CA in the dynamic specification suggest that the latter one should be our preferred choice.

³ Time dimension of 18 observations may, indeed, seem quite large for a panel data set. However, it is rather small for cointegration time-series methods, such as panel ECM models proposed by Pesaran and Smith (1995); Pesaran et al. (1997). What is more, power of unit root test for our data set seems rather low, as their asymptotic properties depend on T and N being large. Instead, we prefer to include lagged dependent variable to control for autocorrelation and year dummies to control for common shocks and possible trends in the data.

⁴ As a cross-check, we also used system-GMM estimates that have better properties if ρ → 1, keeping in mind that due to overidentification (large number of instruments relative to the number of observations) they should be interpreted with great caution (see Roodman, 2009 for extensive discussion). These results, however, are fairly similar to those obtained by OLS and FE estimators and are available on request.