Could the decrease in Belgian government debt-servicing costs offset increased age-related expenditure?

* The authors would like to thank to the two anonymous referees for helpful comments on the paper.

¹ Borio et al. (2017) point out two technical issues with such a structural modeling approach: First, the underlying theoretical models rest on untested assumptions. Second, empirical models tend to be overidentified, which may pose a challenge to out-of-sample predictions. We will get back to that discussion below.

² A simple way to measure inflation expectations is to calculate the spread between yields of inflation-linked bonds and bond yields of the same maturity. A more precise theoretical concept is that of the equilibrium real interest rate; the real interest rate where real GDP equals potential GDP, i.e. where the output gap is zero and where the inflation rate is at the level of the target inflation rate. This unobserved interest rate has received much attention in recent literature and several authors have suggested models for estimating it (Justiniano and Primiceri, 2010; Barsky, Justiniano and Melosi, 2014; Cúrdia et al., 2015; Kiley, 20155; Laubach and Williams, 2015; Lubik and Matthes, 2015). See also Taylor and Wieland (2016) for a discussion on the shortcomings of these model-based approaches.

³ In Belgium, inflation, as measured by changes in the consumer price index (CPI), has remained remarkably constant over the past decades (see figure A1 in the appendix I). Currently, all Belgian government bonds yield negatively when adjusted by inflation (see figure A2 in the appendix I).

⁴ This shift in itself may be driven by aging societies, see Liu and Spiegel (2011). 

⁵ See IMF (2014a:19) for details on the empirical specification estimated. Essentially, the authors regress real private investment on a measure of lagged real GDP. They then analyse the structure of the forecast error under the hypothesis that it is negative if real investment declined more than what can be predicted by the lagged real output term.  

⁶ Borio et al. (2017) attempt to tackle this issue by analyzing long-term data over more than a century and including both monetary regime change dummies as well as variables that capture savings and investment drivers in an empirical specification. They then proceed to argue that variation in country-specific real interest rates is mostly explained by the regime-change dummies. However, this approach is econometrically difficult: Since all trends coincide, the regime change dummies are essentially country-specific era fixed effects. In such a specification, the investment and saving proxy variables are estimated based on the variation left around the mean within these country-by-era fixed effects. This could still be informative if there was heterogeneity across countries in monetary regime changes which do not coincide with trends in relevant structural parameters; however, there is very little of such heterogeneity in advanced countries (see table 9, p. 27 in Borio et al., 2017). A further issue with including savings and investment drivers as independent variables simultaneously is the multi-collinearity between these variables. Some of them, such as GDP growth, population growth and life expectancy are by definition highly correlated, which will both affect point estimates (in an a priori unknown manner) and increase the standard errors around these estimates. 

⁷ See for example Breedon, Chadha and Waters (2012), who find that the Bank of England’s QE1 lowered government bond yields by 50 to 100 basis points.

⁸ The ECB’s President, Mario Draghi, summed up the situation well in a recent speech: “[raising real interest rates can] only be achieved by structural reforms that elicit a structural rebalancing of saving and investment”.

⁹ Arguably, government gross debt to GDP is the more relevant measure for short-to medium term debt sustainability analysis. However, for our purposes, it is important to consider assets held by the government as their value will be equally affected by changes in the interest rate environment.

¹⁰ Note that net government debt is decreasing in most euro area countries due to the ECB’s sovereign bond purchasing program. In January 2017, the euro area national central banks held 14% of gross government debt in the euro area (with the exception of Greece, which is not eligible to participate in the program).  

¹¹ Data was extracted from Bloomberg.

¹² Since treasury bills have a maximum maturity of one year, we do not expect further gains accruing to them. We thus focus our estimates on all other bonds.

¹³ The Belgian government also entered into a number of forward rate agreements, typically using 3 month and 6 month Euribor as the floating rate. The resulting coupons are all very close to zero. We do not consider them in our analysis for both their negligible small amounts and reasons of simplicity.

¹⁴ This assumption is conservative for two reasons. First, we assume a linear and positively sloped yield curve. Figure A2 in the appendix shows that this assumption is reasonable – the ‘flattening out’ of the yield curve for maturities >20 years only increases potential gains. Second, in reality, public debt management is not static. Increased future growth expectations that would result in a steeper yield curve could easily be compensated for by issuing more short-term debt. 

¹⁵ Data on the Belgian net debt-to-GDP ratio were extracted from the IMF’s World Economic Outlook database.

¹⁶ To see this, add the 0.5 percentage point fall in Figure 5 from 2013 to 2015 to the projected 1.1 percentage points from 2015 to 2030.

¹⁷ The relative riskiness of the investment portfolio (i.e. the composition in terms of equity and debt instruments) across countries is also a factor. If Belgians are more likely to hold equity relative to other nationalities, lower interest rates would affect their investment income comparatively less.

¹⁸ All calculations based on 2016-Q4 data of the Bank of Belgium; GDP (2016) data as reported by Eurostat.

¹⁹ Eurostat data.

²⁰ As pointed out by Gros (2016), purchasing domestic government bonds by national central banks effectively decreases their maturity to 0. As of 31 March 2017, the National Bank of Belgium held Belgian government debt worth €49 billion (11% of outstanding debt) with a weighted average maturity of 10.1 years within the ECB’s public sector purchase program (PSPP). This effectively reduces the weighted average maturity of government debt still traded on the secondary market to 8.7 years and the weighted average maturity of total outstanding debt to 7.9 years, which still constitutes a large increase over the past years. 

²¹ The costs of leaving a monetary union are generally considered too high.