These are the global trends in interest rates over the past 150 years

Interest rates are at their lowest levels of the last 150 years in virtually all advanced economies. This column argues that this unprecedented environment reflects secular global forces that have lowered the trend in the world real interest rate by about two percentage points over the past 30 to 40 years. Whatever forces might lift real interest rates in the future must also be global, such as a sustained pickup in world economic growth, or a better alignment between the global demand and supply for safe and liquid assets.

Interest rates are at their lowest levels of the last 150 years in virtually all advanced economies, as illustrated in Figure 1. Does this phenomenon reflect only headwinds still emanating from the Global Crisis that will eventually dissipate, or is it connected to secular developments? And what are the drivers of this decline? These questions are central to monetary policy in the US and other countries, and are the focus of a growing literature (e.g. Holston et al. 2017, Gourinchas and Rey 2016).

Figure 1. Nominal yields on long term government bonds over the past 150 years

In this column, we argue that this unprecedented low interest rate environment reflects secular global forces that have lowered the trend in the world real interest rate by about two percentage points over the past 30 to 40 years. This decline has been nearly the same in all advanced economies, as the trends in their real interest rates have converged over this period. It was driven by an increase in the convenience yield for safety and liquidity and by lower global economic growth.These conclusions are based on the results of a recent paper (Del Negro et al. 2018) in which we jointly estimate trends in real rates for seven advanced economies using data on short- and long-term interest rates, inflation, and consumption starting in 1870. We do so through a flexible time-series model – a vector autoregression (VAR) with common trends. With this econometric tool, which we used in previous work (Del Negro et al. 2017), we can extract trends from a large number of variables, without having to take a stance on the dynamic interdependencies among their cyclical components. Therefore, we can use economic theory to model and interpret the trends, while remaining agnostic on whether theory holds at other frequencies. For example, the absence of arbitrage opportunities in the long run implies that we can interpret the estimated common trend in real interest rates across countries as the trend in the world real interest rate. The same theoretical framework also suggests a decomposition of this trend into some of its potential drivers, such as global consumption growth.

The first result of our empirical analysis is illustrated in Figure 1, which shows the posterior median of the estimated trend in the world real interest rate (dashed line), together with 68% (dark grey) and 95% (light grey) posterior coverage intervals. According to these estimates, the trend in the world real interest rate was stable around values a bit below 2% through the 1940s. It rose gradually after WWII, to a peak close to 2.5% around 1980, but it has been declining ever since, dipping to about 0.5% in 2016, the last available year of data. The exact level of the estimated trend is surrounded by substantial uncertainty, but the drop over the last few decades is precisely estimated. Its magnitude is also in line with that in the GDP-weighted estimate of r* for the US, Canada, the euro area, and the UK. obtained by Holston et al. (2017). A decline of this magnitude is unprecedented in our sample. It did not even occur during the Great Depression in the 1930s.

Figure 2. The trend in the world real interest rate

This finding might seem quite surprising, given the massive upheavals experienced by the world economy during the Great Depression and the World Wars. For instance, Jordà et al. (2017) also document a fall in real safe rates since the mid-1980s, but even larger declines around the World Wars and also in the 1970s. Their conclusion is that “the puzzle may well be why the safe rate was so high in the mid-1980s, rather than why it has declined so much since then”.

Figure 3. The trend in the world real interest rate and decadal moving averages

Why these differences? Figure 3 addresses this question by comparing our measure of the trend in the world real interest rate (dashed black line with grey bands) with the cross-country average of the decadal moving average of the ex-post real interest rate in each country used by Jordà and co-authors (solid red line). Clearly, the moving average approach yields a measure of the global trend in the real rate that is much more volatile than ours and that plunges deep into negative territory around both World Wars. Why is this the case?We argue that this is because decadal moving averages conflate trends with cyclical variations. This is illustrated by the dashed red line in the figure, which shows that we obtain something much closer to the decadal moving average if we allow the trend to fluctuate substantially from year to year. The implication of allowing the trend to fluctuate so much is that one cannot tell apart cycle from trend, resulting in estimates of the trend that are very uncertain. This is apparent from the large posterior uncertainty associated with the dashed red line (shaded red bands): the 50% posterior coverage intervals for the world interest rate are as large as 10 percentage points at the end of the sample!

Our own prior beliefs on what represents a trend are quite conservative: the expected change in the trend in our model has a standard deviation equal to one over a one-century horizon. However, our results are not very sensitive to modifications of the prior that cut the horizon from one century to a half-century or a quarter of a century, or even a decade, as long as one does not use extreme values such as those implicit in the dashed red line (one year). Summing up, our conclusion that the decline in the global trend in interest rates over the past two decades is unprecedented is indeed robust to a wide range of prior views of what constitutes a trend, as long as one does not allow this trend to be unduly affected by cyclical developments.

Figure 4. Trends in the real interest rates across countries

Figure 4 shows that the trend decline in the world real interest over the past 30 or 40 years has been closely synchronised across advanced economies. The figure reports the (posterior medians of the) estimated trends in the real interest rates in the seven economies in our sample (dotted lines). Over this period, country-specific trends have essentially disappeared, so that the trend in the world real interest rate is now extremely close to the trends for all countries (with the possible exception of France). This convergence in the real interest rate trends is consistent with the well-documented growing integration of global capital markets over this period. As a result, the secular decline in real interest rates over the past three to four decades is an eminently global phenomenon, with the trends in real rates falling by very similar amounts in all advanced economies. The finding that a ‘global trend’ in real interest rates has increasingly overshadowed country-specific trends presents interesting similarities with the results of recent work that emphasises the emergence of a ‘global cycle’ – or global factor – as a growing source of comovement in the returns of risky assets around the world (e.g. Gerko and Rey 2017).

Finally, Figure 5 presents the contribution to the evolution of our estimate of the world real interest rate of some of its fundamental components. In particular, we focus here on two factors that are often mentioned as potential drivers of the low interest rate environment and whose contribution can be measured using data: convenience yields and global economic growth. The convenience yield is the amount of interest investors are willing to forgo in exchange for the liquidity and safety benefits of high quality (sovereign) securities. Convenience yields are likely to be a relevant factor in our dataset because the interest rates that it contains are on either government securities or close substitutes, which are relatively safe and liquid compared to other privately issued assets. As a consequence, our world real interest rate captures the trend in real interest rates on safe and liquidsecurities. To measure this convenience yield, we include in the estimation Moody's Baa corporate bond yield for the United States, as in Krishnamurthy and Vissing-Jorgensen (2012). Under certain assumptions which we discuss in the paper, this information is sufficient to capture the long-run effect of convenience on the world interest rate.

Figure 5. The trends in the convenience yield, growth, and the world real interest rate

a) World real interest rate and convenience yields

b) World real interest rate and global growth

The top panel in Figure 5 reports the contribution of the estimated convenience yield factor to the trend movements in the world real interest rate. This decomposition shows that the trend decline in the world real interest rate over the last few decades is driven to a significant extent by an increase in the trend component of the world convenience yields (the dashed green line is plotting its negative, while the green shaded areas are the 68% and 95% posterior coverage intervals). This increase in convenience yields points to a growing imbalance between the global demand for safety and liquidity and its supply. This contribution is especially concentrated in the period since the mid-1990s, supporting the view that the Asian financial crisis of 1997 and the Russian default in 1998, with the ensuing collapse of Long-Term Capital Management (LTCM), were key turning points in the emergence of global imbalances, as argued for instance by Bernanke (2005), Bernanke et al. (2011), Caballero (2010), Caballero and Farhi (2017), and Gourinchas and Rey (2016),among others. The bottom panel shows the contribution associated with the trend in global growth. The latter is estimated as a common component of trends in the growth rate of per capita consumption across the seven economies in our sample, as suggested by the consumption asset pricing model and further detailed in our paper. A global decline in the growth rate of per-capita consumption, possibly linked to demographic shifts, is a further notable factor pushing global real rates lower. Its contribution is comparable in magnitude to that of the convenience yield since 1980, but only about half as important over the last 20 years (and less precisely estimated).

Taken together, our results suggest that low interest rates in advanced economies are a secular phenomenon driven by global forces that emerged well before the Great Recession and that are unlikely to be connected to country specific factors, such as national policies or other domestic developments. Therefore, whatever forces might lift real interest rates in the future must be global, such as a sustained pickup in world economic growth, or a better alignment between the global demand and supply for safe and liquid assets.