Can Demography Explain Secular Stagnation?
by Axel Gottfries and Coen Teulings , voxeu.org
The secular stagnation hypothesis has gained traction in the aftermath of the Global Crisis. This column argues that demography has played an important role in reducing the interest rates. The increase in life expectancy, which has not been offset by an increase in the retirement age, has led to an increase in the stocks of savings. The latter will go into price increases for assets in fixed supply – such as housing – rather than in adding new capital. Potential remedies for absorbing the extra savings are increasing the retirement age and an extension of the pay-as-you-go benefit systems.
The notion of secular stagnation goes back to Alvin Hansen’s 1938 presidential address, “Economic Progress and Declining Population Growth”. The idea received new interest after Larry Summers’ speech at the IMF last year. At the heart of the debate is the steady decline in the real interest rates since 1980 (see the ebook edited by Coen Teulings and Richard Baldwin 2014 for an overview).
The nominal interest rates for the Eurozone, US, and UK have been virtually zero since the onset of the Global Crisis. Figure 1 shows the average real interest over the business cycle for the Eurozone and the US, showing a steady decline over the last three business cycles.
Figure 1. Trends in real interest rates, US and Eurozone
Source: Eurostat, and FRED online database. EONIA and Fed Funds minus core inflation.
The secular stagnation hypothesis claims that this downward trend has driven the full employment real interest rate (FERIR) into the negative territory. Hence, it will be increasingly difficult for monetary policy to achieve full employment due to the zero lower bound.
Olivier Blanchard, Davide Furceri, and Andrea Pescatori summarise the potential explanations for the decline in the real interest rate in their contribution to the ebook, discussing the role of monetary policy, of a shift in the composition of net demand for assets towards safe assets, and trends in supply and demand for loanable funds. As we will argue in this column, demography has played an important role in the downward trend in the real interest rate, in particular the rise in life expectancy.
Life expectancy and the supply of savings
Over the last 40 years, there has been considerable increase in life expectancy. Table 1 presents data for the world’s four biggest economies – the US, China, Japan, and Germany. The increase in life expectancy has not been accompanied by a corresponding increase in the retirement age. In fact, there has been a small fall in the average retirement age.
Other things equal, this trend must have led to an increased supply of savings; workers have to save more to finance their consumption during a prolonged period of retirement.
Table 1. Average retirement age and life expectancy
Note: For the average retirement rate, we use the minimum of the OECD data male average effective age and official age of retirement. For the retirement rate in China 1970, the value from 1987 is used, and for Germany in 1970 and 1990, the value from 1996 is used. For the life expectancy we use total life expectancy at birth.
The increase in the life expectancy and the approximately constant retirement age are not the only relevant trends. Table 2 documents the changes in population growth and the average educational attainment. Average education attainment has gone up in all four countries, most of all in Germany. Consumption during the years at school and at university must be financed by loans from elderly generations. This partly offsets the increase in savings for retirement. Population growth has come down in all countries, most of all in China. Lower population growth implies that elderly cohorts increase in size relative to younger cohorts. Since the stock of savings is at its maximum at the moment of retirement and since the age of retirement falls well beyond the mean, a lower population growth implies that on average the relative size of cohorts with a high stock of savings increases relative to cohorts with a low stock of savings.
Hence, a lower population growth-rate will increase the desired stock of savings.
Table 2. Average years of education and population growth
Note: We use average years of total schooling for the age group 25-29 from the Barro R. & J.W. Lee Dataset.
Sources: Barro R. & J.W. Lee Dataset and World bank
We propose a simple statistic to summarise these effects in one statistic based on a stylised life cycle model. For this summary statistic, we set the real return on savings equal to zero. We assume that the educational attainment, retirement, and life expectancy are fixed and identical for all individuals. Workers start school at the age of 6. They start working after completing their education and they work until their retirement. Consumption starts at age of 10 and is perfectly smoothed over the workers’ lifetime. Using these assumptions, we can calculate the desired stock of savings for a cohort at each age. Normalising annual labour income to unity, the life time income Y is given by (retirement-educational attainment-6). Smoothing this income to achieve a constant consumption level C over the life time starting from the age of 10 yields C = Y/(Life expectancy-10). The stock of savings for a worker of age a is then equal to the sum of annual savings 1-C from the start of the career till the current date minus consumption during the years at school and at university, C times (Educational attainment + 6 – 10). Integrating the stock of savings over all the cohorts yields the total stock of savings in terms of labour income. We express this stock as a percentage of GDP, assuming a labour share of 2/3. The result of this calculation is presented in Table 3.
In all four countries, the desired stock of savings went up dramatically, on average by about two times GDP. The increase in life expectancy, which has not been offset by a corresponding increase in the retirement age, is the most important driving force behind this trend.
In 1970, the desired stock of savings was still negative in three out of four countries due to the investment in human capital, that is, the consumption during years at school. By 2010, all countries require a positive stock of savings. The necessary stock of savings is the highest in Germany, due to the low retirement age and the low rate of population growth. Since these four economies together account for 45% of world GDP, and since many other countries have experienced similar trends, this increase in desired savings from a lifecycle perspective must have had a major impact on the global capital market.
Table 3. The desired stock of savings
Sources: IMF, OECD, Barro R. & J.W. Lee Dataset and World bank.
This calculation requires some strong assumptions.
On the one hand, we have ignored the impact of individual uncertainty. Allowing for this would imply that individuals would save for precautionary motives, thereby increasing the desired stock of savings.
Blanchard et al. (2010) claim that this plays an important role in China where the rapid economic development raised the relative price of health care and where no public insurance is available.
On the other hand, our calculation is based on a zero return to saving. Allowing for a positive return would reduce the desired stock of savings.
The consumption level drops with retirement, suggesting that consumers have lower needs after retirement (see for example Battistin et al. 2009 for a recent estimate). Other things equal, this would lower desired stock of savings.
More importantly, we ignored the role of policy. Pay-as-you-go (PAYG) pension and health care systems reduce the desired stock of savings. Future generations pay for (part of) the cost of retirement, reducing the necessity to save. Hence, the increase in the desired stock of savings might have been (partially) offset by an increase in the role of pay-as-you-go. We are not aware of major changes in institutions in these four economies in this regard. However, even at constant institutions, part of the increase in the desired stock savings might have been offset by automatic increases in the pay-as-you-go claim, since a higher life expectancy increases the cost of such systems, as in e.g., the US Medicare and Social Security systems.
Finally, we take the retirement age as given. One would expect that an increase in life expectancy and a fall in the real interest rate would lead to an increase in the actual retirement age because more saving is needed and saving is less rewarding.
However, the actual retirement age might be driven more by social norms than by individual optimising behaviour. Hence, the official retirement age as embedded in social legislation and public pension system might have strong impact on actual retirement (see Mastrobuoni 2009 for evidence on this issue). In the calculations, we also treat all agents at each date as identical whereas in fact they will be different depending on the year of birth.
Summarising the previous list of arguments, we do not see any reason why our calculations would either be a grave over- or underestimation of the impact of the increase in life expectancy on savings in the four biggest economies in the world.
Impact on the market equilibrium
The increase in life expectancy is, therefore, likely to have increased the desired stock of savings by about two times GDP. What is the impact of this increase in the supply of savings on the global capital market? One would expect this increased saving to lead to a fall in the cost of capital and the real interest rate, as is documented in Figure 1. This fall in the return of capital will have spurred investment. However, if the interest elasticity of investment is insufficiently high, the spur in investment might not be enough to absorb all the additional savings. Part of the additional savings would then be absorbed by higher asset prices, in particular of assets in fixed supply with a long duration and a fixed real yield. Real estate in city centres is a typical example of this type of assets. A fall in the real interest rate pushes up their prices, similarly to Jean Tirole’s (1985) celebrated analysis of the feasibility of rational bubbles when the real interest rate becomes equal to the growth rate. Part of the increase in the desired stock of savings will go into price increases for assets in fixed supply rather than in adding new capital.
Figure 2. Capital in Britain, 1700-2010
Notes: National capital is worth about seven years of national income in Britain in 1700 (including four in agricultural land).
Sources and series: See piketty.ens.fr/capital21c.
Figure 3. Capital in France, 1700-2010
Notes: National capital is worth almost seven years of national income in Britain in 1910 (including one invested abroad).
Sources and series: See piketty.ens.fr/capital21c.
Thomas Piketty (2014) provides evidence that this is indeed the case. Figure 2 and 3 show the evolution of the value of the capital stock in the UK and France since 1700. The steep increase since 1970 coincides with the upward trend in the desired stock of savings documented in Table 3. The order of magnitude also fits the numbers presented in Table 3. The decomposition shows that the main driving force behind this increase is indeed the value of the stock of housing. The evolution for the US, Canada, and Germany exhibits a similar pattern, though less outspoken than that for the UK and France. The cost of the unification might have delayed the built-up of the German stock of savings, while corporate investment is relatively high in the US. This diagnosis is supported by recent evidence on the evolution of house prices around the world presented by Knoll et al. (2014). They show that house prices have been largely flat since 1870, apart from a sharp increase during the last couple of decades due to an increase in land prices.
We have shown in this column that the demographic shifts, in particular the increase in life expectancy, can go a long way in explaining the decline in real interest rate over the past couple of decades. This is quite likely to be the main explanation for the sharp increase in house prices over the past couple of decades. This trend endangers the possibility for monetary policy to achieve full employment and it is likely to further increase house prices. When thinking about potential remedies for these problems, one has to find other ways to absorb the excess saving. An increase in the retirement age and extension of pay-as-you-go benefit systems are the obvious policies to consider.
Battistin, E, Brugiavini, A, Rettore, E and Weber, G (2009), “The retirement consumption puzzle: evidence from a regression discontinuity approach”, American Economic Review, 99(5), 2209–2226.
Blanchard, O, G Dell’Ariccia and P Mauro (2010), “Rethinking Macroeconomic Policy”, IMF Staff Position Note SPN/10/03.
Knoll, K, M Schularick, and T Steger (20140, “No price like home: Global house prices, 1870 – 1912” CESifo Working Paper no. 5006.
Mastrobuoni, G(2009), “Labor supply effects of the recent social security benefit cuts: Empirical estimates using cohort discontinuities”, Journal of Public Economics, 93, 1224–1233.
Piketty, T (2014), Capital in the 21st century, Harvard University Press.
Summers, L H (2013), “Why Stagnation May Prove To Be The New Normal”, Financial Times, 15 December.
Teulings, C N and R E Baldwin (2014), “Secular stagnation: Facts, causes, and cures- a new Vox ebook”, VoxEU.org, 10 September.
Tirole, J (1985), "Asset Bubbles and Overlapping Generations" Econometrica, 53-6, 1499-1528.