Do demographic changes matter? A cross-country perspective

J. of Multi. Fin. Manag. 30 (2015) 36–61

Contents lists available at ScienceDirect

Journal of Multinational Financial Management journal homepage: www.elsevier.com/locate/econbase

Do demographic changes matter? A cross-country perspective ˇ Aleksandar Sevi c´ ∗, Derek Brawn Trinity College Dublin, School of Business, Dublin 2, Ireland

a r t i c l e

i n f o

Article history: Received 7 June 2014 Accepted 18 December 2014 Available online 29 December 2014 JEL classification: G12 G15 J11 Keywords: Population Population ratio Bond yield Quantitative easing

a b s t r a c t We explore the inflation-adjusted yields on benchmark 10-year government bonds in seven developed countries to see if both demographic and non-demographic variables can explain why real bond yields are so low. We find that over the last 60 years there has been a regime shift in the late-1980s that radically altered output growth in major economies as dependency ratios declined. Because of this change we used two time periods for our models: low-frequency annual data from 1950 to 2012 and quarterly data from 1990 to 2013. We show that population age ratios are important as a determinant in the demand for financial assets. In this paper we also introduce a global proxy for quantitative easing (QE). We find that in developed markets in Europe and North America the QE has lowered bond yields by approximately 50–100 bps since the global credit crisis began. © 2014 Elsevier B.V. All rights reserved.

1. Introduction The recent quest for finding a solution to an ageing population primarily in the developed countries has been initiated by the evaluation of asset values for the retiring baby boom generation in the US. Following the end of World War II the prosperous economic environment in the US, which doubled the industrial output over the previous several years and then was further supported by rebuilding

∗ Corresponding author. Tel.: +353 1896 2699. ˇ c), ´ [email protected] (D. Brawn). E-mail addresses: [email protected] (A. Sevi http://dx.doi.org/10.1016/j.mulfin.2014.12.001 1042-444X/© 2014 Elsevier B.V. All rights reserved.

A. Sˇ evi´c, D. Brawn / J. of Multi. Fin. Manag. 30 (2015) 36–61

37

endeavours, primarily in war torn Western Europe led to the assertion of the US as a global power. An increase in birth rates reflected the benefits of these growth rates that could be only realised through unexpected interruptions in human development. Another impetus for excessive growth was the drive for globalisation and deregulation in the 1980s, followed by a demise of the communist world at the beginning of the 1990s. Even if sweeping political changes did not manage to reach all countries with similar force, the changes in the economic environment were rather unambiguous. To rebuild a country on new entrepreneurial foundations or to risk being exposed to another period of isolation, the consequences of which were quite evident throughout 20th century, were options faced by some Asian countries. Previous studies until Davis and Li (2003) mainly referred to demographic variables and linked them to effects on returns in stock markets and bonds. Based on findings and suggestions enclosed in this and some other seminal papers such as Poterba (2001, 2004) there were multiple attempts to compare various countries and variables in order to examine if there is a relationship with population variables. However, when countries were included in studies they were either analysed to confirm or refute the claims about a single country in a study or the countries were observed together in an attempt to find lasting policy recommendations. In this paper we have tried to include countries that may face the problem of ageing population, but at the same time offer intriguing specificities in dealing with the problem. US, Canada, Australia and New Zealand are historically attractive immigration countries, but there are striking differences among them. Immigration policies have a different impact on the number of immigrants (calculated as the ratio to the total population) in respective countries. While US and Canada, as part of NAFTA and NATO pact maintain strong links to Europe and Central and South America, Australia and New Zealand in spite of their dominant European heritage are increasingly more oriented towards the Pacific Rim and Asia, in particular. Finally, New Zealand faces a potential population loss to Australia, even though there has been a net gain in population (Bedford et al., 2010). Personal disposable income in New Zealand represents only 52% of that in Australia and it seems to be a strong reason for emigration to the Commonwealth of Australia (EIU ViewsWire, 2010). More importantly, Australia and Canada have made a move to funded pension schemes and the global ramifications of this change will become more evident quite soon. Studies refer to prices of publicly traded stocks and fixed income securities and neglect the fact that many of these superfunds are larger than company pension funds and their diversification policies are not necessarily based on CAPM postulates. Multiple divisions are requested to increase exposures to specific industries, but the funds themselves due to their gigantic structures are diversified. Another obvious consequence is that these aggressively growing funds avoid publicly traded assets in order to shun accusations of manipulating the market and most probable liquidity and marketability issues. They invest in alternative assets which market values are difficult to quantify. When these pension funds decide to sell assets in financial markets the impact will be significant and unless they face similar giant buyers from other countries there may be challenges that cannot be accurately predicted at this stage. Siegel’s (1998) concern about the problem of transferring trillions of dollars of baby boomers assets could be applied to a multitude of countries that will face similar challenges. Chamon and Prasad (2008) examined the age profile of savings in China in 2005 by age of head of household and the results were striking in that the highest rate at 30% belonged to the age-25 group, almost double the age-45 cohort. Clearly demographics are important as well as productivity (income) growth in determining savings rates. This early-20s cohort is likely a reflection of the once-child policy adopted in 1979. Global real interest rates fell from a peak of 5% in the mid-1980s to a natural rate of about 2% prior to the onset of the global credit crisis, and have been approximately 0% since 2012 (Blanchard et al., 2014). The authors also suggest that the world is poised for a prolonged period of low real interest rates as shifts in savings, investment, plus shifts in portfolio preference for risk-free assets point to a lower natural rate than the pre-2008 level, exacerbated by accommodative monetary policy to close output gaps and return economies to potential growth. During recent business cycles, monetary policy in the main OECD areas tended to be too loose during booms and then eased too aggressively during busts (Borio, 2012, 2014). Among countries selected in our sample, Germany and UK are important countries in the functioning of the Euro Zone and financial markets in Europe, respectively. UK is an immigration country not only for non-EU nations, but also for recently acceded EU member countries from Central and

38

A. Sˇ evi´c, D. Brawn / J. of Multi. Fin. Manag. 30 (2015) 36–61

Eastern Europe (Matheson, 2010, p. 25; Ellis, 2009). Projection for UK and other countries do not include factors that may slow down or accelerate changes in population growth rates such as geophysical impacts, climatic changes, biological outbreaks of pandemics or social collapses, such as wars (Jones, 2003) and examining future trends by neglecting to critically evaluate historical ones leads to over- or underestimates. By contrast, Thayer (2009) includes great power war, small power war, civil war and terrorism as important social factors that could change global and European population trends. In this study a particular aspect has been paid to the radicalisation of young populations and imbalances in male/female ratios in countries with an increasing political and economic influence across the globe. Population ratios for countries included in our study are depicted in Figs. 1–5 . Japan demonstrates an earlier incidence of declining youth ratios from the beginning of the 1970s and a rather rapid increase in the ratio of the oldest population cohort (aged 65+) since mid-1980s. For all other countries in the sample, major shifts occurred mainly in the mid-1980s and it is difficult to observe any stark differences among country-specific patterns. When population ratios and selected control variables are regressed in our paper on annual real long-term yields bimodal country patterns have been discovered. It seems that the recent shift to defined contributions schemes does not account for different relationships between population cohorts and fixed-income securities yields. The relationship seems to be determined by culturally specific behavioural patterns in market-based economies, such as Germany and Japan as opposed to market-based ones. When quarterly data have been analysed we do not find any relationship between population and total bond returns.1 The quantitative easing (QE) variable as a proxy for global liquidity is relevant for all countries except for the Commonwealth of Australia and New Zealand. In the subsequent Section 2 we refer to articles that have already examined population trends and their relationship with the value of financial assets. Variables and models used will be explained in Section 3, followed by an analysis and policy recommendations in Section 4. Finally, well present concluding remarks in Section 5. 2. Literature review The founding idea supporting the alleged relationship between various generational cohorts and financial markets originated in the life cycle hypothesis purported by Modigliani and Brumberg (1954). The younger generation uses salary income to acquire real estates, while older generations increase savings and invest in financial assets, which could affect real estate and financial asset prices to varying degrees and in different timeframes. The postulates of this theory coincide with the emergence of the Baby Boom generation that includes 79.6 million individuals born in the US between 1946 and 1965. An increase in births followed recession years and World War II and was subsequently curbed by an increase in the use of contraceptives and other forms of birth control in the mid-1960s. The decline in fertility rate and an increase in the average population age has since become a global process, which prompted researchers to focus on projections by mid- and end of this century. Bongaarts (2009) claims that the developing worlds fertility rate of more than six children per female in the 1960s will have declined to fewer than two by mid-century, except for Africa where this ratio would still be above two. These trends lead to the creation of population columns with a subsequent conversion to inverted pyramids if fertility supporting policies fail to materialise. In a study for the UK it is claimed that a gradual increase in retirement age to 68.5 in 2050 and 70.2 in 2070 will keep the ratio of people over state-pension age (SPA+) to the cohort comprising those aged from 20 years until SPA relatively stable (Turner, 2009). An increase in contribution rates would alternatively delay the introduction of preferred retirement age (Lacomba and Lagos, 2006). However, the major threat is still imposed by declining fertility rates and inability of future working generations to absorb without a large illiquidity discount an overhang of assets managed on behalf of the retirement age population. This problem is

1 We examined 10-year government bond index returns from 1988q4 to 2013q1 using Datastream EFFAS indices and found that the compounding effects of coupons reinvested and the constant secular trend over this period distorts the pattern of real changes in the inflation-adjusted risk-free rate. Our primary focus is on yields.

A. Sˇ evi´c, D. Brawn / J. of Multi. Fin. Manag. 30 (2015) 36–61

Fig 1–5. Demographic age profiles Source: United Nations Statistics Division (UNSD).

39

40

A. Sˇ evi´c, D. Brawn / J. of Multi. Fin. Manag. 30 (2015) 36–61

Fig 1–5. (continued)

further aggravated by a transition from pay-as-you-go pension schemes and the expansion of stateowned reserve/sovereign funds.2 One of the first articles that linked the value of assets to population factors examined the impact on housing prices. The US citizens in the cohort between 20 and 35 years exert the strongest impact on the value of housing pricing, which coincides with the baby boom cohort entering the real estate market in the 1980s (Mankiw and Weil, 1989). Thereafter, researchers paid attention to the relationship between population and public stock markets. Using the Euler equation Bakshi and Chen (1994) show that aggregate consumption and demographic fluctuations could forecast risk premia. As the population ages more wealth is transferred to stock markets rather than housing markets. An average-age increase influences political risk ratings, while demographic factors can forecast long-term returns internationally (Claude et al., 1997).

2 Brooks (2002) shows that the retired baby boom generation will live better than parents and children and rejects the idea that these retirees should not support the current transition from defined-benefit to defined-contribution schemes.

A. Sˇ evi´c, D. Brawn / J. of Multi. Fin. Manag. 30 (2015) 36–61

41

By contrast, Poterba (2001) does not find evidence that demographic cohorts have relationship with Treasury bills, long-term government bonds and stock returns. Moreover, there is no abrupt decline in the asset demand over the period 2020–2050 (Poterba, 2004). The author does confirm that results are strongly influenced by the choice of econometric specifications. Davis and Li (2003) examine the relationship between demographic indicators and financial asset prices in selected OECD countries. Real asset returns are supported by the cohort with a population aged between 40 and 64. The cohort comprising people aged 65 and over drives stock prices down and supports an increase in bond yields, which is contrary to Poterbas forecasts and purports the view about depressed asset prices. The inclusion of a bequest in intergenerational transfers reduces the consumption of older generations and mitigates the meltdown of asset prices (Abel, 2001). Goyal (2004) extends the overlapping generations model (OLG) to four generations: Infants, Youth, Middles and Olds and proves that as the ratio of the middle-aged cohort increases the aggregate wealth becomes larger. In empirical tests, the outflow from stock markets is positively associated with changes in the age cohort of people older than 65 years, while it is negatively correlated with the middle-aged cohort (4565 years). The inclusion of the capital as a variable depending on the demographic factors allows for a direct relationship between age-structure and financial markets (Brunetti and Torricelli, 2008). In order to address non-stationarity of variables that mainly affect results in this area of research Arnott and Chaves (2012) apply age polynomials. It is also possible to evaluate the joint effect of all demographic variables and the procedure replaces the ad hoc use of age groups. Polynomials change the weights for age groups in order to obtain strong variables. Changes in demographic shares show that stock and bond returns increase for the 40–64-year cohort and decline thereafter. Bosworth (2014) shows that the linkage between growth and interest rates is at best tenuous and that the extent to which interest rates are determined within a global market versus a domestic context has become an area of research interest. Warnock and Warnock (2009) show that foreign official investors pushed down US nominal government bond yields by 80 basis points from the mid-1980s to around 2005. Similarly, Kaminska and Zinna (2014) find that such interventions reduced real long-term government bond yields between 2001 and 2008 also by 80 basis points. Bouis et al. (2014) highlight the impact of foreign reserve accumulation by central banks in emerging market economies (EMEs), which have grown rapidly since the mid-1990s and a large part of these reserves were invested in US assets, in particular US Treasury securities, given the international status of the US dollar, and the safety and liquidity of these assets. Consequently, official foreign investors, mainly comprising central banks and sovereign wealth funds, doubled their share in the US Treasury securities market from around 25% in the mid-1990s to more than 50% by end-2013. In contrast to many private investors, central banks in EMEs frequently buy US Treasuries with little consideration of their prices as purchases are largely determined by capital inflows (Krishnamurthy and Vissing-Jorgensen, 2012). Whether this rise in foreign reserves reflects a public policy objective or was as a result in the surge in private savings is largely irrelevant, the fact that they were invested in US government or government guaranteed fixedincome securities remains, contributing significantly to a decline in real interest rates in the first decade of 21st century (Beltran et al., 2013). The focus on high saving-to-GDP ratios in Asia and OPEC oilproducing countries was a major theme prior to the financial crisis. This savings ratio increased by more than 10 percentage points between 2000 and 2007, lifting the global savings rate by 1.7 percentage points, with China accounting for around half of all this growth by 2013 (Blanchard et al., 2014). Low rates are good news for fiscal policy, and increases in debt-financed government spending, especially public investment, may not lead to increases in public debt ratios in the medium term (DeLong and Summers, 2012). IMF calculations show that the 1% decline in the level of real rates during 2014, ceteris paribus, would reduce the advanced economy debt-to-GDP ratio 5 years ahead by about 4 percentage points (IMF, October 2013). The IMF (2014) suggests that there are upward pressures in interest rates in the medium term: population ageing and lower growth in emerging markets (implying lower savings rates), and according to the McKinsey Global Institute (2010) further financial deepening in emerging market economies would also reduce borrowing constraints and thereby net saving, thus arguing for increases in worldwide real interest rates. However, the effects of the global financial crisis persist, most notably evidenced by a sharp and persistent decline in investment in advanced economies that is unlikely to recover to pre-crisis levels in the next 5 years (IMF, 2014).

42

A. Sˇ evi´c, D. Brawn / J. of Multi. Fin. Manag. 30 (2015) 36–61

As a consequence of depressed economic activity globally, unconventional monetary policy has turned to quantitative easing (QE). Many studies investigating the impact of QE, mainly for the United States and the United Kingdom, have focused on event studies. They show that within a span of few days around the announcement of QE nominal government bond yields declined by 8 basis points on average for each 1% of GDP of asset purchase (Bouis et al., 2013). In the United States, Kaminska and Zinna (2014) find that QE measures in 2009–2012 reduced 10-year real US government bond yields by around 140 basis points (i.e. 13 basis points for purchases of 1% of 2009 GDP). Similarly, in the United Kingdom, QE is estimated to have pushed down real long-term interest rates by about 80 basis points in 2009 with the increase in inflation expectations playing only a marginal role (Hofmann and Zhu, 2013; Meaning and Zhu, 2011). Blundell-Wignall and Roulet (2014) show that foreign official and Fed interventions between the early 2000s and end-2012 reduced nominal yields by around 200 basis points. This accounted for around half of the decline in nominal yields during this period. Finally, between 2008 and 2013, certain empirical studies (D’Amico et al., 2012; Joyce et al., 2011) provide evidence that QE, in the form of long-term asset purchases, may have compressed real term premiums on long-term government bonds in the United States and United Kingdom, partially explaining the increase in the equity premium. QE in the advanced economies has only mitigated the effects of the zero lower bound, suggesting that natural real rates likely are negative now (IMF, 2014). 3. Methodology 3.1. Sources of data The primary source of demographic data used is the United Nations Statistics Division (UNSD) or UN database of annual population estimates (from 1950 to 2010). This is supplemented by data from; national statistics offices for each of the seven countries examined in this paper, for example the German agency de Statis produces annual estimates and projections by age and gender from 2009 to 2060, however, there has been only one population census since unification in 1990.3 Despite the infrequency of census surveys, detailed population estimates exist using monthly births, deaths and migration data, complimented by additional labour market surveys.4 For the UK, additional data was drawn from the Office for National Statistics (ONS) Demographic Analysis Unit who provide detailed intercensal population estimates by age, gender and region with projections from 1992 to 2033, as well as annual data from 1971 to 2011 by quinary age groups.5 Both the United States and Japan have ultra-long dated historical data available, from 1900 and 1920 to the present, respectively,6 and both include separate estimates for males, females and even by single-year age cohorts for the US. The US Department of Commerce: Census Bureau produces high frequency (monthly) population estimates by gender and age, with figures available from April 1990 to December 2012. This is not the norm though, with the European Union (EU) from 2011 insisting that all 27 member states carry out a population count at least once a decade, as is the case in the US. Additional data was sourced from the Australian Bureau of Statistics and Statistics New Zealand who publish annual resident population estimates by broad age group (for both genders) from 2002 to 2012 as well as quarterly estimates of resident population based upon changes due to natural increase or due to net migration. Intercensal population estimates often do not reconcile with the periodic census count, but for statistical purposes what is important is the ratios and whether or not they are consistent through time.7 This is what we focus on specific age structure cohorts, either as a proportion of the total population or as ratios to each other.

3 Source: www.destatis.de. The most recent German census took place on May 9, 2011. Prior to this date the last full population census in the former West Germany was conducted in 1987, whereas the last census taken in what was East Germany occurred in 1981. The first ever German census was taken in 1871. 4 The U.S. Department of Labour: Bureau of Labour Statistics is one example. 5 Quinary or 5-year age cohorts are the norm, typically starting with the 0–4 age group, then 5–9, 10–14, etc. 6 Annual Japanese population estimates from 1920 to 2000 are available, and from 1900 to 2012 for the U.S. 7 Technical data section UK National Statistics see: http://www.statistics.gov.uk/hub/population/population-change/ population-estimates/index.html.

A. Sˇ evi´c, D. Brawn / J. of Multi. Fin. Manag. 30 (2015) 36–61

43

Our approach to modelling the changing demographic age structures influence on long-term government bond yields is to include these population ratios alongside other financial and economic time series variables. Accordingly, we sourced inflation (CPI), Industrial Production and Gross Domestic Product (GDP) from the Thomson-Reuters Datastream database. For data on currencies, specifically Effective Exchange Rate Indices (EERI), we relied upon the Bank for International Settlements (BIS) monthly series that commences in January 1964.8 In some cases, using national central bank databases, we were able to go back as far as 1960 (Canada) or before. Finding 10-year government bond yields plus 3-month treasury-bill discount rates for seven different countries as far back as 1950 proved to be a challenge. However, we did source interest rate data for 1950–1989 from Homer and Sylla (2005) and continued the series using yield data from the various domestic central banks (as well as Datastream and Bloomberg for validation purposes). We also sourced 10-year Government Bond Returns Index (RI) data from Datastream (1988q4–2013q1). 3.2. Model specification Given the constraints of the availability of low frequency data, the maximum possible estimation period is 1950–2012 and we use an OLS five-factor model for each of the seven countries chosen (US, UK, Japan, Germany, Canada, Australia and New Zealand): RLRt = ˛ + ˇ1 SRt + ˇ2 YCt−1 + ˇ3 GDPt + ˇ4 Log(EERI)t + ˇ5 PopVart

(1)

where, RLR is the real long-rate, so 10-year government bond yield minus current annual inflation, SRt is the change in the three-month T-Bill rate (from year to year), YCt−1 is the one-period lagged change in the yield curve slope (long-bond less short-rate), GDPt represents the annual growth in economic output, IPt is the growth rate in Industrial Production (a proxy for economic ‘output’ where there was insufficient GDP time series data available, for example; Japan and New Zealand), Log(EERI)t is the log difference of the real effective exchange rate index, and PopVart is the particular demographic age structure metric of choice. We use 5 different population ratios (PopVart ) which are run in separate iterations of model to overcome the problem of correlated independent variables.9 The five selected demographic age structure ratios are: • • • • •

AGE20−39 is the number of persons aged 20–39 as a proportion of total population, AGE40−64 is the number of persons aged 40–64 as a percentage of total population, AGE65+ is the number of citizens aged 65 and over as a percentage of total population, DEP is the traditional Total Dependency ratio,10 and MY Ratio which is the proportion of middle-aged (age 40–64) workers over the proportion of the population that are younger workers (aged 20–39).

3.3. Variables: definition and interpretation We choose real 10-year bond-yield as our dependent variable, because we are interested in modelling one of the most significant financial asset determinants globally – the inflation-adjusted risk-free rate. Real yields on benchmark 10-year government bonds have been negative in recent years for some of the major markets; US (2009), UK (since 2010), Canada (2011) and Germany (in 2012), with other markets recording low single-digit real yields. Our hypothesis is that shifting demographic age structures may be increasing the demand for fixed-income assets. In order to test our theory, we

8

The B.I.S. real EERI are both trade-weighted and CPI-adjusted (see http://www.bis.org/statistics/eer/). The danger of using multiple demographic ratios is the potential problem of “over-fitting” (Poterba, 2001), thus to eliminate this problem, our proposal is to run successive models, changing the choice of population age structure ratio each time. Simple covariance and correlation analysis amongst the demographic ratios themselves shows that increased levels of cross-correlation occur, the lower the natural growth rate in total population. 10 Total Dependency ratio is defined as the ‘young’ (aged 0–19) and ‘old’ (aged 65+) as a proportion of the ‘working-age’ population (aged 20–64). Essentially it is the ratio of non-workers to workers (in theory). 9

44

A. Sˇ evi´c, D. Brawn / J. of Multi. Fin. Manag. 30 (2015) 36–61

need to model the inflation-adjusted 10-year government bond yield, using not just macro-economic variables, but also incorporating demographic age structures as well. We test for the significance of changing population age ratios as fertility rates have fallen11 and life expectancy has risen, to see if there is any potential attribution to the phenomenon of dramatic ageing in major developed countries on declining yields.12 Using a similar approach to Davis and Li (2003), our proposed model specification of the real long-term bond yield is derived in part from the expectations theory of the term structure (which applies strictly to the nominal long rate with zero inflation), suggesting that the long-term interest rate is based on future expected short rates. If the future short rate exceeds the current one the term structure will be upward sloping and vice versa if the future short-rate is below. Thus we include the change in the three month bill rate as our proxy for prospective monetary tightening or easing, and we include the slope of the yield curve itself as a reflection of both the financial markets perception of the future direction of official discount rates and as a proxy for credit liquidity. We allow for some mean reversion in the term structure by including a lag of the term structure differential. Economic expansion or contraction as measured by gross domestic product (GDP) or alternatively by a measure of manufacturing or industrial production, tends to have a direct effect on both real and nominal bond yields, not just because of the shift in inflationary expectations, but also due to the beneficial impact on total government debt ratios. In some cases growth may facilitate the repayment of the national debt, thus reducing the supply of government securities. Accordingly, we expect that there exists a positive relationship between economic growth and bond yields. Given that benchmark government bond markets differ in their geographic spread of investors, we deem it necessary to include a foreign-exchange rate variable as a proxy for foreign demand. The trade-weighted index fits these criteria. So we use the inflation-adjusted measure known as the real effective exchange rate index (real-EERI) as a predictor variable. An appreciating currency, in theory, is a function of low-inflation coupled with comparatively higher real interest rates.13 Population age structure ratios are deemed important because of what they tell us about the productive capacity of a nation. A baby-boom followed by a drop in fertility rates will lead to a demographic dividend in subsequent decades (as the population pyramid adjusts from ‘pine-tree’ shaped to a more columnar or ‘barrel-shaped’ one), as those persons enter the labour market reducing the overall dependency ratio. Further, if this increase in age cohort distribution is concentrated in the younger worker age category aged 20–39 then stronger economic growth will likely be experienced due to the higher productivity rate.14 And as this particular group gets older and we see an expansion of the ‘middle-aged’ worker population, the reverse is true – they save for retirement, invest in financial assets such as stocks and bonds. As workers reach late-50s and early 60s in their lives they seek to minimise capital risk and pursue income-protection strategies, switching out of higher risk securities to the safety of benchmark government bond. Hence, we expect to see a greater level of demand for fixed-income assets as the ratio of middle-aged to younger (we use MY ratio) increases. As the proportion of the population that is over 65 years old increases, or if the total dependency ratio rises, there may be either a positive or negative impact on bond yields depending on other factors such as: the savings ratio, real-asset alternatives or decumulation. In Tables 1 and 2 we summarise our expectations as to the influence of variables on the expected real long-term government bond yield. 3.4. Quarterly data – model specification Due to the lack of long-term low-frequency economic data we ran similar models using quarterly data from 1989 to 2013, again to test the significance of the population age structure variables. The

11 Most OECD developed world countries do not have a Total Fertility Rate (TFR) above the natural replacement rate of 2.1 (TFR is measured as the number of offspring per female of child-bearing aged 16–42). 12 Countries like Italy, Japan, and Germany have witnessed actual population declines since 2010. 13 Bi-lateral exchange rates at a fundamental level are a function of interest-rate plus inflation differential. 14 Young workers’ expenditure typically exceeds income as they borrow to purchase homes and capital items, plus life-cycle theory also suggests that they work harder for less money for career advancement.

A. Sˇ evi´c, D. Brawn / J. of Multi. Fin. Manag. 30 (2015) 36–61

45

Table 1 Demographic age structure variables and their pressure on real 10-year government bond yields.

Real 10-year bond yield

AGE20 –39

AGE40−64

AGE65+

DEP

MY ratio

+

−

±

±

−

Table 2 Economic variables and their expected influence on real 10-year government bond yields.

Real 10-year bond yield

SR

YCSlope

GDPGrowth

Ind. prod.

real-EERI

−

±

+

+

±

Fig. 6. US Federal Reserve balance sheet history (1994–2013).

regime shift is evident in demographic age profiles (see Fig. 1) from the late-1980s. Principally the rise in the number of those of working age combined with a sizeable reduction in dependency ratios globally reinforces our view that we should examine the last two decades separately. Our proposed model uses the same factors as those used in the longer-term annual model, however we added one additional variable to capture the impact of US quantitative easing on the global liquidity and demand for benchmark fixed-income assets.15 Regular data from the early-1990s is available from the US Federal Reserve (the Fed) pertaining to their purchase and outright holdings of US Treasury securities.16 We only use Fed purchases of US treasuries so we exclude other categories of bonds such as mortgage-backed and other non-treasury securities (Fig. 6). Our revised real bond yield model specification (using quarterly data): RLRt = ˛ + ˇ1 SRt + ˇ2 YCt−1 + ˇ3 GDPt + ˇ4 Log(EERI)t + ˇ5 Log(FED)t + ˇ6 PopVart where, Log(FED) is simply the log of the dollar amount of US treasuries on the Fed balance sheet. As Fig. 2 shows, prior to the financial market turmoil in 2008,17 the Fed held approximately $700 billion of US treasury securities on its own balance sheet, having already doubled from the early-2000s when quantitative easing (QE) first became a monetary policy tool utilised by central banks.18 As of May 29, 2013 some $3.1 trillion of securities are held by the Fed, of which US Treasury securities

15 We specifically used the US Federal Reserve’s quantitative easing (QE) programmes as the US Dollar is the primary global reserve currency, plus the US Treasury market is the largest and most liquid, thus serving as a global bond yield benchmark. Empirical evidence supports the idea that individual country investors’ usually look to two markets – their own (domestic investor bias) and that of the United States. 16 Factors affecting reserves balances http://www.federalreserve.gov/releases/h41/. 17 Specifically, the collapse of the US investment bank Lehman Brothers in mid-September 2008. 18 Quantitative easing (QE) was first used by the Bank of Japan in March 2001 to fight domestic deflation.

A. Sˇ evi´c, D. Brawn / J. of Multi. Fin. Manag. 30 (2015) 36–61

46

account for $1.88 trillion.19 Our rationale is to use this unusual treasury purchase activity as a control variable, to determine if demographic age structures are still relevant when QE measures are also taken into account. 3.5. Panel data estimation technique It is almost uniformly accepted to assume that initial procedures for modelling panel data sets include the application of: (a) the OLS on pooled time-series and cross-section data, (b) the fixedeffects model that more realistically purports the fact that omitted variables lead to the change in the intercept and (c) the random-effects model that assumes the existence of both cross-section and time-series disturbances. Let us specify the following variables:yit – the value of the dependent variable representing the individual unit i observed at time t,Xit – is the itth observation on K explanatory variables. Using the specification as in Baltagi (1995) and Pindyck and Rubinfeld (1998) in the somewhat changed form the basic panel data equation is: yit = ˛ + ˇXit + εit ,

i = 1, . . ., N;

t = 1, . . ., T

(2)

where εit = ui + it . It is important to distinguish that uit is unchanged over time and varies across individuals, while it denotes the common disturbance term that freely fluctuates in time-series and cross-sectionally. In the vector form the Eq. (2) is presented as follows: y = ˛NT + Xˇ + ε

(3)

and more generally y = Zı + ε. The decomposition of the error term in the vector form implies: ε = Z u + 

(4)

The combination of Eqs. (3) and (4) renders: y = ˛NT + Xˇ + Z u + 

(5)

i.e. y = Zı + Z u + . The model is evaluated using OLS estimation and implies extensive calculation due to the inclusion of N individual dummies. The Eq. (5) is multiplied by Q: Qy = QXˇ + Q

(6)

since the intercept part followed by the time-invariant disturbance term has trivial solutions. The modified OLS estimator is therefore: ˜ = (X  QX)−1 X  Qy ˇ

(7)

This procedure is similar to the GLS procedure that also renders the similar result. If Eq. (5) is rewritten for the simple regression, and averaged across all observations we obtain: y¯¯ = ˛ + ˇx¯¯ + ¯¯

N

(8)

u = 0, for it assures the avoidance Suits (1984) attaches particular attention to the fact that i−1 i of the dummy variable trap caused by perfect multi-collinearity. Another potential pitfall is the large loss of degrees of freedom because of the N − 1 additional parameters that are to be estimated. Greene (2000) underlines that the initial formulation (Eq. (2)) can be transformed by the deviations from the group means and in terms of group means that can be consistently estimated using OLS. The author also presents ways of calculating within- and between groups ␤ by calculating moment matrices.

19

The US Fed is now the largest single holder of US Treasury securities ahead of large creditor nations like China and Japan.

A. Sˇ evi´c, D. Brawn / J. of Multi. Fin. Manag. 30 (2015) 36–61

47

Table 3 Explanatory variables: ADF unit root tests (using intercept & trend) annual data.

Canada (p-values) United States Germany United Kingdom Japan Australia New Zealand

GDP

EERI

RLR

SR

−3.252* (0.0841) −3.844**

−7.426*** −6.491*** −6.828*** −8.70*** −6.327*** −5.369*** −0.582 (0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0002) (0.9765) −6.840*** −5.310*** −7.009*** −5.070*** −9.466*** −5.763*** −2.198

−5.037*** −6.081*** (0.0006) (0.0000) −5.686*** −6.104***

(0.0206) −3.595** (0.0395) −4.560***

(0.0000) −6.872*** (0.0000) −7.465***

(0.0003) −5.645*** (0.0001) −4.172***

(0.0000) −5.976*** (0.0000) −5.122***

(0.0008) −7.483*** (0.0000) −7.605***

(0.0000) −1.940 (0.6212) −2.140

(0.0001) −2.226 (0.4669) −2.669

(0.4818) −3.268* (0.0814) −1.837

(0.0001) −4.213*** (0.0076) −2.479

(0.0000) −0.886 (0.9506) −0.882

(0.0028) −3.277* (0.0832) −4.715*** (0.0017) −3.509**

(0.0000) −6.612*** (0.0000) −7.205*** (0.0000) −7.183***

(0.0085) −3.109 (0.1164) −4.678*** (0.0019) −3.445*

(0.0005) −7.460*** (0.0000) −6.342*** (0.0000) −7.396***

(0.0000) −6.983*** (0.0000) −7.28*** (0.0000) −6.551***

(0.5133) −4.068** (0.0114) −2.966 (0.1503) −7.058***

(0.2530) −1.816 (0.6848) −2.101 (0.5348) −2.130

(0.6729) −1.979 (0.6020) −1.222 (0.8968) −4.468**

(0.3371) −2.132 (0.5178) −4.338*** (0.0053) −2.983

(0.9511) −2.376 (0.3879) −2.669 (0.5532) −3.026

(0.0471) Panel (all 7) −2.153** (0.0156 (Levin, Lin and Chu)

YC−1

AGE20−39

AGE40−64

AGE65+

DEP

MY

(0.0000) (0.0548) (0.0000) (0.0000) (0.0000) (0.5187) (0.0039) (0.1455) (0.1338) −9.362*** −6.733*** −13.00*** −9.461*** −10.71*** −6.554*** −2.604*** −4.332*** −3.725*** (0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0000) (0.0046) (0.0000) (0.0001)

Note: RLR is the real long-rate, i.e. 10-year bond yield (nominal minus current inflation), SR the change in the short-rate (3 m bills), YC−1 the lagged slope of the term structure, GDP is economic output (as measured by GDP), IP is a change in industrial production, and EERI−2 a lagged foreign-exchange rate proxy (inflation-adjusted and trade-weighted effective exchange rate index) to reflect demand from overseas investors. Demographic independent variables are: AGE20−39 the percentage of the total population that is aged 20–39, AGE40−64 the percentage of the total population that is aged 40–64, AGE65+ the percentage of the total population that is aged over 65, DEP the ratio of young (0–19) and old (65+) as a percentage of the working population (aged 20–64). We also tested MY40−64/20−39 a ratio of AGE40−64 to AGE20−39 . * ** ***

Significant at 10% level. Significant at 5% confidence level. Significant at 1% confidence level.

Finally, the OLS estimator is denoted as the summation of the product of aforementioned beta vectors and moment matrices terms: w b bt = [Sxx + Sxx ]

−1 w w Sxx b

w b + (I − [Sxx + Sxx ]

−1 w Sxx )bb

(9)

where superscripts have the following notation t-total, w-within groups and b-between groups. Presented results are obtained on the basis of balanced panels, but in the real world missing observations are not rare. In this essay we will use unbalanced samples. 4. Variables 4.1. Testing the variables Table 3 summarises unit-root test results for all of the annual model variables. Stationarity issues arise as expected with certain population variables. Demographic age structure ratios by nature have a systematic secular change in mean over time (trend) and transition at a slow pace.20 In most instances the economic variables are unit-root free. If we were dealing with nominal bond yields this would not necessarily be true. Rather than using the average population age for each country, most empirical studies use aggregate age groups as a proportion of the total population. For example, Poterba (2001) uses the age cohort 40–64 as one of their variables and Yoo (1994) and Goyal (2004) use a variable based on the percentage of retirees in their studies.

20 We did expect a priori for unit-root problems to present when using non-stationary time series ratios. In most cases unit root issues did arise particularly with the proportion of persons over 65 years.

A. Sˇ evi´c, D. Brawn / J. of Multi. Fin. Manag. 30 (2015) 36–61

48

Table 4 Real long-term bond yield (panel data – annual frequency). Variables/model

I

II

III

IV

V

Constant (p-values) SR

−13.939*** (0.0000) −44.607*** (0.0000) 0.159*** (0.0030) 0.172*** (0.0000) 2.199*** (0.0062) 56.758*** (0.0000)

9.717*** (0.0000) −0.533*** (0.0000) 0.139 (0.1357) 0.193*** (0.0002) 1.430 (0.1135)

3.132*** (0.0000) −0.497*** (0.0000) 0.138* (0.0657) 0.192*** (0.0003) 1.138 (0.2462)

6.258*** (0.0000) −0.544*** (0.0000) 0.107 (0.1590) 0.299*** (0.0000) 0.713 (0.5526)

9.896*** (0.0001) −0.470*** (0.0001) 0.162** (0.0365) 0.127*** (0.0019) 2.066 (0.0082)

YC−1 GDP EERI−2 AGE20−39

−24.416*** (0.0001)

AGE40−64

−388.739*** (0.0001)

AGE65+

−5.717*** (0.0039)

DEP

−7.022*** (0.0000)

MY40−64/20−39 Adj-R2 F-statistic Durbin–Watson Observations

0.569 35.742 1.207 290

0.363 15.978 0.925 290

0.312 12.920 0.879 290

0.279 11.165 0.912 290

0.516 29.058 1.076 290

Note: The dependent variable is the real 10-year bond yield (nominal minus current inflation. The independent variables used are as follows: SR the change in the short-rate (3 m bills), YC−1 the lagged slope of the term structure, GDP is economic output (as measured by GDP or Industrial Production, whichever time series is more complete) and EERI−2 a lagged foreign-exchange rate proxy (inflation-adjusted and trade-weighted effective exchange rate index) to reflect demand from overseas investors. Demographic independent variables are: AGE20−39 the percentage of the total population that is aged 20 to 39, AGE40−64 the percentage of the total population that is aged 40 to 64, AGE65+ the percentage of the total population that is aged over 65, DEP the ratio of young (0–19) and old (65+) as a percentage of the working population (aged 20–64). We also tested MY40−64/20−39 a ratio of AGE40−64 to AGE20−39 . * Significant at 10% level. ** Significant at 5% confidence level. *** Significant at 1% confidence level.

4.2. Results for real bond yields: panel model Table 4 displays results from the estimation of the change in real 10-year government bond yields using panel estimation for all seven countries. The non-demographic specification is satisfactory with coefficient estimates generally as expected with the exception of the lagged real EERI variable, however as we shall see later in the individual country results, this is because currency is important for some countries, less so for others.21 Reductions in short-term interest rates are highly significant and reduce long-term real bond yields, particularly when the younger-worker demographic variable is included, as this age cohort is more leveraged thus more rate sensitive. Likewise, positive growth in economic output consistently leads to higher levels of real yields. The results for lagged term structure slope changes is quite consistent too, with every 100 basis points (bps) increase in curve steepening translating to a 10–15 bps increase in real 10-year bond yields. Again, highly intuitive. Conversely, the current low rate of economic growth globally, combined with moderately steep term structure but extremely low short rates22 more than

21 In commodity-based countries like Canada, New Zealand and Australia, the real effective exchange rate appears less important, unlike major reserve-currency markets like the UK and US where it is highly significant. 22 3 m bill yields in five out of the seven countries are sub-1% (with the exception of Australia and New Zealand).

A. Sˇ evi´c, D. Brawn / J. of Multi. Fin. Manag. 30 (2015) 36–61

49

offsetting the yield curve slope explains why negative or zero real 10-year benchmark government bond yields were still evident in 2012. All coefficient estimates for demographic age structure appear to be highly significant. Model variants I and V show the strongest influence. The increased proportion of young productive workers predicts a significant increase in real yields, whilst the higher proportion of middle-aged workers in the latter model indicates that real yields drop as the proportion of older; less productive, but higher earning workers increases relative to the younger proportion. A similar pattern is seen in variant II when only this older worker age cohort is used. 4.3. Results for real bond yields: individual country models Looking at the individual country results (see Appendix A: Tables A.2–A.7) we first notice that there is an omitted variable problem with highly significant constant or intercept coefficients, particularly in countries such as Canada, the US and Germany. However, the short-rate variable is showing a consistently negative sign across all seven countries and by a similar magnitude in each, for every given combination of population ratio variable. The term structure slope is barely statistically significant in Japan, but it is significant in Germany and highly so in Australia. There is an explanation for the highly significant negative slope coefficient for the United Kingdom for technical supply reasons, dating back to the introduction of the Minimum Funding Requirement (MFR) by New Labour in 1997. The legacy of this statutory-imposed demand for longer-dated UK gilts, combined with greater issuance at the shorter-dated maturities by the British Debt Management Office (DMO), has resulted in lower than normal nominal bond yields at a time when in recent years UK inflation was persistently above the Bank of England official target rate. A steeper gilt curve may prompt British institutional investors to purchase more longer-dated bonds despite unusually high levels of inflation.23 In the individual country equations, GDP positively affects the real yield in five out of the seven countries. In Japan and New Zealand, there was not enough historical data, so industrial production growth was used instead of GDP. This proved to be significant for Japan whereby growth in industrial output has a positive relationship with the real JGB yield. The difficulty with New Zealand is the larger gap in available data when compared to other countries.24 Our currency proxy variable, the real EERI, was highly significant in the US where effective exchange rates have a positive relationship with yields. A stronger US Dollar attracts higher levels of inward capital flows as evidenced by the enormous inflows into the US during the 1990s – into stocks and bonds. Li and Qiu (2013) also find that a rapidly appreciating currency (Chinese Yuan in this instance, from 2005 to 2008) attracts even more capital inflows. Given that the dollar remains the dominant global reserve currency, and that the US Treasury market is the largest and most liquid in the world, this strong exchange rate relationship is unsurprising. Since Germany is part of a common currency-bloc, we did not expect a significant result. Again the UK market varies with the US, in this instance effective exchange rate changes have a negative relationship with real yields suggesting that institutional fund investors favour UK gilts over foreign treasuries when the British Pound is strong. Examining German and UK bond yields, Swanson and Williams (2014) find that both the USD/GBP and USD/EUR exchange rates have been largely unaffected by the zero lower bound in recent years. Currency factors seem less important for commodity-based countries like Canada, Australia & New Zealand, in part, because their exchange rates are directly tied to the prices of global commodities.25

23 UK inflation, as measured by the Retail Prices Index (RPI) has been above 4.5% for 2010–2011 and remained above 3% during 2012. British real 10-year bond yields have been negative since 2009. 24 New Zealand GDP data is only available from 1987 onwards, with total manufacturing volume data beginning in June 1977. 25 For instance, when the price of a barrel of crude oil rises sharply, the Canadian Dollar appreciates substantially vis-a-vis currencies like the Japanese Yen (as Japan is a net importer of energy and Canada is world’s 8th largest exporter of crude oil). Thus in this respect it is the global price of internationally traded commodities (such as oil, wheat or precious metal) that determines these countries exchange rates.

50

A. Sˇ evi´c, D. Brawn / J. of Multi. Fin. Manag. 30 (2015) 36–61

Looking at the results of the regressions from a demographic perspective, several interesting themes emerge. The young working Canadian population is more interested in real estate and stocks which may be the reason for such a high positive impact for the AGE20–39 group. The same could be said for the United States, Australia, Germany, UK and New Zealand. All models are highly significant, indicating a strong positive relationship between younger-working age populations and higher bond yields. In deflationary Japan, the younger salaried worker is more interested in saving provided real rates are positive. The increase in the older or middle-aged working wealthiest cohort and the one with the largest disposable income should drive real yields downward. An important observation is that both the 40–64 age cohort and the over 65s in Japan appear to have positive impact on yields, which may be a country specific phenomenon, possibly reflecting a bank-based economy and less interest in financial markets in general. A recent study shows that the savings rate in Japan by age group over the 1998–2010 period has increased for both the 30–44 and 45–49 age cohorts, whilst the 60+ group has turned negative, especially when compared to the prior 1991–1997 period (Anderson et al., 2014). In each of the countries selected where we examined the influence of the over-65 age cohort we observed a significant positive relationship, indicating that the older population spends retirement income and may be more involved in wealth transfer rather than being concerned with the ‘bequeath motive’. Older people engage in decumulation whether it is through the purchase of annuities; where the proceeds are then reinvested by Life Assurance/Insurers across multiple asset classes, not just risk-free bonds, or alternatively by drawing down their wealth over time. Lower absolute bond yields and negative real yields reduce the attractiveness of long-term government bonds. However, further research into the persistently low absolute yield phenomenon would be required. Examination of the results for total dependency shows some rather unexpected results for US, Canada, Australia and New Zealand whereby the coefficient estimates suggest that a relative increase in non-working population reduces the bond yield. It must be noted though that from observing the charts in Figs. 1–5, that two distinct phenomena have occurred in recent decades. Firstly, total dependency ratios have fallen dramatically at a time when the over-65 age cohort has risen substantially – an indicator of dependency being dominated by seniors and a serious decline in fertility rates. Secondly, whilst the total dependency level across all countries studied has been comparatively stable for the last two decades, the secular decline in global bond yields has continued unabated. Thus it could be down to external factors prompting seniors to save more or it could be because decumulation has not occurred. The latter could be partly explained by the poor returns on stocks over the first decade of this century. Results of our derived variable: the Middle-to-Young or MY ratio is rather revealing. These ratios are highly significant for Canada, US, UK, and Germany, indicating that the wealthiest age cohort invests heavily in bond markets. The opposite appears to be the case with Japan where results were highly significant in one of the world’s most rapidly ageing economies, prompting the question if the wealthiest age cohort are not investing in bonds for retirement, then where are they putting their savings? This is why we postulate that Japan may be a bank-dominated market. 4.4. Results of the quarterly data analysis Table A.1 in Appendix A summarises the unit-root tests for all variables used in our second dataset, based on quarterly data. We report that the majority of the macroeconomic variables are significant at both the 1 and 5% confidence level. In some countries industrial production growth was superior as a proxy for economic growth, whereas in others GDP growth proved to be better. New Zealand is a case in point. Consistent with the annual data, few of the population age ratios were free of unit-root issues. Again, this was anticipated. We show the test for our global QE variable separately below this table to avoid unnecessary repetition. The LOG(FED) variable passed our unit-root test. Comparing across markets and between both datasets – annual and quarterly – is highly informative. Taking the younger worker age cohort first, the AGE20–39 , we see that this variable is highly

A. Sˇ evi´c, D. Brawn / J. of Multi. Fin. Manag. 30 (2015) 36–61

51

significant in all of our countries except Japan. In addition, for countries such as Canada, UK, Australia and New Zealand the coefficient estimate remains high or increased. These countries are destinations for young migrant workers. In Japan this variable was negative but consistent across datasets. Likewise when looking at the older worker age cohort Japan also shows an unexpected coefficient sign in the annual data model. The fact that the opposite sign for the AGE40–64 is observed in annual data, but does not persist in the quarterly series is also interesting. Country effect obviously matters. Results for this age group across the remaining countries show that both the coefficient estimates for US and Germany are consistently significant in that this wealthier cohort drives down bond yields. Canada, Australia and the UK all show a large increase in the importance of this group using the more recent quarterly dataset. New Zealand was unique with the coefficient sign changing from positive to expected negative when more recent data are used. Looking next at the older population and dependency ratios across countries only Germany is consistent between both datasets with ageing population and rising dependency depressing bond yields. Australia and New Zealand are interesting in that their coefficient estimates switch signs when moving from the longer-dated annual series to the more recent quarterly data and all are statistically significant. Using the quarterly series for both countries we witness a large negative sign with respect to the over 65s, as expected having a big impact on lower bond yields. Like Australia, the UK shows this AGE65+ variable to be extremely important using the higher frequency dataset, plus Canada shows a similar pattern but with similar magnitude as New Zealand. Perhaps the quarterly dataset is capturing more information. Finally the results of the MY-ratio is as expected, significant and consistent across Canada, UK, US, and Germany showing that where older wealthier workers dominate the downward pressure on yields exists. Both Australia and New Zealand are now highly significant for this older-younger worker ratio when examining the more recent quarterly data, but not significant using the earlier annual series. Japan once more remains an enigma with their MY-ratio highly significant albeit with opposite signs, just like other population age structure variables before. 5. Conclusion The main empirical findings are as follows; we detect a significant shift in the difference between quarterly and annual data for Japan as well as unexpected coefficient estimate sign; we measure the impact of US quantitative easing and we recognise that the lack of non-demographic data is a major constraint when performing low frequency data modelling that is necessary when dealing with shifting population age structures that change at a slow pace. We see that the impact of the US Federal Reserves actions in their domestic government bond market, purchasing almost two trillion dollars of treasuries has had a pronounced effect on bond yields in US, Canada, Germany and the UK. Based on the coefficient estimates for this variable we can measure this influence of quantitative easing (QE) over the past 5-year period and it shows that QE has at the very least reduced real 10-year bond yields by approximately 50–100 basis points. It also explains why real yields have been negative in major bond markets. Country effects do exist; Japan is a prime example of a rapidly ageing nation with some of the lowest bond yields, yet the population age structure ratios appear to be contradictory in terms of their coefficient estimate signs. This suggests to us that other influences may dominate there such as a large banking sector. Our results indicate that alongside traditional macroeconomic variables and exceptional factors like quantitative easing, demographic age structures have a strong influence on real bond yields. However, further country research would seem logical, both to broaden the base as well as the inclusion of countries with rapidly ageing populations such as; Belgium, Italy, Denmark and other Scandinavian countries. Appendix A. Regression results output See Tables A.1–A.8.

52

Table A.1 Explanatory variables – summary of ADF tests (intercept and trend) using quarterly data.

Germany United Kingdom Japan Australia New Zealand Panel (all 7) (Levin, Lin and Chu)

SR

−3.106 (0.1110) −7.049*** (0.0000) −3.408* (0.0562) −4.499*** (0.0025) −4.583*** (0.0019) −4.394*** (0.0036) −6.007*** (0.0000) 0.178 (0.5705)

−9.136 (0.0000) −3.913** (0.0152) −5.756*** (0.0000) −6.071*** (0.0000) −6.847*** (0.0000) −6.390*** (0.0000) −6.362*** (0.0000) −9.311** (0.0000)

GDP

YC−1 ***

−9.112 (0.0000) −8.362*** (0.0000) −7.240*** (0.0000) −7.682*** (0.0000) −11.20*** (0.0000) −8.047*** (0.0000) −7.881*** (0.0000) −20.31*** (0.0000) ***

IP

−3.446 (0.0519) −3.834** (0.0188) −7.054*** (0.0000) −2.728 0.2280) −2.870 (0.1770) −4.416*** (0.0034) −9.329*** (0.0000) −0.205 (0.4186) *

EERI

−6.331 (0.0000) −5.494*** (0.0001) −6.216*** (0.0000) −9.776*** (0.0000) −8.742*** (0.0000) −9.302*** (0.0000) −2.765 (0.2140) −5.977*** (0.0000) ***

−7.746 (0.0000) −9.123*** (0.0000) −9.462*** (0.0000) −8.064*** (0.0000) −4.380*** (0.0038) −9.175*** (0.0000) −9.143*** (0.0000) −20.72*** (0.0000) ***

AGE20−39

AGE40−64

AGE65+

DEP

2.140 (1.0000) −0.467 (0.9834) −1.806 (0.6941) −3.500** (0.0451) −2.597 (0.2829) −2.918 (0.1618) −2.931 (0.1575) 1.838 (0.9670)

−5.313 (0.000) 0.941 (0.9998) −1.467 (08343) −1.753 (0.7198) −2.199 (0.4838) −1.532 (0.8114) −2.032 (0.5762) 0.838 (0.7989)

−2.243 (0.4600) −0.609 (0.9759) −3.659** (0.0301) −0.040 (0.9952) −2.806 (0.1989) −1.045 (0.932) −2.278 (0.4412) 0.409 (0.6587)

−4.011 (0.0117) −2.268 (0.4467) −0.230 (0.9915) −0.087 (0.9945) −2.688 (0.2440) −2.141 (0.5162) −1.534 (0.8111) −3.275*** (0.0005)

***

MY **

−2.113 (0.5315) −0.186 (0.9925) −1.567 (0.7985) −3.053 (0.1241) −2.193 (0.4880) −2.694 (0.2418) −2.883 (0.1726) −0.215 (0.4150)

Note: RLR is the real long-rate, i.e. 10-year bond yield (nominal minus current inflation), SR the change in the short-rate (3 m bills), YC−1 the lagged slope of the term structure, GDP is economic output (as measured by GDP), IP is a change in industrial production, and EERI−2 a lagged foreign-exchange rate proxy (inflation-adjusted and trade-weighted effective exchange rate index) to reflect demand from overseas investors. Demographic independent variables are: AGE20−39 the percentage of the total population that is aged 20–39, AGE40−64 the percentage of the total population that is aged 40–64, AGE65+ the percentage of the total population that is aged over 65, DEP the ratio of young (0–19) and old (65+) as a percentage of the working population (aged 20–64). We also tested MY40−64/20−39 a ratio of AGE40−64 to AGE20−39 . * Significant at 10% level. ** ***

Significant at 5% confidence level. Significant at 1% confidence level.

A. Sˇ evi´c, D. Brawn / J. of Multi. Fin. Manag. 30 (2015) 36–61

Canada (p-values) United States

RLR

Table A.2 Canada – real yield bond yield models. Variables/model

Constant (p-values) SR

GDP or IP EERI−2

Quarterly

I

II

III

IV

V

I

II

III

IV

V

−10.62*** (0.0004) −0.419** (0.0118) 0.045 (0.8487) 0.315** (0.0429) 1.734 (0.6946)

5.267* (0.0646) −0.336* (0.0924) −0.081 (0.8034) 0.121 (0.5841) −2.801 (0.6003)

0.420 (0.8509) −0.368* (0.0650) −0.380 (0.2313) 0.386* (0.0980) −4.148 (0.4280)

6.817*** (0.0002) −0.412** (0.0318) −0.438 (0.1257) 0.539** (0.0151) −3.297 (0.5061)

7.587*** (0.0001) −0.343* (0.0655) 0.151 (0.6100) 0.025 (0.8927) 0.045 (0.9929)

−8.472 (0.1212) 0.746*** (0.0051) 0.984*** (0.0010) 3.286 (0.5361) 1.213 (0.7365) −1.414*** (0.0010) 70.906*** (0.0000)

28.41*** (0.0000) 0.769*** (0.0052) 0.995*** (0.0013) 0.193*** (0.0002) 0.782*** (0.1135) −1.454*** (0.0016)

29.06*** (0.0000) 0.738** (0.0219) 0.875** (0.0146) 12.35** (0.0508) −1.53 (0.4207) −1.531* (0.0611)

−19.424* (0.0547) 0.821*** (0.0059) 1.001*** (0.0029) 6.606 (0.2964) −0.313 (0.9379) −1.701*** (0.0011)

21.862*** (0.0000) 0.774*** (0.0056) 0.991*** (0.0016) 4.402 (0.4319) 0.619 (0.8698) −1.543*** (0.0008)

LOG(FED) AGE20−39

42.58*** (0.0000) −7.814 (0.3902)

AGE40−64 AGE65+

−48.78*** (0.0001) −124.61*** (0.0062)

20.517 (0.2483) −6.395** (0.0149)

DEP

−4.793*** (0.0075)

MY40−64/20−39 Adj-R2 F-statistic Durbin–Watson Observations

54.543*** (0.0000)

0.341 6.275 0.704 52

0.010 1.101 0.561 52

0.023 1.237 0.631 52

0.117 2.345 0.719 52

0.140 2.567 0.596 52

−7.782*** (0.0000) 0.801 52.041 0.931 77

0.787 47.848 0.878 77

0.705 31.219 0.686 77

0.749 38.925 0.776 77

0.780 45.940 0.855 77

A. Sˇ evi´c, D. Brawn / J. of Multi. Fin. Manag. 30 (2015) 36–61

YC−1

Annual

Note: The dependent variable is the real 10-year government bond yield (nominal yield minus current annual inflation). Independent variables include: the change in the 3-month short-rate (SR), the lagged change in the slope of the yield curve (YC−1 ), an economic output proxy or growth in GDP (GDP) or change in industrial production (IP), an exchange rate measure – the lagged change in the real effective exchange rate index (EERI−2 ), plus the US Fed’s balance sheet holding of US Treasury securities in LOG [US$ billions] designated below by LOG (FED). The demographic variables are population age ratios such as proportion of the total population aged 20–39 (AGE20−39 ), aged 40–64 (AGE40−64 ), and over the age of 65 (AGE65+ ), as well as the total dependency ratio (DEP) and finally, the relative ratio of ‘middle-aged’ workers to ‘younger’ ones (MY40−64/20−39 ). * Significant at 10% level. ** ***

Significant at 5% confidence level. Significant at 1% confidence level. 53

54

Table A.3 United States – real yield bond yield models. Variables/model

Constant (p-values) SR

GDP or IP EERI−2

Quarterly

I

II

III

IV

V

I

II

III

IV

V

−11.641*** (0.0001) −1.108*** (0.0000) −0.070 (0.7502) 0.462*** (0.0019) 13.210** (0.0150)

8.469** (0.0131) −1.118*** (0.0000) 0.357 (0.8034) 0.291 (0.1020) 18.303*** (0.0020)

−8.133* (0.0627) −1.176*** (0.0000) −0.309 (0.3464) 0.634*** (0.0015) 11.320* (0.0877)

9.084*** (0.0084) −1.178*** (0.0000) −0.304 (0.3289) 0.647*** (0.0010) 11.655* (0.0702)

7.110*** (0.0003) −1.073*** (0.0000) 0.229 (0.3354) 0.330** (0.0384) 16.516*** (0.0052)

9.301* (0.0586) 0.586*** (0.0035) 0.447*** (0.0060) 15.836** (0.014) 4.862* (0.0749) −2.067*** (0.0000) 28.112*** (0.0073)

23.579*** (0.0000) 0.604*** (0.0026) 0.449*** (0.0060) 15.918** (0.0117) 4.745* (0.0836) −2.087*** (0.0000)

22.631*** (0.0000) 0.653*** (0.0019) 0.416** (0.0155) 24.13*** (0.0003) 5.644* (0.0535) −2.828*** (0.0000)

4.717 (0.5578) 0.648*** (0.0014) 0.447*** (0.0070) 17.365*** (0.0066) 4.641* (0.0967) −2.249*** (0.0000)

20.87*** (0.0000) 0.606*** (0.0026) 0.451*** (0.0057) 16.633** (0.0130) 4.781* (0.0801) −2.045*** (0.0000)

LOG(FED) AGE20−39

44.04*** (0.0000) −25.86** (0.0278)

AGE40−64 AGE65+

−19.885*** (0.0095) −3.858 (0.8898)

79.40** (0.0333) −10.64** (0.0182)

DEP

−6.033*** (0.0011)

MY40−64/20−39 Adj-R2 F-statistic Durbin–Watson Observations

20.087** (0.0299)

0.577 14.103 1.409 49

0.429 8.222 1.216 49

0.425 8.100 1.060 49

0.439 8.514 1.094 49

0.502 10.678 1.313 49

−3.390*** (0.0075) 0.814 56.523 1.882 77

0.813 56.077 1.871 77

0.794 49.843 1.740 77

0.806 54.145 1.825 77

0.814 56.493 1.883 77

Note: The dependent variable is the real 10-year government bond yield (nominal yield minus current annual inflation). Independent variables include: the change in the 3-month short-rate (SR), the lagged change in the slope of the yield curve (YC−1 ), an economic output proxy or growth in GDP (GDP) or change in industrial production (IP), an exchange rate measure – the lagged change in the real effective exchange rate index (EERI−2 ), plus the US Fed’s balance sheet holding of US Treasury securities in LOG [US$ billions] designated below by LOG (FED). The demographic variables are population age ratios such as proportion of the total population aged 20–39 (AGE20−39 ), aged 40–64 (AGE40−64 ), and over the age of 65 (AGE65+ ), as well as the total dependency ratio (DEP) and finally, the relative ratio of ‘middle-aged’ workers to ‘younger’ ones (MY40−64/20−39 ). * Significant at 10% level. **

Significant at 5% confidence level.

***

Significant at 1% confidence level.

A. Sˇ evi´c, D. Brawn / J. of Multi. Fin. Manag. 30 (2015) 36–61

YC−1

Annual

Table A.4 Australia – real yield bond yield models. Variables/model

Constant (p-values) SR

GDP EERI−2

(−1 annual)

Quarterly

I

II

III

IV

V

I

II

III

IV

V

−37.938*** (0.0001) −0.606** (0.0286) 0.615 (0.1298) 0.317** (0.0394) 4.731 (0.2771)

1.079 (0.8448) −0.642* (0.0664) 0.892* (0.0785) 0.158 (0.3957) 2.328 (0.6711)

−6.028* (0.0915) −0.472 (0.1527) 0.770 (0.1065) 0.146 (0.3983) 0.596 (0.9073)

18.699*** (0.0013) −0.408 (0.1892) 0.741* (0.0994) 0.149 (0.3589) 0.329*** (0.0027)

6.168 (0.1214) −0.695** (0.0422) 0.850* (0.0880) 0.203 (0.2712) 3.518 (0.5134)

−29.386*** (0.0004) 0.626** (0.0203) 0.737*** (0.0035) 0.199*** (0.0067) 6.795** (0.0374) −0.4598 (0.3178) 118.311*** (0.0000)

30.168*** (0.0000) 0.594** (0.0323) 0.738*** (0.0044) 0.191** (0.0113) 6.746** (0.0439) −0.147 (0.7784)

32.767*** (0.0000) 0.615* (0.0618) 0.653** (0.0289) 0.101 (0.2290) 6.052 (0.9526) −0.057 (0.9526)

−59.356*** (0.0025) 0.577* (0.0613) 0.700** (0.0135) 0.147* (0.0683) 6.557* (0.0744) 0.0281 (0.9690)

19.966*** (0.0000) 0.615** (0.0262) 0.738*** (0.0043) 0.194*** (0.0098) 6.861** (0.0400) −0.312 (0.5317)

LOG(FED) AGE20−39

129.62*** (0.0000)

AGE40−64

−86.79*** (0.0000)

1.874 (0.9247)

AGE65+

−231.51*** (0.0000)

69.71** (0.0306) −23.51*** (0.0027)

DEP

−5.044 (0.2367)

MY40−64/20−39 Adj-R2 F-statistic Durbin–Watson Observations

93.58*** (0.0001)

0.395 6.492 0.621 43

0.045 1.396 0.392 43

0.160 2.595 0.391 43

0.254 3.859 0.415 43

0.081 1.372 0.426 43

−14.741*** (0.0000) 0.647 24.195 0.70 77

0.628 22.381 0.812 77

0.494 13.371 0.543 77

0.548 16.395 0.637 77

0.630 22.573 0.826 77

A. Sˇ evi´c, D. Brawn / J. of Multi. Fin. Manag. 30 (2015) 36–61

YC−1

Annual

Note: The dependent variable is the real 10-year government bond yield (nominal yield minus current annual inflation). Independent variables include: the change in the 3-month short-rate (SR), the lagged change in the slope of the yield curve (YC−1 ), an economic output proxy or growth in GDP (GDP) or change in industrial production (IP), an exchange rate measure – the lagged change in the real effective exchange rate index (EERI−2 ), plus the US Fed’s balance sheet holding of US Treasury securities in LOG [US$ billions] designated below by LOG (FED). The demographic variables are population age ratios such as proportion of the total population aged 20–39 (AGE20−39 ), aged 40–64 (AGE40−64 ), and over the age of 65 (AGE65+ ), as well as the total dependency ratio (DEP) and finally, the relative ratio of ‘middle-aged’ workers to ‘younger’ ones (MY40−64/20−39 ). * Significant at 10% level. ** ***

Significant at 5% confidence level. Significant at 1% confidence level. 55

56

Table A.5 Germany – real yield bond yield models. Variables/model

Constant (p-values) SR

GDP or IP EERI−2

Quarterly

I

II

III

IV

V

I

II

III

IV

V

−10.589*** (0.0000) −0.369*** (0.0010) 0.186** (0.0470) 0.084 (0.3967) −1.094 (0.7756)

9.906** (0.0012) −0.379* (0.0570) 0.192 (0.1027) 0.068 (0.6033) −0.766 (0.8729)

10.416*** (0.0000) −0.350*** (0.0027) 0.221** (0.0273) −0.036 (0.7487) −2.350 (0.5635)

5.569** (0.0096) −0.417*** (0.0037) 0.111 (0.3548) 0.225* (0.0901) −0.919 (0.8546)

9.948*** (0.0000) −0.359*** (0.0023) 0.222** (0.0281) 0.009 (0.9364) −1.001 (0.8062)

6.801** (0.0359) −0.370** (0.0347) 0.397*** (0.0033) 0.0867*** (0.007) 0.399 (0.9248) −1.712*** (0.0000) 24.95*** (0.0000)

26.262*** (0.0000) −0.413** (0.0195) 0.419*** (0.0020) 0.0796** (0.0145) −2.200 (0.6057) −1.531*** (0.0000)

19.676*** (0.0000) −0.386** (0.0267) 0.409*** (0.0023) 0.092*** (0.0038) 0.349 (0.9334) −1.795*** (0.0000)

27.496*** (0.0000) −0.309* (0.0956) 0.370*** (0.0096) 0.109*** (0.0014) 1.687 (0.7087) −2.385*** (0.0000)

17.543*** (0.0000) −0.415** (0.0175) 0.420*** (0.0017) 0.082** (0.0107) −1.360 (0.7543) −1.540*** (0.0000)

LOG(FED) AGE20−39

48.53*** (0.0000) −21.164** (0.0186)

AGE40−64

−40.144*** (0.0000) −44.47*** (0.0000)

AGE65+

−30.72*** (0.0000) −3.953 (0.2101)

DEP

−5.896*** (0.0000)

MY40−64/20−39 Adj-R2 F-statistic Durbin–Watson Observations

−15.361*** (0.0005)

0.481 10.080 1.363 50

0.192 3.322 0.879 50

0.422 8.155 1.194 50

0.114 2.265 0.906 50

0.413 7.909 1.184 50

−3.877*** (0.0000) 0.859 78.422 0.873 77

0.859 78.295 0.879 77

0.862 80.301 0.907 77

0.841 67.890 0.816 77

0.863 80.609 0.900 77

Note: The dependent variable is the real 10-year government bond yield (nominal yield minus current annual inflation). Independent variables include: the change in the 3-month short-rate (SR), the lagged change in the slope of the yield curve (YC−1 ), an economic output proxy or growth in GDP (GDP) or change in industrial production (IP), an exchange rate measure – the lagged change in the real effective exchange rate index (EERI−2 ), plus the US Fed’s balance sheet holding of US Treasury securities in LOG [US$ billions] designated below by LOG (FED). The demographic variables are population age ratios such as proportion of the total population aged 20–39 (AGE20−39 ), aged 40–64 (AGE40−64 ), and over the age of 65 (AGE65+ ), as well as the total dependency ratio (DEP) and finally, the relative ratio of ‘middle-aged’ workers to ‘younger’ ones (MY40−64/20−39 ). * Significant at 10% level. **

Significant at 5% confidence level.

***

Significant at 1% confidence level.

A. Sˇ evi´c, D. Brawn / J. of Multi. Fin. Manag. 30 (2015) 36–61

YC−1

Annual

Table A.6 United Kingdom – real yield bond yield models. Variables/model

Constant (p-values) SR

GDP EERI−1

Quarterly

I

II

III

IV

V

I

II

III

IV

V

−14.44** (0.0151) −0.588*** (0.0017) −0.392** (0.0299) 0.609** (0.0002) −9.853*** (0.0078)

8.992 (0.1007) −0.662*** (0.0008) −0.572*** (0.0012) 0.685*** (0.0001) −9.191** (0.0172)

−0.815 (0.8289) −0.643*** (0.0015) −0.599*** (0.0009) 0.733*** (0.0000) −8.906** (0.0237)

4.103 (0.4059) −0.635*** (0.0017) −0.577*** (0.0019) 0.161*** (0.0000) −8.834** (0.0241)

8.02** (0.0156) −0.631*** (0.0010) −0.482*** (0.0064) 0.641*** (0.0001) −9.650** (0.0106)

−19.676* (0.0541) 0.634* (0.0676) 1.156*** (0.0002) −0.268*** (0.0011) 3.372 (0.4029) −2.053*** (0.0003) 127.77*** (0.0000)

45.989*** (0.0000) 0.671* (0.0540) 1.152*** (0.0002) −0.278*** (0.0009) 3.596 (0.3717) −2.143*** (0.0001)

63.616*** (0.0000) 0.642 (0.1005) 0.989*** (0.0032) −0.192** (0.0420) 7.291 (0.1013) −3.148*** (0.0000)

−53.534*** (0.0000) 0.737** (0.0404) 1.117*** (0.0004) −0.272*** (0.0017) 4.728 (0.2496) −2.658*** (0.0000)

32.757*** (0.0001) 0.675* (0.0515) 1.158*** (0.0002) −0.288*** (0.0006) 3.397 (0.3968) −2.029*** (0.0003)

LOG(FED) AGE20−39

55.771*** (0.0099) −26.737 (0.1370)

AGE40−64 AGE65+

−93.843*** (0.0000) −253.58** (0.0182)

11.238 (0.6464) −4.361 (0.5139)

DEP

−6.456* (0.0900)

MY40−64/20−39 Adj-R2 F-statistic Durbin–Watson Observations

106.74*** (0.0000)

0.552 12.819 1.687 49

0.502 10.694 1.652 49

0.478 9.806 1.606 49

0.481 9.900 1.593 49

0.531 11.894 1.676 49

−14.98*** (0.0000) 0.799 51.556 1.014 77

0.799 51.472 1.015 77

0.747 38.478 0.854 77

0.788 48.229 0.983 77

0.801 52.130 1.021 77

A. Sˇ evi´c, D. Brawn / J. of Multi. Fin. Manag. 30 (2015) 36–61

YC−1

Annual

Note: The dependent variable is the real 10-year government bond yield (nominal yield minus current annual inflation). Independent variables include: the change in the 3-month short-rate (SR), the lagged change in the slope of the yield curve (YC−1 ), an economic output proxy or growth in GDP (GDP) or change in industrial production (IP), an exchange rate measure – the lagged change in the real effective exchange rate index (EERI−1 ), plus the US Fed’s balance sheet holding of US Treasury securities in LOG [US$ billions] designated below by LOG (FED). The demographic variables are population age ratios such as proportion of the total population aged 20–39 (AGE20−39 ), aged 40–64 (AGE40−64 ), and over the age of 65 (AGE65+ ), as well as the total dependency ratio (DEP) and finally, the relative ratio of ‘middle-aged’ workers to ‘younger’ ones (MY40−64/20−39 ). * Significant at 10% level. ** ***

Significant at 5% confidence level. Significant at 1% confidence level. 57

58

Table A.7 Japan – real yield bond yield models. Variables/model

Constant (p-values) SR

IP EERI−2

Quarterly

I

II

III

IV

V

I

II

III

IV

V

14.10*** (0.0029) −1.912*** (0.0020) 1.135* (0.0746) 12.227* (0.053) −1.094 (0.8643)

−11.901** (0.0113) −1.810*** (0.0047) 0.861 (0.1800) 11.671* (0.0740) −0.861 (0.8513)

−3.642 (0.1511) −2.121*** (0.0016) 1.367* (0.0721) 8.615 (0.1984) −2.093 (0.6638)

3.132 (0.7901) −2.038*** (0.0045) 0.726 (0.3242) 4.355 (0.5116) −3.852 (0.4376)

−7.737** (0.0117) −1.885*** (0.0029) 1.086* (0.0937) 11.837* (0.0661) −0.853*** (0.0053)

15.354* (0.0516) −1.074** (0.0170) 0.880*** (0.0027) 6.366*** (0.0081) −1.846 (0.3266) −1.156*** (0.0053) −22.685 (0.2670)

−11.160 (0.4983) −1.023** (0.0138) 0.875*** (0.0029) 5.444** (0.0234) −1.244 (0.5104) −0.365 (0.4163)

4.660** (0.0290) −1.106** (0.0129) 0.865*** (0.0030) 5.132** (0.0328) −1.144 (0.5418) −0.140 (0.7782)

6.304*** (0.0011) −1.120** (0.0130) 0.887** (0.0027) 6.176** (0.0113) −1.728 (0.3638) −0.988** (0.0418)

−0.618 (0.8949) −1.014** (0.0232) 0.869*** (0.0028) 6.311*** (0.0075) −1.781*** (0.0004) −1.116*** (0.0004)

LOG(FED) AGE20−39

−50.43*** (0.0021)

AGE40−64

37.10** (0.0182)

AGE65+

44.57 (0.2741)

−4.608 (0.7931)

DEP MY

−10.603 (0.1474)

18.702* (0.0852)

40−64/20−39

Adj-R2 F-statistic Durbin–Watson Observations

2.926 (0.6275) ***

7.551* (0.0964)

6.495 (0.0053) 0.313 5.108 1.128 46

0.269 4.310 1.032 46

0.190 3.117 0.915 46

0.129 2.332 0.805 46

0.283 4.555 1.068 46

0.309 6.664 0.833 77

0.308 6.654 0.767 77

0.317 6.894 0.759 77

0.299 6.403 0.815 77

0.324 7.072 0.834 77

Note: The dependent variable is the real 10-year government bond yield (nominal yield minus current annual inflation). Independent variables include: the change in the 3-month short-rate (SR), the lagged change in the slope of the yield curve (YC−1 ), an economic output proxy or growth in GDP (GDP) or change in industrial production (IP), an exchange rate measure – the lagged change in the real effective exchange rate index (EERI−1 ), plus the US Fed’s balance sheet holding of US Treasury securities in LOG [US$ billions] designated below by LOG (FED). The demographic variables are population age ratios such as proportion of the total population aged 20–39 (AGE20−39 ), aged 40–64 (AGE40−64 ), and over the age of 65 (AGE65+ ), as well as the total dependency ratio (DEP) and finally, the relative ratio of ‘middle-aged’ workers to ‘younger’ ones (MY40−64/20−39 ). * Significant at 10% level. **

Significant at 5% confidence level.

***

Significant at 1% confidence level.

A. Sˇ evi´c, D. Brawn / J. of Multi. Fin. Manag. 30 (2015) 36–61

YC−1

Annual

Table A.8 New Zealand – real yield bond yield models. Variables/model

Constant (p-values) SR

IP or GDP EERI−1

Quarterly

I

II

III

IV

V

I

II

III

IV

V

−13.847 (0.1713) −0.467 (0.1075) 0.173 (0.7160) −3.765 (0.7894) −1.962 (0.7823)

−1.529 (0.7791) −0.351 (0.2443) −0.300 (0.5398) 9.223 (0.5375) −5.490 (0.4623)

−12.015* (0.0957) −0.230 (0.4213) −0.567 (0.2320) 17.354 (0.2302) −7.833 (0.2695)

26.141*** (0.0002) −0.215 (0.3918) −0.453 (0.2504) 18.003 (0.1458) −8.465 (0.1775)

3.309 (0.3786) −0.404 (0.1864) −0.148 (0.7670) 4.340 (0.7751) −4.063 (0.5910)

−9.927 (0.1685) 0.681*** (0.0085) 0.817*** (0.0048) 0.0195 (0.8959) 1.355 (0.6761) −0.487 (0.3549) 59.119*** (0.0001)

21.217*** (0.0000) 0.678*** (0.0089) 0.819*** (0.0048) 0.012 (0.9358) 1.264 (0.6978) −0.518 (0.3250)

24.539*** (0.0000) 0.687** (0.017) 0.714** (0.0219) 0.102 (0.5424) 1.689 (0.6332) −1.133 (0.1076)

−36.107** (0.0253) 0.676** (0.0111) 0.800*** (0.002) 0.020 (0.8967) 1.117 (0.7389) −0.875* (0.0870)

14.793*** (0.0000) 0.682*** (0.0080) 0.832*** (0.0039) −0.0011 (0.9941) 1.332 (0.6796) −0.402 (0.4496)

LOG(FED) AGE20−39

58.0* (0.0893)

AGE40−64

−46.60*** (0.0001)

18.262 (0.3591)

AGE65+

−110.85* (0.0617)

137.05** (0.0333) −30.27*** (0.0008)

DEP MY40−64/20−39 Adj-R2 F-statistic Durbin–Watson Observations

64.81*** (0.0008) −7.878*** (0.0001)

0.176 (0.9647) 0.053 1.383 1.030 35

−0.017 0.886 1.010 35

0.106 1.809 1.208 35

0.291 3.792 1.507 35

−0.047 0.692 0.961 35

0.569 17.761 0.879 77

0.566 17.582 0.873 77

0.490 13.172 0.737 77

0.543 16.073 0.825 77

0.574 18.086 0.890 77

A. Sˇ evi´c, D. Brawn / J. of Multi. Fin. Manag. 30 (2015) 36–61

YC−1

Annual

Note: The dependent variable is the real 10-year government bond yield (nominal yield minus current annual inflation). Independent variables include: the change in the 3-month short-rate (SR), the lagged change in the slope of the yield curve (YC−1 ), an economic output proxy or growth in GDP (GDP) or change in industrial production (IP), an exchange rate measure – the lagged change in the real effective exchange rate index (EERI−1 ), plus the US Fed’s balance sheet holding of US Treasury securities in LOG [US$ billions] designated below by LOG (FED). The demographic variables are population age ratios such as proportion of the total population aged 20–39 (AGE20−39 ), aged 40–64 (AGE40−64 ), and over the age of 65 (AGE65+ ), as well as the total dependency ratio (DEP) and finally, the relative ratio of ‘middle-aged’ workers to ‘younger’ ones (MY40−64/20−39 ). * Significant at 10% level. ** ***

Significant at 5% confidence level. Significant at 1% confidence level. 59

60

A. Sˇ evi´c, D. Brawn / J. of Multi. Fin. Manag. 30 (2015) 36–61

References Abel, A.B., 2001. Will bequests attenuate the predicted meltdown in stock prices when baby boomers retire. Rev. Econ. Stat. 83 (4), 589–595. Anderson, D., Botman, D.P.J., Hunt, B., 2014. Is Japan’s population aging deflationary? IMF Working Paper No. 14/139. Arnott, R.D., Chaves, D.B., 2012. Demographic changes, financial markets, and the economy. Financ. Anal. J. 68 (1), 23–46. Bakshi, G.S., Chen, Z., 1994. Baby boom, population aging, and capital markets. J. Bus. 67 (2), 165–202. Baltagi, B.H., 1995. Econometric Analysis of Panel Data. John Wiley & Sons, Chichester, New York. Bedford, R., Callister, P., Didham, R., 2010. Missing men and unacknowledged women: explaining gender disparities in New Zealand’s prime adult age groups, 1986–2006. N. Z. Popul. Rev. 36, 1–26. Beltran, D.O., Kretchmer, M., Marquez, J., Thomas, C.P., 2013. Foreign holdings of U.S. treasuries and U.S. treasury yields. J. Int. Money Financ. 1 (32), 1120–1143. Blanchard, O.J., Furceri, D., Pescatori, A., 2014. A prolonged period of low interest rates? In: Teulings, C., Baldwin, R. (Eds.), Secular Stagnation: Facts, Causes and Cures. A VoxEU.org eBook. Blundell-Wignall, A., Roulet, C., 2014. Problems in the international financial system. OECD J. Financ. Mark. Trends 1, 99–121. Bongaarts, J., 2009. Human population growth and the demographic transition. Philos. Trans. Biol. Sci. 364 (1532), 2985–2990. Borio, C., 2012. The financial cycle and macroeconomics: what have we learnt? BIS Working Papers No. 395. Borio, C., 2014. Monetary policy and financial stability: what role in prevention and recovery? BIS Working Papers No. 440. Bosworth, B.P., 2014. Interest rates and economic growth: are they related? Center for Retirement Research at Boston College Working Paper No. 2014-8. Bouis, R., Rawdanowicz, Ł., Renne, J.-P., Watanabe, S., Christensen, A.K., 2013. The effectiveness of monetary policy since the onset of the financial crisis, OECD Economics Department Working Papers No. 1081. OECD Publishing. Bouis, R., Kei-Ichiro, I., Rawdanowicz, L., Christensen, A.K.,2014. Factors behind the decline in real long-term government bond yields. In: OECD Economics Department Working Papers No. 1167. OECD Publishing. Brooks, R., 2002. Asset-market effects of the baby boom and social-security reform. Am. Econ. Rev. 92 (2), 402–406. Brunetti, M., Torricelli, C., 2008. Demographics and asset returns: does the dynamics of population ageing matter? Ann. Financ. 6, 193–219. Chamon, M., Prasad, E., 2008. Why are saving rates of urban households in China rising? NBER Working Paper Series No. 14546. Claude, B.E., Harvey, C.R., Viskanta, T.E., 1997. Demographics and international investments. Financ. Anal. J. 53 (4), 14–28. D’Amico, S., English, W., Lopez-Salido, D., Nelson, E.,2012. The Federal Reserve’s large-scale asset purchase programs: rationale and effects. In: Finance and Economics Discussion Series Working Paper No. 2012-85. Federal Reserve Board, Washington, DC. Davis, E.P., Li, C., 2003. Demographics and financial asset prices in the major industrial economies, Brunel University Working Paper, http://www.ephilipdavis.com/ DeLong, J.B., Summers, L.H., 2012. Fiscal Policy in a Depressed Economy, Brookings Papers on Economic Activity., pp. 233–297. EIU ViewsWire, 2010. Ellis, A., 2009. UK resident population by country of birth, Population Trends 135. Goyal, A., 2004. Demographics, stock markets flows, and stock returns. J. Financ. Quant. Anal. 39 (1), 115–142. Greene, W.H., 2000. Econometric Analysis. Prentice Hall International, London. Hofmann, B., Zhu, F., 2013. Central bank asset purchases and inflation expectations. BIS Q. Rev. (March), 23–35. Homer, S., Sylla, R., 2005. A History of Interest Rates, 4th ed. John Wiley & Sons, Hoboken, NJ. International Monetary Fund, 2013. World Economic Outlook: Transitions and Tensions. IMF, Washington, DC, October. International Monetary Fund, 2014. World Economic Outlook: Recovery Strengthens, Remains Uneven. IMF, Washington, DC, April. Jones, E., 2003. The European Miracle: Environments, Economics and Geopolitics in the History of Europe and Asia, 3rd ed. Cambridge University Press, New York. Joyce, M., Lasaosa, A., Stevens, I., Tong, M., 2011. The financial market impact of quantitative easing in the United Kingdom. Int. J. Cent. Bank. 3 (7), 113–161. Kaminska, I., Zinna, G., 2014. Official demand for U.S. debt: implications for U.S. real interest rates, IMF Working Papers No. 66. Krishnamurthy, A., Vissing-Jorgensen, A., 2012. The aggregate demand for treasury debt. J. Polit. Econ. 120 (2), 233–267. Lacomba, J.A., Lagos, F., 2006. Population aging and legal retirement age. J. Popul. Econ. 19, 507–519. Li, M., Qiu, J., 2013. Speculative capital inflows, adaptive expectations, and the optimal renminbi appreciation policy. China Econ. Rev. 25, 117–138. Mankiw, N.G., Weil, D.N., 1989. The baby boom, the baby bust, and the housing market. Reg. Sci. Urban Econ. 19, 235–258. Matheson, J., 2010. The UK populations: how does it compare? Population Trends 142. McKinsey Global Institute, 2010. Farewell to Cheap Capital? The Implications of Long-Term Shifts in Global Investment and Saving. McKinsey and Co., Washington, DC. Meaning, J., Zhu, F., 2011. The impact of recent central bank asset purchase programmes. BIS Q. Rev. (December), 73–83. Modigliani, F., Brumberg, R., 1954. Utility analysis and the consumption function: an interpretation of cross-section data. In: Kurihara, K. (Ed.), Post-Keynesian Economics. Rutgers University Press, New Brunswick, NJ. Pindyck, R.S., Rubinfeld, D.L., 1998. Econometric Models and Economic Forecasts. Irwin McGraw-Hill, Boston. Poterba, J.M., 2001. Demographic structure and asset returns. Rev. Econ. Stat. 83 (4), 565–584. Poterba, J.M., 2004. The impact of population aging on financial markets, NBER Working Paper 10851. Siegel, J., 1998. Stocks for the Long Run, 2nd ed. McGraw Hill, New York. Suits, D., 1984. Dummy variables: mechanics vs. interpretation. Rev. Econ. Stat. 66 (1), 177–180.

A. Sˇ evi´c, D. Brawn / J. of Multi. Fin. Manag. 30 (2015) 36–61

61

Swanson, E.T., Williams, J.C., 2014. Measuring the effect of the zero lower bound on yields and exchange rates in the U.K. and Germany. J. Int. Econ. 92 (Suppl. 1), S2–S21. Thayer, A.B., 2009. Considering population and war: a critical and neglected aspect of conflict studies. Philos. Trans. Biol. Stud. 364 (1532), 3081–3092. Turner, A., 2009. Population ageing: what should we worry about? Philos. Trans. Biol. Sci. 364 (1532), 3009–3021. Yoo, P.S., 1994. Age distribution and returns of financial assets Federal Reserve Bank of St. Louis Working Paper 94-002B. Warnock, F.E., Warnock, V.C., 2009. International capital flows and U.S. interest rates. Journal of International Money and Finance 28 (6), 903–919.