Population Aging and the Possibility of a Middle-Income Trap in Asia

ABSTRACT We present three conditions for a demography-driven middle-income trap and show that many economies in East, South, and Southeast Asia satisfy all of them. The conditions involve (1) the support ratio of workers to consumers has an impact on economic growth, (2) economic development accompanies more investment in human capital and lower fertility due to the quantity–quality trade-off, and (3) a current low level of fertility corresponds to very low support ratios for keeping up with frontier economies in the long run. Panel analysis for 178 countries shows that (1) and (2) are satisfied for Asia with higher elasticity than others. As for (3), we set up a dynamic model for simulations, showing that approximately two-third of Asia’s developing countries have an unsustainable level of support ratios, implying possibility of a middle-income trap due to future demographic headwinds.

before getting rich? If being old means hardships in economic growth, the answer could be that Asia is getting into a "middle-income trap" due to various demographic factors.
In this article, we focus on the possibility of a demography-driven middle-income trap in Asian countries. As for a middle-income trap, using a transition matrix analysis, Han and Wei (2016) argued that a middle-income country that grows at the average or median rate of the middle-income group will clearly and surely become a high-income country, reaching and surpassing the living standard in today's United States or France. Im and Rosenblatt (2013) also found no support for the notion of a middle-income trap. However, from the viewpoint of demographic dynamics, this only reflects the data for the past several decades, when demographic factors have played a positive role for Asian countries. Indeed, Han and Wei (2016) pointed out that fast-growing countries have a relatively large working-age population, which we see as only temporary.
The possibility of a demography-driven middle-income trap has been proposed by Ha and Lee (2016) where the role of a support ratio (the number of workers to the number of consumers) in the process of economic convergence has been analyzed using the panel data of more than 100 countries. They argued that the level of the support ratio is closely related to the speed of convergence or growth rate of GDP per capita, and the levels of economic development and average human capital have a negative relationship with fertility. As an economy grows, fertility declines and ultimately leads to low support ratios and lower speed of convergence, which implies the possibility of a middle-income trap.
In this research, we first present more extended analyses of Ha and Lee (2016) for different regions and income groups for Asia. We then present our analyses of the link between Asian fertility and support ratios. This additional work is important because fertility affects support ratios in different ways over time. In the short term, a decline in fertility raises the support ratio, but in the long term, the motion of the support ratio depends on the current level of the support ratio and other demographic variables. We model such dynamics to analyze the long-term effects of fertility for each country to see whether a middle-income trap is likely to develop.
Moreover, we discuss what information the National Transfer Accounts (NTAs) data provide about the middle-income trap. We demonstrate that the NTAs' data show a more vivid picture for intergenerational resource allocation and the dynamic relationship between fertility and support ratios.

Related Studies
According to Han and Wei (2016), who rejected the notion of an unconditional middle-income trap, the term "middle-income trap" was introduced by Gill and Kharas (2007). However, similar concepts such as a non-convergence trap were developed in the 1990s to model various sources of the trap. These theories are all based on the premise that developing countries could face stagnation if they fail to "change gears" properly at the point where old growth regimes reach their limit during the convergence process.
In this type of model, success of the past becomes a hurdle for the future at a certain point. Acemoglu, Aghion, and Zilibotti (2006) argued that a growth strategy should change from investmentbased to innovation-based one to avoid this trap. Basu and Weil (1998) and Aghion, Howitt, and Mayer-Foulkes (2005) emphasized that technologies and institutions, respectively, should change appropriately following the country's level of economic development.
Unlike previous works on this trap, this article focuses on demographic factors as important sources of the middle-income trap, at least temporarily. The role of demographic factors and demographic dividends have been significant in the development of Asia. Young (1995), Williamson (1998), Bloom, Canning, andMalaney (2009), Mason and Kinugasa (2008), Deaton and Paxson (1997), Lee, Mason, and Miller (2003), Mason and Lee (2007), Lee and Mason (2010), Mason, Lee, and Lee (2010), and Lee, Mason, and Park (2012) all have stated that demographic transition accounts for significant parts of Asia's economic growth. This naturally leads to the possibility that this growth could suffer from adverse effects once demographic dividends turn negative, which implies a stagnating force due to demographics. Ha and Lee (2016) provided a framework for the analysis of a demography-driven middle-income trap. They focused on the key variable that effectively summarizes demographic structure: the support ratio, or the ratio of (effective) workers to (effective) consumers. They argued that in the early stages of development, Asia's fast demographic transition raised support ratios that created a huge demographic dividend, thereby encouraging Asia's quick convergence and economic development. However, they projected that in later stages, low fertility, which economic development brought about through the quantity-quality trade-off, eventually leads to falling support ratios and a negative demographic dividend. 1 If the support ratio falls to a level too low for catching up with the frontier, then the economy can get into a non-convergence trap.
Based on Ha and Lee (2016), we develop this concept by providing and analyzing the three conditions for the trap and the dynamic aspect of support ratios both theoretically and empirically. We also try to evaluate the sustainability of the high support ratio of Asia with some comparisons between regions and income groups.

Three Conditions for a Demography-Driven Middle-Income Trap
We suggest three conditions for a demography-driven middle-income trap. First, demographic factors should matter for economic growth and convergence. In this article, we look at the relationship between support ratio and growth rates. We see that support ratios can affect saving and investment rates and eventually growth rates of GDP per capita, which determines the speed of convergence to the frontier. From the viewpoint of the NTAs, a high support ratio means that more resources can be allocated within productive generations through saving and investment.
Second, there should be a negative relationship between the level of development and fertility. As economy develops, it needs more human capital for learning and adopting frontier technologies, which leads to higher returns on education and more investment in education. More education accompanies lower fertility following the mechanism of the quantity-quality trade-off.
Third, fertility should be low enough to lead to a low support ratio. This is more complicated than other conditions and creates more complex issues because low fertility raises the support ratio in the short term, yet it could pull it down in the long run. Moreover, according to Lee and Mason (2014), low fertility does not always lead to long-term low suboptimal support ratios because there could be cases in which the maximum steady-state support ratios correspond to subreplacement fertility (less than the fertility level of 2.1). We delve into this issue later in this article. Ha and Lee (2016) analyzed the first and second conditions for Asia and the world, and this article shows the three conditions for Asian Development Bank (ADB) economies and others in Asian regions. In analyzing the third condition, we model the dynamics of the support ratio and use the concept of steady state support ratios. The structure of the trap and the conditions can be summarized, as shown in Figure 2.

Testing for Conditions of a Demography-Driven Middle-Income Trap
To test the conditions for middle-income trap, we conducted annual panel analyses using the Penn World Table (PWT) 9.0 2 and World Development Indicators for 178 countries, among which 30 were ADB regional members with sufficient data for the years 1970-2014. 3 We present the results for both random-effects models and fixed-effects models because the Hausman test results are ambiguous for many cases. 4 We use parsimonious models to describe the relationship between major variables.

Condition (1): Support Ratio Matters for Economic Growth
The first condition states that support ratios and economic growth should have positive relationships. In the panel analyses in this section, the support ratio is defined as the population share of age 15-64. The NTAs' definition-ratio of effective workers to effective consumers-would be more precise, but time series and country coverage are quite limited for NTAs. Table 1 shows the results. The two variables are indeed closely related. The coefficients indicate this relationship is much stronger for ADB member countries than for others, implying that Asian countries have used demographic dividend more intensively for economic growth. This was possible because of the rapid decline in fertility during the early stages of demographic transition.
We also looked at the relationship between support ratio and saving/investment rates (see Tables S1 and S2, available online). All of the results show that saving/investment rates are more closely tied to the support ratio in Asia, just as with the case of growth rates. This is consistent with the literature that has shown a large part of Asia's growth has been the realization of the first and second demographic dividends. The first dividend, which is the direct effect of increased labor supply among the population, accounts for more than 10% of growth in many countries (Ha and Lee 2016). However, it seems that the second dividend-increased saving and capital accumulation-is never smaller than the first dividend in Asia.
Moreover, we looked at the difference between regions in Asia. As for the effect of support ratio on saving rate, there are no big differences among regions in Asia, but the coefficients are a little high in Central Asia and a little low in Southeast Asia (Tables S3 and S4, available online). Meanwhile, regarding the coefficient of support ratio on investment rates, Central Asia shows no significance, but other regions show significant correlations, with South and Southeast Asia having high elasticity with respect to support ratios (Tables S5 and S6, available online).
These analyses show that, in general, Asia has exploited demographic dividends more intensely than other regions, and high and rising support ratios have been important in catching up with the frontier. Therefore, the first condition of a middle-income trap is satisfied in Asia; support ratio matters more for growth in Asia.

Condition (2): Development Causing a Low Fertility-Quantity-Quality Trade-Off
The second condition is about the relationship between the level of development and fertility. This issue has been discussed widely within the framework of the quantity-quality trade-off, in which increased investment in human capital accompanies lower fertility following households' optimization of decision-making. As an economy develops and moves closer to the frontier, more human capital is needed to learn and adopt the frontier technologies, and the return to human capital could rise. This makes households put more weight on the quality rather than the quantity of children. Table 2 shows that fertility is highly sensitive to the level of economic development in Asia when measured by either GDP per capita or the level of average human capital as in PWT 9. If we look at the differences across regions in Asia, human capital accumulation has greater adverse effects on fertility, especially in East and Southeast Asia (Tables S7 and S8). In Central Asia fertility is not so sensitive to economic development. Table 3 shows that for all income groups, human capital accumulation has a strong negative effect on fertility. 5 Fertility seems to be more sensitive to economic development in Asia than on other continents. There may be various reasons. Asia's education is much costlier because of factors such as the high cost of learning the English language and tough competition due to education's role in screening workers. Moreover, the massive increase in human capital needed during the process of economic catch-up has inevitably brought about a rapid decline in fertility in Asia. ADB = The 30 Asian Development Bank regional member countries as in footnote 3, GDP = gross domestic product. Note: All independent variables are logged. Numbers in parenthesis are standard errors; *, **, and *** refer to the significance levels of 10%, 5%, and 1%, respectively. Source(s): Authors' calculation.
POPULATION AGING AND ASIAN MIDDLE-INCOME TRAP 1229 Table 2. Gross domestic product per capita, level of human capital, and fertility.

Random Effect Fixed Effect Random Effect Fixed Effect Random Effect Fixed Effect Random Effect Fixed Effect
Gross domestic product per capita

Condition (3): Low Fertility Leading to a Support Ratio Too Low to Catch-Up
According to Lee and Mason (2014), low fertility does not necessarily mean a low support ratio in the long term. To put it simply, there can be long-term optimal fertility that can maximize the support ratio. One can easily conjecture that extremely high long-term fertility will lead to a low support ratio because of the presence of too many children for each household, whereas extremely low long-term fertility will cause a low support ratio due to the ever-shrinking number of workers compared with retired generations. Therefore, the optimal fertility that maximizes the support ratio is somewhere in the middle, which does not necessarily coincide with the replacement level of fertility. This means that, depending on economic conditions, the everincreasing or ever-decreasing population size could be optimal to keep the support ratio high. Hence, it is possible that relatively high or low fertility can last a long time. An average household can certainly choose fertility that is too low or too high to maximize its within-household support ratio. Considering these issues, the condition that low fertility leads to a too-low support ratio for catch-up is more complicated than it seems. Fertility affects the support ratio with a time lag and complex dynamics. To analyze this problem, we set up a model based on Ha and Lee (2015) and extended it to show the relationship between fertility and a steady state support ratio. Then we put actual numbers in the model to compare the steady-state support ratio with the current support ratio, which enabled us to evaluate sustainability of a high support ratio. This analysis is important because if high support ratios due to low fertility in recent years correspond to a very low support ratio in the long term, then the high support ratio cannot be sustained. In other words, if fertility decline raises the support ratio in the short run but lowers it in the long run, then the speed of convergence or catch-up will erode and lead to a trap. Therefore, we need to evaluate the sustainability of the current support ratio for each country to see whether the country will face a decline in its support ratio. In doing so, we use Ha and Lee (2015) model to obtain the dynamics of a support ratio, and then make up the formula for actual applications. In the model, the key state variables are the current support ratio (l) and the steady-state support ratio (l ss ), the latter of which is the steadystate value of the support ratio implied by current birth-death rate combinations. The dynamics are as follows.
The support ratio l is where L is the working age (age 15-64) population and N is the total population. The law of motion of support ratio is simply  and the motion of the working age population L is where z is labor force inflow rate, which is the flow of population into the labor force as a ratio of population, m is retirement rate of L, and d is the mortality rate of L. Then the motion of N is Putting (3) and (4) into (2), we have: Letting (5) equal to zero, the steady state value of l is obtained This is consistent with the findings of Lee and Mason (2014), where l is constant, but population growth rate is either positive or negative. We apply this concept to Asian countries, and thus we introduce several simplifying assumptions: (1) labor force inflow is the same as the number of people born 15 years ago; (2) L survives until 64 if life expectancy is greater than 64; and (3) labor force outflow is the number of people born 64 years ago if life expectancy is greater than 64. Otherwise, labor force outflow is the mortality rate multiplied by the population. Then, we have: We define diffsr ¼ l À l ss , which converges to 0 in the long run. Then we have the following proposition. Proposition: As the absolute value of diffsr becomes larger, the speed and magnitude of the support ratio adjustment become larger during the process of convergence to the steady state, as long as the current values of birth-death rates persist.
This occurs because diffsr measures the amount of transitional factors and thus the need for future adjustment. Figure 3 shows the relationship between the current support ratio and the steady-state support ratio. What is interesting here is that the two variables have an inverse U relationship, implying that too low or too high support ratios are not sustainable. Also, it seems that high support ratios that exceed 0.65 cannot be sustained if current birth-death rates do not change drastically. In the case of Asia, many middle-income countries have already reached 0.7 or above. Figure S4, available online, presents the trends in the two support ratios for selected countries where the gaps are quite large. The ROK and the PRC have exceptionally large gaps, predicting large long-term adjustments of support ratios. Japan and Germany seem to show a little bit of adjustment, whereas the United States and India show relatively smaller gaps and the possibility of less drastic adjustments. However, except for India, the size of the gap (more than 10%) is substantial in every aspect and requires close attention when predicting future growth paths.
Likewise, Figure 4 shows the trends in the size of needs for adjustment for Asian countries. In a number of countries the gap is quite small, but countries like Armenia, Azerbaijan, Brunei Darussalam, Bhutan, the PRC, Hong Kong, the ROK, Sri Lanka, the Maldives, Malaysia,    Singapore, Thailand, and Vietnam have a support ratio gap greater than 10 percentage points. These are 13 countries out of 30. Moreover, Bangladesh, Georgia, Indonesia, Mongolia, and Uzbekistan have a gap higher than 7 percentage points. This means about two thirds of Asian economies are facing serious headwinds of a declining support ratio in the near future.

Evaluating the Possibility of a Middle-Income Trap in Asia
As we have seen in this section, most countries in East, South, and Southeast Asia meet the three conditions of the middle-income trap driven by demographic factors. Specifically, the middle-income countries with low steady-state support ratios are quite likely to end up in the trap if they do not get into the high-income group quickly or fail to raise fertility. It seems that the PRC and Thailand are facing this possibility ( Figure 5). Japan, the ROK, Hong Kong, and Singapore are also facing a similar trap although they are classified in the high-income category.

Effects of Extending Retirement Age
Thus far, steady-state support ratios have been entirely determined by birth-death dynamics with a fixed retirement age of 65. However, one can consider extending this retirement age to make the size of the working age population larger and the number of retirees smaller. This can be analyzed by modifying Equation (7) as follows: Here, the only difference is that retirement age is now 70 instead of 65, and Equation (8) gives us the new steady-state support ratios for given birth-death rates. In most cases, this change raises the steadystate support ratios substantially.  state support ratios increase by 3.45 percentage points in the ROK, 4.38 percentage points in the PRC, and 4.45 percentage points in Thailand. This narrows the gap between actual and steady-state support ratios by 10-18% for these countries. Therefore, an extended retirement age could certainly relieve the aging problems in Asia. 6 However, one should keep in mind that this has only a partial effect and cannot change the whole direction of aging dynamics as long as retirement itself exists. 7 First, NTAs show a more detailed picture for a quantity-quality trade-off. NTAs show that intergenerational resource allocation affects fertility and the future course of economic development. One can see a stronger and clearer relationship between human capital investment and fertility through the lens of NTAs. In particular, human capital spending and consumption by age group 0-24 have strong negative correlations with fertility. Human capital spending and transfer to young generation data from NTAs have much clearer negative relationships with fertility than do PWT 9 data ( Figure S5, available online). Table 5 also shows this and that the NTAs human capital investment data have a stronger correlation with fertility.
Second, NTAs show the transitional path of support ratios more accurately. The adverse effect of declining support ratios can last for decades, implying the possibility of a middle-income trap for many years. Future changes in support ratios imply that transitional effects can last for decades and lead to a temporary but long-term trap. Figure 6 shows that for 2015-2035 some countries have positive projected growth rates of their support ratio, whereas others have negative rates. 8 However, for 2015-2055, more countries fall into the area of negative demographic dividends. The calculation of the model shows that in the long run with the current birth-death rates, the countries, except the Lao People's Democratic Republic, face demographic headwinds.

Conclusion
We have shown that there are three conditions for a demography-driven middle-income trap, and our main finding is that Asian countries generally meet those conditions, especially in East and Southeast Asia. Thus far, low fertility has caused a rapid increase in support ratios and more demographic dividends in Asia, which have led to a faster catch-up with the frontier. But for many countries in this region, this positive effect cannot be sustained for many decades to come.
Declining support ratios are a strong headwind for growth, leading to at least a temporary or transitional middle-income trap. We would like to emphasize that the transition may take decades, not years.
Policy implications for our analysis are straightforward. First, a fertility decline should not be that fast and shocking to the economy. Second, fertility should come back to more sustainable levels in many Asian countries. Transfers to younger generations due to human capital investment are needed for continuing catch-up, but this may be too costly for some countries as the "quantity-quality tradeoff" leads to a fertility that is too low and thus erodes the basis of catch-up itself. For example, it is well known that in East Asian countries, households spend substantial amounts of income on English language education. This may be to prepare their kids for the catch-up development, but at the same time, this makes the private cost of having kids too high. Therefore, when economies develop, the private cost of human capital investment relative to income should become significantly cheaper to avoid any future trap. In the meantime, however, carefully extending retirement age, such as to 70, will certainly help alleviate these problems to some degree. generally considered to be developing economies. Other ADB developing members that have no sufficient data are not included. In this paper, ADB member countries refers to the 30 economies.
6. However, this could temporarily lead to much tougher competition in the job market and a higher youth unemployment rate.
7. This means that if there is no retirement in the first place, age structure would not matter that much. 8. The prediction of support ratios in the NTAs data sheet for 2016 uses the United Nations fertility forecast and the current intergenerational resource allocation structure.