The economic effect of changes to China’s one-child policy

(1)

Lisanne Karsten 5941326

lisanne.karsten@student.uva.nl

Supervisor: Dr. K.B.T. Thio

Date: 15 September 2014

Word count: 14.314

The economic effect of changes

to China’s one-child policy

Master Thesis – Revised Version

In November 2013 China has relaxed the regulations of its one-child policy. This is expected to lead to a sudden increase in the fertility rate, which raises the population growth rate and the youth dependency ratio. In the long run, it lowers the old-age dependency ratio relative to what it would be without the relaxation. This research provides an indication of how future demographic values will influence future GDP per capita growth in China. It consists of two stages. In the first stage the effect of demographic values on GDP per capita growth is empirically estimated. In the second stage, predictions for future GDP per capita growth are created by multiplying the estimated coefficients with the population projections up to the year 2050. These predictions are made for three scenarios, using three different population projections. For the base scenario a population projection is used from before the relaxation of November 2013. The relaxed one-child policy scenario uses the population projection under current one-child policy regulations. And a hypothetical eliminated one-child policy scenario, where a population projection for an immediate elimination of the one-child policy is used. The analyses indicate that China has most to gain from future demographics under the base scenario, thus under one-child policy regulations from before November 2013. It appears that the depressing effects from an increased youth dependency ratio on GDP per capita growth are more significant than the stimulating effects from a lower old-age dependency ratio. As a result the base scenario is the most beneficial scenario for future GDP per capita growth in China.

(2)

- 1 -

1. Introduction

In 1979 the Chinese government introduced the one-child policy in China. The one-child policy limits couples to having one child and was installed in order to control rapid population growth. China’s National Bureau of Statistics (CNBS) estimated in 2011 that the one-child policy has prevented approximately 400 million births and decreased the average fertility rate, the number of births per woman, from 2.9 to 1.6 (The Economist, 2011). These figures make the one-child policy successful in containing the size of the Chinese population. However, studies also show that there are negative consequences to the one-child policy. One of the critiques on the policy is that it has increased the rate of aging in China, possibly leading to high economic costs for future generations (Feng 2005, Shobert 2013, Hamori et al. 2013). In 1990 the share of the elderly population, aged 65 and older, was still 5.8 percent. In 2050 it is expected to be at 23.9 percent (The World Bank, 2014). Figure 1 illustrates China’s age-profiles for 1990 and 2050, it shows that the elderly population can put a large burden on the working-age population (ages 15 to 64) in the future.

Source: Worldbank – World Development Indicators

In November 2013 China has relaxed the rules to the one-child policy. The relaxation allows more couples to have a second child, which will increase population growth as well as the youth-dependency ratio (The Economist, 2013). The youth youth-dependency ratio is the ratio between the amount of people aged 0 to 14 over the amount of people aged 15 to 64. By allowing this demographic group to grow, the age dependency ratio is expected to fall in the future. The old-age dependency ratio refers to the number of people old-aged 65 and over divided by the number of people aged 15 to 64. This research analyses how the relaxation of the one-child policy affects Chinese economic growth.

80 60 40 20 0 20 40 60 80 0-4 5-9 10-14 15-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75-7980+

Figure 1b: Expected Age Structure 2050

Female Male 80 60 40 20 0 20 40 60 80 0-4 5-9 10-14 15-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75-7980+

Figure 1a: Age Structure 1990

(4)

- 3 - Previous literature has examined the effect of these demographic variables on economic growth in multiple ways. Studies show that population growth enhances economic growth, however this positive relationship has proven to be caused by growth of the working-age population (Bloom and Williamson 1997, Bloom Canning and Malanay 2000, Dao 2012). The youth-dependency ratio has presented a constant negative relationship with economic growth, whereas the old-age dependency ratio has generally given undetermined results (Bloom and Williamson 1997, Wei and Hao 2010, Dao 2012).

This research adds to the existing literature by analyzing how alterations to the one-child policy affect GDP per capita growth in China during the period from 2020 to 2050. The analysis consists of two stages. First, the effects of the sizes as well as the growth rates of the demographic variables in Asia on GDP per capita growth will be estimated. The existing literature presents good approximations of these values. However, the studies do not estimate the effect of both size- and growth rate values on economic growth. For this reason, a regression analysis is constructed in stage one of this research. The second stage presents predictions for future GDP per capita growth in China under three different scenarios. The first scenario will be the base scenario, it uses projections for Chinese population growth for the one-child policy as it was before the relaxation of November 2013. The second scenario concerns population projections after the relaxation in November 2013. And the third scenario is a hypothetical situation where an immediate complete elimination of the one-child policy is projected.

The findings of this research suggest that Chinese demographics contribute most to future GDP per capita growth under the base scenario. It implies that the Chinese economy will benefit most from future demographics when the one-child policy remains the way it was before the relaxation of November 2013. This result is generated by the fact that the costs from an increase in the youth-dependency ratio outweigh any benefits from a relaxed one-child policy.

The next chapter describes the history and consequences of the one-child policy. Chapter 3 presents a theoretic framework for economic growth regressions and gives an overview of previous empirical studies that have estimated the effect of demographic variables on GDP per capita growth. Chapter 4 performs the regression analysis after which the estimated coefficients will be used to generate predictions about demographic contributions to Chinese GDP per capita growth. And finally, chapter 5 concludes the research and provides suggestions for further research.

(5)

- 4 -

2. The Chinese One-Child Policy

To be able to analyse the implications of the one-child policy it is relevant to understand the origins of the one-child policy. Containing population growth has been part of China’s governmental policies for many decades. Since the year 1956 China has initiated four population-control campaigns, stimulating the use of contraception, abortion, late marriages and small families. The belief that governmental intervention in population size was required, was encouraged by the outcome of the first national census in 1953. The census revealed that China had a population of almost 600 million people and was growing at a pace of 2 percent per year. The government advocated that population control was needed in order to guide the country towards economic prosperity. China’s rapid population growth hampered the goals of the so called ‘four modernizations’ referring to industrial-, agricultural-, scientific- and defence improvement (Chen and Kols, 1982).

The first two campaigns aimed at controlling population growth, initiated in 1956 and 1962, proved to be unsuccessful in lowering the fertility rate. Figure 2 shows that the fertility rate was little affected by the campaigns and remained relatively stable at roughly six children per woman. The years between 1958 to 1961 form an exception in Chinese population data, this is due to the widespread famine China was exposed to. The famine caused extremely high death- and very low birth rates, resulting in low and negative population growth rates (Chen and Kols, 1982). The third campaign, the later-longer-fewer campaign of 1971, had a greater effect on the fertility rate and was able to reduce it to 2.98 births per woman in 1978. This campaign stimulated couples to marry later, have longer intervals between first and subsequent children, and have fewer children in general.

Source: Worldbank – World Development Indicators

6,3 6,161 5,2 2.982 1,52 1,663 0 1 2 3 4 5 6 7 1950 1952 1954 1956 1958 1960 1962 1964 1966 1968 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 Figure 2: Fertility Rate (Births per Woman)

Introduction of the one-child policy Introduction of the later-longer-fewer campaign First campaign Second campaign

(6)

- 5 - The final drop in the fertility rate was brought about by the initiation of the one-child policy in 1979. This policy lowered the fertility rate even further to a steady level ranging between 1.52 and 1.67 births per woman.

The one-child policy was introduced in 1979 under Deng Xiaoping. The foundations for the one-child policy were based on a research conducted by Jian Song, a Chinese mathematics- and technology engineer. He estimated that China would reach optimal welfare if it was able to reduce its population size to range between 650 and 700 million by 2050 (Bongaarts and Greenhalgh, 1985). Jian Song was inspired by Olsder and Strijbos (1970) who approached population size as a mathematical optimization problem and presented a solution to how a population can be “steered” towards a desired size via an optimal birth rate. Based on Song’s estimates of China’s optimal population size the Deng Xiaoping regime created a new legal structure to incentivize single-child families.

The one-child policy is enforced by offering great incentives to single-child families and highly discouraging the birth of a second or third child. If a couple commits to having a single child they can apply for a One-Child Certificate. This certificate grants the family access to a number of benefits. An urban family receives additional monthly payments of 5 to 8 percent of the parents’ wages during the first 14 years of the child’s life. Rural couples can receive up to one month extra wages each year until the child is 14 years old. In terms of housing, single child families get preferential treatment when registering for public housing as well as the opportunity to live in a house suitable for two-child families. Furthermore, couples with a one-child certificate receive two weeks additional maternity leave, higher pension payments, priority access to school programs and higher priority in job assignments (Chen and Kols, 1982).

When a family does not meet the terms of the one-child policy they will be subject to a number of penalties. The family is obliged to pay back the benefits it had received when applying for the One-Chid Certificate. The second child cannot be registered for a cooperative medical care program, nor does it receive any priorities in school or job assignments. When a couple gives birth to a third child the parents will see their average monthly wages drop by 10 percent, the family will have to live in a two-child living space and the parents are not eligible for any subsidies in times of economic hardship. A number of provinces also exclude the parents from promotions, wage increases or bonuses for three subsequent years (Chen and Kols, 1982).

The one-child policy was originally designed as a drastic measure that would allow a very small portion of the couples to have a second child if circumstances justified the family to have two children. These specific conditions were thought to be relevant for approximately 5 percent of the urban population and 10 percent of the rural population. However, public resistance caused the government to be more lenient concerning the rules of the one-child policy which led towards a

(7)

- 6 - more permissive program (Bongaarts and Greenhalgh, 1985). The permissiveness is visible in the number of exceptions to the one-child policy. Most importantly, the one-child policy is only strictly applied to Han Chinese, an East Asian ethnicity that covers approximately 92 percent of China’s population. Other ethnicities in China are considered minorities and are subject to rules that are less strict than the one-child policy, they are usually allowed to have two children (Liao, 2013). Nonetheless, some Han Chinese families are also allowed to have a second child. For example, a second child is permitted if the first child has died, is disabled, diseased or adopted due to a presumed sterilization of the mother. Also, a family may have a second child if the first child comes from a previous marriage (Chen and Kols, 1982). There are also some regions where it is allowed to have a second child if the first child is a girl (Bongaarts and Greenhalgh, 1985). Another set of exceptions to the one-child policy refers to the occupational characteristics of the parents. For example, families may have a second child if the father is a deep sea fisherman or if the father works in underground mining for more than 5 years (Liao, 2013). The final exception to the one-child policy worth mentioning is the regulation that allows couples to have second child if both parents come from a single child family. This rule has been changed in November 2013, since then couples are allowed to have a second child if one of the parents comes from a single child family (The Economist, 2013).

The rules and exceptions described above, exclusive of the relaxation of November 2013, make that approximately two thirds of China’s total population is restricted to a single child (Feng, 2005). The one-child policy is believed to be the cause for a number of social and economic problems. The most commonly cited consequences are a rapidly aging population, imbalances in the sex-ratio and increased female infant mortality rates (Feng 2005, Hesketh et al. 2005, Liao 2012). This research is limited to the economic consequences of changes the one-child policy, therefore only the effects of population demographics will be taken into the analysis.

In 2004, research concerning population aging in China has stated that the old-age dependency ratio has increased from 5 percent to 7.5 percent between 1982 and 2004 and is expected to rise to 15 percent in 2025 (Hesketh, et al., 2005). This ratio is still lower than the ratios in most industrialized countries. For example, the current old-age dependency ratio is roughly 27.5 percent in Europe and 21 percent in the United States (Worldbank, 2014). However, China is believed to have an inadequate pension system which is why the Chinese population is expected to face a situation where 70 percent of its elderly population is dependent on the support of their offspring. With the one-child policy in place this means that a working-age couple will be responsible for the care of four parents and one child, the so-called “4:2:1 phenomenon” (Hesketh et al., 2005). China

faces a unique aging problem since the country is expected to get old before it gets rich. Shobert

(8)

- 7 - extracting from China’s resources. The approaching growth in the elderly dependent population, together with China’s limited health and pension funds have caused several researchers to believe that the one-child policy is unsustainable.

The demographic dynamics in China have put the government in a balancing position with population growth control on one side, and reducing the future weight from the elderly population on the other side. During the past 20 years a number of researchers have suggested alternatives to the one-child policy that would relieve the population from some of the burdens it will be facing in the future. For example, allowing couples to have two children would still limit the fertility rate but also reduce the old-age dependency ratio and normalize the skewed sex-ratio (Hesketh et al., 2005). In response to the suggested alternatives the government made an official statement in 2002 announcing that it would not fundamentally change the one-child policy but that instead a number of regulatory aspects would be relaxed. The new regulations allowed couples freedom of choice in which contraception to use, and couples were no longer obliged to get permission in advance to have their first child (Hesketh et al., 2005). The relaxation in November 2013 was the first fundamental change to the one-child policy since its implementation.

(9)

- 8 -

3. Literature Review

The alteration of the one-child policy in November 2013 is believed to affect population growth, the youth dependency ratio and the old-age dependency ratio. The possible effects of demographic changes on economic growth have been researched extensively. This section provides a brief overview of the existing literature. It starts with a theoretic approach to the effect of population demographics on economic growth. Next, a brief overview of previous empirical studies is presented. And the final section formulates an expectation of the effect the relaxed one-child policy will have on Chinese economic growth.

3.1 Population demographics and economic growth

One of the earliest theories that relates population variables to economic growth dates back to 1798. Thomas Malthus presented a theory where economic growth is hampered by population growth due to disproportionate growth rates of food and population (Seidl and Tisdell, 1999). Malthus based his theory on the assumption that population growth increases at an exponential pace, whereas food supply increases linearly. The inequality between these two growth rates would ultimately cause a country’s population to outgrow food supply with catastrophic consequences as a result. Seidl and Tisdell (1999) argue that Malthus’ expectations of the future turned out to be unrealistic since population growth behaves differently than he assumed, and the growth rate of food production has grown exponentially rather than linearly.

3.1.1 Solow growth model

Recent theories concerning population demographics are generally based on the Solow growth model. The Solow growth model (Solow, 1956) attempts to relate long-run economic growth to four factors: output (𝑌), capital (𝐾), labour (𝐿) and knowledge (𝐴). By multiplying knowledge (𝐴) and labour (𝐿) effective labour is found, this means that changes in productivity, due to technological change, appear in the model via increased effectiveness of labour (Romer, 2012). The relationship between output, capital and effective labour is presented in the production function

𝑌𝑡 = 𝐹(𝐾𝑡, 𝐴𝑡𝐿𝑡). (3.1)

For analytic purposes the four factors of the model are expressed per unit of effective labour, thus the production function above is divided by 𝐴𝑡𝐿𝑡. This results in an output per unit of effective labour

function

𝑦𝑡 = 𝑓(𝑘𝑡), (3.2)

where 𝑦𝑡 = 𝑌𝑡

𝐴𝑡𝐿𝑡 is the output per unit of effective labour and 𝑘𝑡 = 𝐾𝑡

𝐴𝑡𝐿𝑡 is the amount of capital per

(10)

- 9 - Figure 3: Solow Growth Model

0, 𝑓′(𝑘) > 0, 𝑓′′_{(𝑘) < 0 and finally 𝑓}′_{(𝑘) ≈ 0 if 𝑘 is large. As a result, when 𝑘 is small, 𝑓}′_{(𝑘) is large}

and the change in actual capital is large. Since 𝑓′′(𝑘) < 0 this effect reduces as 𝑘 grows larger. This means that when there is relatively little capital per unit of effective labour, adding one amount of capital will increase output by a large quantity. When the amount of capital per unit of effective labour rises the increase in output grows less. The 𝑘 variable is essential for analysing the model, in particular the rate of change in 𝑘 is important, denoted by 𝑘̇. If 𝑘̇ = 0 the rate of change of capital per unit of effective labour is zero, the economy is in a steady state and on a balanced growth path; each variable in the model grows at a constant rate (Romer, 2012). In the Solow growth model 𝑘̇ is a function of existing capital 𝑘𝑡 which takes the following form

𝑘̇(𝑘𝑡) = 𝑠𝑓(𝑘𝑡) − (𝑛 + 𝑔 + 𝛿)𝑘𝑡 . (3.3)

The first term in the equation, 𝑠𝑓(𝑘𝑡), represents actual investment per unit of effective labour,

where 𝑠 signifies the investment rate: the fraction of output that is invested. The second term of the equation, (𝑛 + 𝑔 + 𝛿)𝑘𝑡, embodies the break-even investment. In this term 𝑛 signifies the growth

rate of labour, 𝑔 is the growth rate of knowledge and 𝛿 is the rate of depreciation of existing capital. As a result, the equation of 𝑘̇ states that the rate of change of capital per unit of effective labour is equal to actual investments minus the break-even investment. The amount of capital per unit of effective labour can only be kept constant if actual investment equals break-even investment (𝑘𝑡̇ =

0). Figure 3 shows this relationship graphically.

At point 𝑘∗ the economy is in its steady state. The economy grows at a constant rate. The assumption that 𝑘 is subject to the Inada Conditions means that the economy will always converge back to its steady state, to 𝑘∗. Shocks to either the savings rate, the growth rate of labour, the

𝑘∗ In vest m ent o r O u tpu t p er u n it o f effec tiv e lab o u r 𝑘 𝑓(𝑘)

Output per unit of effective labour

(𝑛 + 𝑔 + 𝛿)𝑘

Break-even investment 𝑠𝑓(𝑘)

(11)

- 10 - growth rate of knowledge or the depreciation rate (𝑠, 𝑛, 𝑔, 𝛿) can cause the steady state to shift to another level which will induce a temporary change in the growth rate of output per worker and a permanent change in the actual output per worker.

The speed at which an economy converges to its steady state level is called the rate of convergence and determines the convergence equation. The convergence equation is often used for theoretical and empirical analysis of economic growth studies, and will also be used for this research. To find the convergence equation a first-order Taylor approximation is performed on the production function 𝑦𝑡 = 𝑓(𝑘𝑡) (eq. 3.2) around the point where 𝑦𝑡 = 𝑦∗. In order to do so, the production

function should be rewritten as

𝑦𝑡 = 𝑓(𝑔(𝑦)), (3.4)

where 𝑔(𝑦) = 𝑘𝑡. Taking the first-order Taylor approximation of equation 3.4 around the point

where 𝑦𝑡 = 𝑦∗ leads to the following equation

𝑦𝑡 ≃ 𝑦∗+ 𝑒−𝜆𝑡[𝑦0− 𝑦∗], (3.5)

which can be rewritten into

𝑦𝑡− 𝑦∗≃ 𝑒−𝜆𝑡[𝑦0− 𝑦∗]. (3.6)

Equation 3.6 implies that 𝑦𝑡 approaches 𝑦∗ at a rate proportionally equal to the gap between initial

output (𝑦0) and steady state output (𝑦∗). This equation is then converted into a growth equation,

which gives us the convergence equation

𝑔𝑦= −𝜆(𝑦0− 𝑦∗). (3.7)

The term −𝜆 signifies the speed of convergence1_{(Romer, 2012). Equation 3.7 thus states that the} growth rate of output per unit of effective labour is a function of the gap between initial output and steady state output, multiplied by the speed of convergence (−𝜆). The speed of convergence appears as a negative term in the equation because if 𝑦0 is slightly lower than 𝑦∗, output per unit of effective

labour is expected to rise, hence 𝑔𝑦 is positive. Similarly, if 𝑦0 is slightly higher than 𝑦∗, output per

unit of effective labour is expected to fall.

3.1.2 Solow growth model and population growth theories

As previously mentioned, the Solow growth model is often the basis for literature concerning economic growth theory. Population growth theories are not an exception to this rule and can essentially be divided into three categories: the so called “population pessimists” who believe in a negative relationship between population growth and economic growth, “population optimists” who argue that population growth stimulates economic growth, and “population neutralists” who believe

1_{In Romer (2012) it is explained that 𝜆 = −𝑓}′_(𝑘∗_{) = −[𝑠𝑓}′_(𝑘∗_{) − (𝑛 + 𝑔 + 𝛿)] = (𝑛 + 𝑔 + 𝛿) − 𝑠𝑓}′_(𝑘∗_{) =} (𝑛 + 𝑔 + 𝛿) −(𝑛+𝑔+𝛿)𝑘∗𝑓′(𝑘∗)

(12)

- 11 - that population growth has no significant effect on economic growth (Bloom and Williamson, 1998). The theories generally assume that the economy is initially in its steady state and that the changes to the population growth rate cause the steady state to shift either up or down.

The theory proposed by Coale and Hoover (1958), supporters of the pessimistic view, can be approached via the Solow growth model. The authors explain that population growth impedes economic growth due to insufficient investments. A rise in the population growth rate means that there are more people using the existing capital. If a country is unable raise its investments accordingly, the amount of capital will be insufficient to maintain the same level of productivity and output per unit of effective labour will fall. In terms of figure 3 an increase in the population growth rate implies a rise in 𝑛, which makes the (𝑛 + 𝑔 + 𝛿)𝑘 line steeper. When actual investments do not catch up, due to and inadequate savings rate (𝑠), the steady state level 𝑘∗ will shift to the left and output per unit of effective labour falls (Coale and Hoover, 1958). In line with this theory, the instalment of the one-child policy in 1979 enhanced economic growth, and the relaxation of November 2013 is expected to harm economic growth.

Population optimists, on the other hand, advocate that population growth enhances economic growth because of increased productivity (Bloom and Williamson, 1998). This changes the Solow growth model from figure 3 to a model with endogenous growth. In figure 3 the growth rate of knowledge (𝑔) is treated as an exogenous force. The theories proposed by population optimists use an expanded version of the Solow growth model. In the expanded model knowledge is endogenously determined, it is influenced by the population growth rate. This appears in the model via an equation relating the growth rate of knowledge (𝑔) to the amount of labour (𝐿) and capital (𝐾) devoted to research and development. As a result, the model implies that output per unit of effective labour is an increasing function of the population growth rate (Romer, 2012).

Theoretically population growth can stimulate productivity via two mechanisms: technological progress and increased human capital. According to Kremer (1993) population growth enhances technological progress, because a higher population growth rate increases the number of people able to invent something. The author points out that the costs of inventing new technologies stands apart from the number of people using it, thus a bigger population implies an increased number of inventions without increasing the costs of those inventions. Becker, Glaeser and Murphy (1999) explain how population growth increases human capital. Population growth creates a higher population density, which allows for improved labour allocation and lower coordination costs. As a result the returns on investments in education increase which raises the amount of investments in human capital and increases the productivity of the population. According to these theories, the relaxation of the one-child policy will enhance Chinese economic growth via increased productivity.

(13)

- 12 - Population neutralists base their belief on empirical research rather than on economic models. The general claim is that there is insufficient support for the hypotheses of either a negative or a positive relationship between population growth and economic growth (Bloom and Williamson, 1998). Empirical findings concerning these claims will be discussed in section 3.2.

3.1.3 Solow growth model and population composition

At present there is a general acceptance among researchers that the composition of a population influences economic development. Bloom and Williamson (1998) describe how changes in the age structure of a population can affect economic growth. They write that a rise in the fertility rate, raising the number of children born at a certain point in time, will first negatively affect economic growth, due to need for spreading output over a larger group of people, but will stimulate economic growth after approximately two decades when the new-borns enter the labour force (Bloom and Williamson, 1998).

In terms of the Solow growth model, these movements can be explained via the savings rate. A person’s economic needs and contributions change over a lifetime. The working-age population tends to save, because they consume less than their wages. The youth- and elderly populations dis-save since they primarily consume and earn very little. As a result, the savings rate (𝑠) changes when the relative size of one of the population groups changes. It increases when the relative size of the working-age population grows and decreases when the relative size of the youth- or old-age population grows (Bloom, Canning and Fink 2010). In figure 3, it can be seen that a rise in the savings rate 𝑠 shifts the actual investment line (𝑠𝑓(𝑘)) upwards. This shifts the steady state level outwards and results in a higher output per unit of effective labour and increased output per worker. Intuitively, a higher savings rate causes more investments which raises the amount of capital, and more capital raises output per unit of effective labour (Romer, 2012).

A growing old-age dependency ratio is also referred to as aging and is believed to negatively affect economic growth (Bloom, Canning and Fink 2010). However, Bloom, Canning and Finlay (2010) claim that the overall effect of aging is negligible since the negative effects from an increased old-age dependency ratio are offset by households’ behavioural responses. They explain that aging is driven by two factors; a decline in the fertility rate and an increase in life expectancy. These two factors each provoke behavioural responses that stimulate economic activity. First, the decline in the fertility rate allows a higher female participation rate in the labour market as there is less need for them to stay home. Second, the increase in life expectancy changes savings behaviour since people expect they will have to consume for more periods, making them save accordingly. And finally, a combination of the two factors provides people with higher incentives to invest more in education causing a higher productivity rate. These responses to population aging stimulate the economy such

(14)

- 13 - that the downward pressures of aging are nearly or completely offset (Bloom, Canning and Finlay 2010).

3.2 Previous empirical estimations

Theories about the effect of demographic variables on economic growth have been empirically tested in a number of studies. These studies report differing, and in some cases opposing, results. This section provides an overview of previous empirical findings.

3.2.1 From the Solow growth model to empirical regressions

Nearly all of the empirical studies attempting to capture the relationship between population dynamics and economic growth use the Solow growth model as the foundation for their regression analysis. The key equation for these empirical studies is the convergence equation derived in section 3.1.1. Recall that the convergence equation is given by 𝑔𝑦= −𝜆(𝑦0− 𝑦∗) (eq. 3.7).

The steady state output level of an economy (𝑦∗) is not a given value in economic data, which is why economic growth regression generally portray it as a function of 𝑿, such that 𝑦∗= 𝜷𝑗𝑿. The

term 𝑿 denotes a vector containing a number of variables that are believed to influence a country’s steady state, such as human capital, government expenditures, the savings rate, and others. Implementing this in the convergence equation (eq. 3.7) yields a generalized form of the regression that is often used in empirical literature concerning economic growth

𝑔𝑦= 𝛼𝑦0+ 𝜷𝑗𝑿 + 𝜀. (3.8)

The term 𝑔𝑦 functions as the dependent variable and indicates a measurement of output per capita.

The 𝑦0 variable represents output per capita in the base year and captures the convergence effect.

The term 𝜷𝑗𝑿 determines the steady state. And finally 𝜀 is a stochastic term which captures omitted

influences.

The objective of the papers discussed in this section is to estimate the effects of population dynamics on economic growth. These authors suggest that vector 𝑿 in equation 3.8 contains a number of variables that signify demographic variables such as population size and population growth. In this case the growth regression in equation 3.8 can be rewritten as follows:

𝑔𝑦= 𝛼𝑦0+ 𝜸𝑖𝒏 + 𝜷𝑗𝑿 + 𝜀, (3.9)

where 𝒏 is a vector representing demographic explanatory variables and 𝑿 contains the remaining determinants of the steady state.

(15)

- 14 - 3.2.2 Results from previous studies

Table 1 shows the results from previous empirical studies which have estimated a regression similar to the one given in equation 3.9. The papers discussed here have run multiple regressions, however, table 1 is limited to display only the outcomes of the demographic variables from the most relevant regressions.

The coefficients estimating the effect of population size on economic growth are estimated to be significantly positive, significantly negative and insignificant. For example, Bosworth and Collins (2003) report that a one percent increase in population size corresponds to a 0.0011 percent increase in output per capita growth. On the other hand, Hamori et al. (2013) present a significant negative result of -1.80, implying that a one percent increase in log population size causes GDP per capita to fall by 1.29 percent. A number of researches estimate the effect of population growth rather than population size. The results show that the population growth rate has a significant negative effect on economic development. Only regression (A) in column 2 reports a significant positive relationship between the population growth rate and economic growth, it seems that this result is generated by the positive influences of the growth rate of the working-age population.

With respect to the total dependency ratio, the ratio where the youth- and the old-age dependency ratio are grouped together, results suggest a negative relationship to economic growth. Both the size and the growth rate of the total dependency ratio report a negative effect, but only the effect of the size of the total dependency ratio is significant. The growth rate gives an insignificant negative coefficient.

When the total dependency ratio is split into the youth and the old-age dependency ratio we see differing results. The youth dependency ratio shows a significant negative relationship with respect to economic growth, whereas the old-age dependency ratio seems to have both significant negative and insignificant positive effects on economic growth. The interaction terms between the dependency ratios and the population growth rate, visible in column 7 of table 1, show to have significantly positive effects on economic development. This implies that in combination with population growth, both the youth- and the old-age dependency ratio apply positive pressures on economic growth (Dao, 2012).

(16)

- 15 -

Note: The values in parentheses are the corresponding t-statistics. The stars indicate the significance level: (***) significant at the 1 percent level, (**) significant at the 5 percent level and (*) significant at the 10 percent level.

Table 1: Regression results from previous studies

Column 1 Column 2 Column3 Column 4 Column 6 Column 7

Bosworth and Collins 2003

Bloom and Williamson 1997 Bloom,

Canning and Malanay 2000

Wei and Hao 2010 Hamori,

Kinugasa and Yao 2013

Dao 2012

Regression A Regression B Regression C Regression A Regression B

Type of Regression Panel data regression

Panel data regression Panel data

regression

Panel data regression Time-series regression

Cross-sectional

Dependent Variable Growth in output per worker

Growth rate of real GDP per capita Growth rate of real GDP per capita

Growth rate of GDP per capita GDP per capita

Growth rate of GDP per capita

Log Population Size 0.29*** (4.80) -0.362 (0.85) -0.121 (-0.28) -1.80*** (4.60)

Population Growth Rate 0.56***

(3.50) -1.03*** (-2.58) -0.980** (2.13) -13.732*** (-4.157)

Log Total Dependency Ratio

-4.227** (-2.27)

Total Dependency Ratio Growth -2.747 (-0.19) Working-age Population Growth Rate 1.46*** (4.29) 0.81*** (4.15) 1.265*** (3.34)

Log Youth Dependency Size

-2.724** (-2.12)

Youth Dependency Ratio Growth Rate

-0.37*** (3.97)

-8.859 (-0.80)

Log Old-age Dependency Size 1.605 (0.94) -1.604*** (-3.853) Old-age Dependency Ratio Growth Rate

0.08 (1.00)

11.996 (1.42)

Log Ratio of Woking-age Population to Total Population in Base Year

0.167* (1.76)

Population Growth & Youth Dependency Ratio

0.074*** (3.350)

Population Growth & Old-age Dependency Ratio

0.757*** (2.934)

(17)

- 16 - These results point out that existing empirical estimations of the effects of demographic variables on economic development are mixed. Heady and Hodge (2009) propose that the inconclusiveness of these studies arises from the fact that both negative and positive effects of population growth are present. They claim that the results differ from one another due to differences in the time frames considered, the types of countries included in the data, the definitions of the relevant variables and the statistical methods employed. In their own research they apply a meta-regression analysis to a data sample consisting of 471 results available in the existing literature. This type of analysis allows for new insights into relationships or patterns that might be present in the results. The strongest result of their analysis showed that a possible negative effect of population growth on economic growth has strengthened after 1980, suggesting that the negative pressures of a growing population have gotten more relevant over time. They also find that the downward pressures of a growing youth dependency ratio on economic growth are larger than the positive effects of a growing labour force, and that the size of the positive effects of a growing labour force are dependent on the institutional quality of a country which may cause different results on a country specific level.

3.3 Expectations of one-child policy alterations for Chinese economic growth

Despite the inconsistency of previous studies, an expectation can be formed about the outcome of the question at hand; to what extent will the relaxation of the one-child policy hamper or enhance economic growth relative to the situation without changes to the one-child policy? The relaxation or elimination of the one-child policy is expected to cause a higher population growth rate as well as a higher youth dependency ratio relative to the base scenario. In the long run, the relaxation will cause a lower old-age dependency ratio. Based on the Solow growth model, the increase in the population growth rate and the population size are expected to negatively influence Chinese economic growth. Similarly for the youth-dependency ratio, its increase means a higher burden on the working-age population applying downward pressures on economic growth. Thus, in the short run, the relaxed policy is expected to cause lower economic growth than the base scenario. However, when the increased amount of new-borns have grown old enough to join the economically active population, the economy is likely to be stimulated. Therefore after 2 decades, Chinese economic growth may be higher under the relaxed one-child policy than under the base scenario. But this does not have to the case. Considering the inconsistent results from the empirical literature in combination with the theories proposed by Bloom, Canning and Finlay (2010), the negative effects from an aging population may be offset by behavioural responses causing economic stimulus rather than economic harm.

(18)

- 17 -

4. Empirical analysis

The empirical studies described in the previous chapter show that there are many estimations of the relationship between demographic factors and economic growth. To find the effects of changes to the one-child policy there is need for a regression that estimates the coefficients of the sizes as well as the growth rates of the demographic groups. The regressions in the existing literature do not provide a clear estimation of these effects. Therefore this research performs its own regression analysis. First, the regression will be explained. Next, the data will be introduced after which the results from the regression analysis are presented. These results will then be used to create predictions for demographic contributions to future GDP per capita growth. The final section of this chapter performs a robustness-check of the results by running the same regression whilst including a binary variable specific for China.

4.1 The regression

The foundation for the analysis is parallel to the regressions described in chapter 3, and most similar to the study executed by Wei and Hao (2010). Wei and Hao (2010) use the convergence equation from the Solow growth model. They manipulate this equation such that it is able to incorporate the effect of changes to population composition. They refer to their model as “the extended convergence framework”.

4.1.1 The extended convergence framework

The regular convergence equation derived in chapter 3 states that the growth rate of output per unit of effective labour depends on the gap between the initial level of output and steady state output multiplied by the speed of convergence, 𝑔𝑦= −𝜆(𝑦0− 𝑦∗) (eq. 3.7). This equation gives the

growth rate of output in terms of effective labour, it assumes stability of demographic composition. The extended convergence framework from Wei and Hao (2010) relaxes this assumption. It allows for changes in the composition of the population via the dependency ratio. They derive a function determining the growth rate of output per capita rather than per unit of effective labour.

The authors start by explaining that output per capita (𝑌

𝑃) consists of output per unit of

effective labour (𝑌

𝐴𝐿) multiplied by the share of effective labour in the total population ( 𝐴𝐿 𝑃) 𝑌_𝑃= (𝑌 𝐴𝐿) ( 𝐴𝐿 𝑃). (4.1)

The share of effective labour in the total population can also be written in terms of the total dependency ratio (𝐷), such that 𝐷 =𝑃−𝐴𝐿_𝐴𝐿 and 𝐴𝐿_𝑃 = 1

1+𝐷. This changes equation 4.1 to 𝑌 𝑃= ( 𝑌 𝐴𝐿) ( 1 1+𝐷). (4.2)

(19)

- 18 - The authors derive a natural logarithm equation where 𝑦̃𝑡 =

𝑌

𝑃 , output per capita, and 𝑦𝑡= 𝑌 𝐴𝐿,

output per unit of effective labour,

𝑙𝑛(𝑦̃𝑡) = 𝑙𝑛(𝑦𝑡) − 𝑙𝑛 (1 + 𝐷)𝑡. (4.3)

This equation implies that the natural logarithm of output per capita (𝑙𝑛 (𝑦̃)) consists of the natural logarithm of output per unit of effective labour (𝑙𝑛 (𝑦)) minus the natural logarithm of the total dependency ratio (𝑙𝑛 (1 + 𝐷)). When equation 4.3 is differentiated with respect to time an output per capita growth rate equation arises:

𝑔𝑦̃= 𝑔𝑦− 𝑔1+𝐷. (4.4)

Equation 4.4 shows that the growth rate of output per capita (𝑔𝑦̃) equals the growth rate of output

per unit of effective labour (𝑔𝑦) minus the growth rate of the total dependency ratio (𝑔1+𝐷). The

growth rate of output per unit of effective labour (𝑔𝑦) is determined by the convergence equation.

Wei and Hao (2010) use the same equation, however they express it in logarithmic terms

𝑔𝑦= −𝜆(𝑙𝑛(𝑦0) − 𝑙𝑛(𝑦∗)). (4.5)

Initial output (𝑙𝑛(𝑦0)) can be derived from equation 4.3.

𝑙𝑛(𝑦0) = 𝑙𝑛(𝑦̃0) + 𝑙𝑛(1 + 𝐷)0. (4.6)

And the steady-state output level (𝑙𝑛(𝑦∗)) is determined by a number of variables depicted by the vector (𝑿), resulting in

𝑙𝑛(𝑦∗_{) = 𝜷}

𝑗𝑿. (4.7)

Finally, if equation 4.4 is rewritten using equations 4.5, 4.6 and 4.7, the extended convergence equation is obtained:

𝑔𝑦 ̃ = 𝜆(𝜷𝑗𝑿 − 𝑙𝑛(1 + 𝐷)0− 𝑙𝑛(𝑦̃0)) − 𝑔1+𝐷 (4.8)

The regression I estimate is based on equation 4.8. The term 𝑔𝑦̃ is measured by GDP per

capita growth, the vector 𝑿 is split into six demographic variables plus six control variables grouped together as 𝜶𝑗𝑪𝑖,𝑡−1, and the term 𝑙𝑛(𝑦̃0) is represented in the regression as the logarithm of GDP

per capita in the previous period. This results in the following regression:

𝐺𝐷𝑃 𝑝𝑒𝑟 𝑐𝑎𝑝𝑖𝑡𝑎 𝑔𝑟𝑜𝑤𝑡ℎ𝑡 = 𝛼1𝑙𝑜𝑔(𝑃𝑂𝑃𝑖,𝑡−1) + 𝛼2𝑔𝑃𝑂𝑃+ 𝛼3𝑙𝑜𝑔(𝑌𝐷𝑅𝑖,𝑡−1) +

𝛼4 𝑔𝑌𝐷𝑅+ 𝛼5𝑙𝑜𝑔(𝑂𝐷𝑅𝑖,𝑡−1) + 𝛼6𝑔𝑂𝐷𝑅+𝜶𝑗𝑪𝑖,𝑡−1− 𝜆 𝑙𝑜𝑔(𝑌𝑖,𝑡−1) + 𝜀𝑖𝑡 (4.9)

Equation 4.9 deviates from Wei and Hao (2010) by including the population growth rate as an explanatory variable. This is relevant since the relaxation of the one-child policy is expected to increase the population growth rate, which in turn is expected to negatively affect GDP per capita growth. Moreover, Wei and Hao (2010) use data from 31 Chinese regions over a 15 year period. This research uses data from 14 Asian countries over a 41 year period. And finally, the control variables used by Wei and Hao (2010) differ from the ones employed in this research. Table 2 provides a description of the variables in regression 4.9.

(20)

- 19 - Table 2: Regression variables

1. 𝐺𝐷𝑃 𝑝𝑒𝑟 𝑐𝑎𝑝𝑖𝑡𝑎 𝑔𝑟𝑜𝑤𝑡ℎ𝑡 The dependent variable, which is the GDP per capita growth rate

2. 𝑙𝑜𝑔(𝑃𝑂𝑃𝑖,𝑡−1) The natural logarithm of population size

3. 𝑔𝑃𝑂𝑃 The growth rate of population size

4. 𝑙𝑜𝑔(𝑌𝐷𝑅𝑖,𝑡−1) The natural logarithm of the youth dependency ratio

5. 𝑔𝑌𝐷𝑅 The growth rate of the youth dependency ratio

6. 𝑙𝑜𝑔 (𝑂𝐷𝑅𝑖,𝑡−1) The natural logarithm of the old-age dependency ratio

7. 𝑔𝑂𝐷𝑅 The growth rate of the old-age dependency ratio

8. 𝐶𝑖,𝑡−1 The steady state control variables. These consist of human capital,

government consumption as a percentage of GDP, the level of openness of a country, the change in the capital stock, a corruption index and the coast to land ratio

9. 𝑙𝑜𝑔(𝑌𝑖,𝑡−1) The natural logarithm of GDP per capita of the year before. This is

included to capture the income convergence effect explained in the Solow growth model

4.1.2 Expectations of the coefficient estimates

According to the Solow growth model the relationship between population size (𝑙𝑜𝑔(𝑃𝑂𝑃𝑖,𝑡−1)) and GDP per capita growth, as well as between population growth (𝑔𝑃𝑂𝑃) and GDP

per capita growth is expected to be negative. The Solow growth model predicts that a rise in the population growth rate results in a lower steady state output level.

The coefficients estimating the relationship between the youth-dependency ratio (𝑙𝑜𝑔(𝑌𝐷𝑅𝑖,𝑡−1) and 𝑔𝑌𝐷𝑅) and GDP per capita growth are also expected to be negative. A rise in the

youth-dependency ratio lowers the savings rate of the economy, which will eventually lead to a lower steady state level.

The estimates for the effect of the size of the old-age dependency ratio (𝑙𝑜𝑔 (𝑂𝐷𝑅𝑖,𝑡−1))

and the growth rate of the old-age dependency ratio (𝑔𝑂𝐷𝑅) on GDP per capita growth are expected

to be insignificant or perhaps even positive. Bloom, Canning and Finlay (2010) have explained that an aging population can rationally be expected to alter their behaviour in a way that may stimulate the economy rather than harm the economy. They refer to behavioural changes such as increased labour participation, increased life savings and increased human capital investments. Due to China’s limited pension funds it is imaginable that the Chinese population follows these patterns in order to secure themselves with economic stability during their retirement.

Lastly, the coefficient depicting the relationship between the logarithm of GDP per capita in the previous year (𝑙𝑜𝑔(𝑌𝑖,𝑡−1)) and GDP per capita growth in the current year is expected to be

negative. When GDP per capita is low relative to its steady state level, it is expected to grow in order to converge to its steady state. The larger the distance to the steady state, the higher the growth rate is expected to be (Romer, 2012).

(21)

- 20 - 4.2 Data description

The data consists of annual panel data from 14 East and Southeast Asian countries. The definition of East and Southeast Asia is based on a classification made by Bloom and Williamson (1998), it includes: China, Cambodia, Hong Kong, Indonesia, Japan, The Republic of Korea, Laos, Macao, Malaysia, Philippines, Singapore, Taiwan, Thailand and Vietnam. These countries are included in the regression because they are considered to have been part of ‘The East Asian Miracle’. The countries have experienced comparable economic growth surges, and improvements in living environments. Moreover, this region is believed to have gone through similar demographic transitions since 1965 (Bloom and Williamson, 1998). The regression will be run over a period from 1970 to 2011. Most of the data were retrieved from the Penn World Table version 7.1 with the exception of the dependency ratios and the corruption index, these were available via the World Development Index (World Bank) and the TI Corruption Index respectively. In appendix 1, 2 and 3 a detailed description of the data, the summary statistics and the correlations table can be found.

Using panel data for this regression minimizes the small sample bias and allows us to control for country specific effects. A country fixed effects regression controls for the omitted variables that vary from one country to the next but remain constant over time. The regression is estimated with a time lag in order to eliminate possible endogeneity between variables. It is assumed that the one-period lagged values of the explanatory- and control variables are uncorrelated with the error term of the present, thus making the variables exogenous (Wei and Hao, 2010).Furthermore,Bloom and Williamson (1998) examined whether the feedback effect of GDP per capita on demographic variables was of significant influence on their regression results. They performed their economic growth regression both via Ordinary Least Squares (OLS) regression as well as via Instrumental Variable (IV) regression. Population policy indicators and information about the religious composition of the population were used as instrumental variables. The results from both regressions proved not to be significantly different from one-another, providing evidence that the regression is not significantly affected by simultaneous causality.

The population projections, which will be used to generate forecasts about future Chinese economic growth, are based on the United Nations World Population Prospects. The base projection consists of a population projection for China in case of no changes to the one-child policy. This data is readily available. The projections concerning the relaxed child policy and the eliminated one-child policy are constructed using Spectrum DemProj software. In this program alterations to one or more aspects of the forecast can be made and it calculates future demographics accordingly.

(22)

- 21 - 4.3 Regression results

Table 3 displays the results from running OLS regressions on equation 4.9. The coefficients in column (1) are the regression estimates without controlling for country fixed effects. Column (2) shows the results from running the exact same regression but with the addition of controlling for country fixed effects. Controlling for country fixed effects is relevant for this research since the data employed here consists of 14 different countries that are all characterized by varying fixed variables such as the climate of a country, whether or not it has access to a port, whether the country contains a jungle or desert, the number of languages spoken in the country etc. The fixed effects regression eliminates the influences from these omitted variables. Column (3) and (4) estimate the coefficients using bigger time-lags for the youth dependency ratio.

The results show that a positive relationship exists between population size and GDP per capita growth, and also between population growth and GDP per capita growth. This is in direct contrast to the expectations that followed from the Solow growth model. However, the coefficients are insignificant for almost all the regressions. Only the population growth rate coefficient reports a significantly positive result in the regression that does not control for fixed effects. The lack of negative results between population growth and GDP per capita growth potentially supports the theory that population growth stimulates economic growth via increased productivity, rather than harm economic growth due to scale effects.

The coefficients for the youth dependency ratio variables appear to be highly significant and negative throughout all the regressions, regardless whether one, two or three time-lags are used. This outcome supports the theories proposed earlier, implying that a high youth dependency ratio impedes economic development and has a decreasing effect on GDP per capita growth.

The old-age dependency ratio variables are positive and switch between being significant at the 5 and 10 percent level and being insignificant. This result suggests that the downward pressures on GDP per capita growth from a large old-age dependency ratio are not as substantial as implied by the economic theory. Moreover, it supports the claim made by Bloom, Canning and Finlay (2010) that an aging population performs behavioural alterations that stimulate economic growth.

These results are somewhat comparable to the results from the previous literature. First, the effects of population size on GDP per capita growth changes per regression from positive to negative, and from significant to insignificant. Second, the youth dependency ratio has reportedly generated a greater effect on GDP per capita growth than the old-age dependency ratio. And third, the relationship between the old-age dependency ratio and economic growth has been estimated to be positive, significant and insignificant.

(23)

- 22 -

Note: The values in parentheses are the corresponding t-statistics. The stars indicate the significance level: (***) significant at the 1 percent level, (**) significant at the 5 percent level and (*) significant at the 10 percent level.

It can be argued that some of the explanatory variables may act differently if they were interacting with the control variables. Column (5) in table 3 contains two interaction term estimates. Following Wei and Hao (2010) and Bloom and Williamson (1998), the population growth variable interacts with human capital and the change in the capital stock. Theoretically it can be argued that Table 3: OLS Regression Results – Dependent Variable: Real GDP per Capita Growth

Column (1) Column (2) Column (3) Column (4) Column (5)

log (population sizet-1) -0.187 (1.13) 2.288 (1.21) 1.927 (1.00) 1.978 (0.99) 1.987 (1.05) growth rate: population 0.540** (2.31) 0.154 (0.63) 0.133 (0.55) 0.121 (0.49) 0.872 (0.94)

log (youth dependency ratiot-1) -3.332*** (-3.59) -9.682*** (-5.57) -10.026*** (-5.62)

log (youth dependency ratiot-2)

-9.934*** (-5.42)

log (youth dependency ratiot-3)

-10.807*** (-5.43)

growth rate: youth dependency ratio -0.574*** (-4.99) -0.447*** (3.62) -0.514*** (-4.11) -0.581*** (-4.43) -0.346*** (-2.79)

log (old-age dependency ratiot-1) 0.755 (1.11) 2.789** (2.32) 2.202** (1.78) 2.149* (1.67) 2.729* (1.84)

growth rate: old-age dependency ratio 0.269** (2.50) 0.195* (1.68) 0.169 (1.44) 0.135 (1.11) 0.051 (0.42)

log (GDP per capitat-1) -2.293*** (-4.71) -8.101*** (-7.54) -8.489** (-7.53) -9.049*** (-7.52) -8.636*** (-7.86) human capitalt-1 1.848** (2.26) 5.484** (2.33) 7.160*** (2.94) 7.826*** (3.07) 5.897** (2.45) government consumption t-1 8.391*** (2.59) 12.988** (2.44) 13.862** (2.56) 14.463*** (2.61) 14.663*** (2.76) openness t-1 0.000 (0.02) -0.008 (1.60) -0.008 (-1.63) -0.009* (-1.71) -0.008 (-1.47)

capital stock change 54.636*** (10.21) 64.187*** (9.40) 66.873*** (9.45) 69.083*** (9.28) 97.742*** (8.52) corruption index 0.096 (0.83) -0.827*** (-2.8) -0.823 (-2.76) -.808*** (-2.65) -0.7558*** (-2.59) coast-to-land ratio 0.002*** (2.77) population growth_ human capital -0.077 (-0.15) population growth_ capital stock change

-18.369*** (-3.77) constant 25.777*** (3.32) 50.221 (1.62) 57.905* (1.81) 63.685* (1.90) 59.120** (1.85)

country fixed effects No Yes Yes Yes Yes

observations 574 574 574 574 574

(24)

- 23 - the greater the level of human capital the more positive or less negative the effect of population growth may be on GDP per capita growth. However, the results in column 5 show that the interaction term between population growth and human capital shows no significant relationship. For the interaction between population growth and the change in the capital stock, it is to be expected that a higher population growth has a more negative effect on GDP per capita growth if the capital stock remains constant. The Solow growth model suggests that an increased population growth rate lowers GDP per capita if the investment rate does not change accordingly. The coefficient estimate in column (5) shows a significant negative relationship, supporting the Solow growth model theory. The coefficients for the interaction terms between population growth and the dependency ratios, similar to the regression run by Dao (2012), showed no significant results and are therefore not reported in table 3.

4.4 Consequences for future GDP per capita growth

In this stage of the analysis the regression results from table 3 will be used to generate predictions about demographic contributions to future GDP per capita growth in China. The regression from column (2) is deemed most appropriate for this purpose. This choice is based on the fact that controlling for country fixed effects is relevant to the nature of the data, and because the additional time-lags for the youth-dependency ratio are not significantly different from the one-lag coefficient. The interaction term will not be used for the forecast since the forecast is initially based on the assumption that all the control variables remain unaltered. However, it is useful to bear in mind that an increase in the population growth rate, for example due to the relaxation of the one-child policy, could have depressing effects on the GDP per capita growth rate if the capital stock change remains constant.

The results in column (2) from table 3 lead to the following equation: 𝐺𝐷𝑃 𝑝𝑒𝑟 𝑐𝑎𝑝𝑖𝑡𝑎 𝑔𝑟𝑜𝑤𝑡ℎ𝑡 = 50.221 + 2.288 𝑙𝑜𝑔(𝑃𝑂𝑃𝑖,𝑡−1) + 0.154𝑔𝑃𝑂𝑃

−9.682 𝑙𝑜𝑔(𝑌𝐷𝑅𝑖,𝑡−1) − 0.447𝑔𝑌𝐷𝑅+ 2.789 𝑙𝑜𝑔(𝑂𝐷𝑅𝑖,𝑡−1) + 0.195𝑔𝑂𝐷𝑅 (4.10)

−8.101 𝑙𝑜𝑔(𝑌𝑖,𝑡−1) + 5.484 ℎ𝑢𝑚𝑎𝑛 𝑐𝑎𝑝𝑖,𝑡−1+ 12.988 𝑔𝑜𝑣. 𝑐𝑜𝑛𝑠𝑖,𝑡−1

− 0.008 𝑜𝑝𝑒𝑛𝑛𝑒𝑠𝑠𝑖,𝑡−1+ 64.187 𝑐𝑎𝑝𝑖𝑡𝑎𝑙 𝑠𝑡𝑜𝑐𝑘 𝑐ℎ𝑎𝑛𝑔𝑒𝑖,𝑡− 0.827 𝑐𝑜𝑟𝑟𝑢𝑝𝑡𝑖𝑜𝑛 𝑖𝑛𝑑𝑒𝑥𝑖,𝑡

This equation is used to estimate the influence future demographics will have on GDP per capita growth. As mentioned before, the forecasts are made for three different scenarios. The first scenario concerns the ‘base forecast’, this forecast is uses population projections if China did not alter the one-child policy in November 2013. The second forecast is based on projections of the current relaxed one-child policy.It is estimated that this policy increases the fertility rate from 1.6 to 1.8 births per woman (Chaime, 2014). By altering the fertility rate in the Spectrum DemProj software

(25)

- 24 - and allowing it to interpolate the new data a new population projection is obtained. The third forecast is based on projections for a hypothetical situation where China completely eliminates the one-child policy. For this scenario an estimation from The Development Research Centre of the State

Council is used. Evidence suggests that the current voluntary fertility rate in China has decreased

significantly since 1979. A survey amongst 39,600 Chinese woman showed that 35 percent of the women wish to have only one child, and 57 percent prefers to have two children (Hesketh et al, 2005). The Research Centre has estimated the fertility rate to initially rise to a level 1.9 births per woman and thereafter slowly increase to 2.3 (Tyers and Golley, 2010). Figure 4 shows the movements for the population projections under all three scenarios. The exact values are displayed in appendix 4.

Figure 4a shows that under all three scenario’s the population size of China is expected to drop in the future, this is due to the expected negative population growth rate visible in figure 4b. China’s current fertility rate is well under the replacement level of 2.1 births per woman, which means that the population is currently facing declining population growth rates which will become below zero in the future, resulting in a decreasing population size. Figure 4c and 4d show the movements for the size and growth rate of the youth dependency ratio. It shows that the growth rate of the youth dependency ratio is expected to be negative up to 2035 after which it will be predominantly positive. This effect is a direct result from the expected drop in the working-age population. The fertility rate dropped significantly after 1975 causing an inequality between the amount of people entering and leaving the working-age population around the year 2035. This fall in the working age population causes the youth dependency ratio to grow. The same effect is visible in the growth rate of the old-age dependency ratio in figure 4f. Between the years 2025 and 2035 the large bulk of people currently part of the working-age population will reach retirement age and the old-age dependency ratio will grow. After this period, the amount of people entering the retirement population will drop causing a large fall in the old-age dependency ratio.

If the demographics of one-child policy alterations are compared to the base scenario it can be seen that under the relaxed- and the eliminated one-child policy the projection for population size and population growth rates are higher than under the base projection. Similarly, relaxations of the one-child policy cause the youth dependency ratio to be higher than the base scenario, except during the years 2030 to 2040. In this period, the youth dependency ratio growth rate under the altered policies are lower. These dynamics result from the rise in the working-age population due to the increased rate of new-born children reaching the age of 15. The differences in the old-age dependency ratio are only visible in the long run, around the year 2030, when the higher rate of new-borns enter the economically active population.

(26)

- 25 - -2 -1,5 -1 -0,5 0 0,5 1 1,5

2Figure 4d: Youth Dependency Ratio Growth

Source: United Nations World Population Prospects, available in Spectrum Demproj Software

The next step in analysing the effect of changes to the one-child policy is to examine how the demographic projections from figure 4 affect GDP per capita growth in the future. Following Bloom and Williamson (1998) the regression coefficients from equation 4.10 are multiplied with the

Base Projections

Relaxed one-child policy Projections Eliminated one-child policy projections

-0,8 -0,6 -0,4 -0,2 0 0,2 0,4

0,6 Figure 4b: Population Growth

1200 1250 1300 1350 1400 1450

1500 Figure 4a: Population Size

0 5 10 15 20 25

30 Figure 4c: Youth DependencyRatio Size

0 5 10 15 20 25 30 35 40

45Figure 4e: Old-age Dependency Ratio Size

0 1 2 3 4 5 6 mln

Figure 4f: Old-age Dependency Ratio Growth %

% %

(27)

- 26 - demographic projections for all three scenarios. The forecasts give insight into how future demographics will contribute to future GDP per capita growth. The results are displayed in table 4 and figures 5a to 5c. A more detailed presentation of the calculations can be found in appendix 5.

Table 4: Demographic contributions to future annual GDP per capita growth

2020 2025 2030 2035 2040 2045 2050 Si ze co n tr ib u tion s Base -0.060% -0.051% -0.041% -0.035% -0.032% -0.032% -0.032% Relaxed -0.067% -0.063% -0.058% -0.050% -0.047% -0.047% -0.049% Eliminated -0.068% -0.064% -0.060% -0.052% -0.050% -0.052% -0.056% Gr o wt h r ate co n tr ib u tion s Base 1.138% 1.108% 1.088% 0.770% 0.288% -0.001% 0.002% Relaxed 0.643% 0.618% 1.140% 0.870% 0.230% -0.160% -0.141% Eliminated 0.550% 0.547% 1.108% 0.871% 0.157% -0.323% -0.373% To tal co n tr ib u tion s Base 1.078% 1.058% 1.047% 0.735% 0.257% -0.032% -0.031% Relaxed 0.577% 0.556% 1.083% 0.821% 0.183% -0.208% -0.191% Eliminated 0.483% 0.482% 1.048% 0.819% 0.108% -0.374% -0.429% -0,6 -0,4 -0,2 0,0 0,2 0,4 0,6 0,8 1,0 1,2 2020 2025 2030 2035 2040 2045 2050 Figure 5c: Total contribution to future

GDP per capita growth Base projection contributions to GDP _{per capita growth}

Relaxed one-child policy projection contributions to GDP per capita growth Eliminated one-child policy projection contributions to GDP per capita growth -0,08 -0,07 -0,06 -0,05 -0,04 -0,03 -0,02 -0,01 0,00 2020 2025 2030 2035 2040 2045 2050 Figure 5a: Size contributions to future

GDP per capita growth

-0,6 -0,4 -0,2 0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4 2020 2025 2030 2035 2040 2045 2050 Figure 5b: Growth rate contributions to

future GDP per capita growth %

%