Intergenerational Transfers of the Sandwich Generation in Taiwan An - - PDF document

Intergenerational Transfers of the Sandwich Generation in Taiwan – An Application of the National Transfer Accounts Method

Kevin Yu-Ching Hsieh (yqxie1996@hotmail.com ) and An-Chi Tung (actung@econ.sinica.edu.tw) September, 2017 (Very preliminary draft: Please do not quote without permission) keywords：intergenerational transfers, intra-familial transfers, public transfers, sandwich generation, National Transfer Accounts, Taiwan

• 1. Motivation

At any point of time in any society, resources transfer among generations, through both private and public channels. The working-aged population, who are sandwiched between young children and aging parents, and are called the Sandwich Generation (SG, hereafter) in this study, are usually the net providers of resources to

• ther generations. In many countries, the financial burden of the SG has been

escalating in recent years, because elders live longer, and because college fees are higher and the unemployment rate for young people is rising (see for example, Parker and Patten, 2013). Yet a different, and probably opposite, trend has also been observed. In Japan, for example, young elders transfer resource downwards, because they enjoy decent pensions, while the working-aged population are financially weak due to prolonged economic recession (Ogawa et al, 2009). In Taiwan, population aging, education lengthening, and economic recession are all taking place, and in a faster speed than anywhere in the world. However, it remains a puzzle whether the financial burden of the SG has become heavier or lighter in Taiwan, as we do know too well the magnitude and direction intergenerational resource flows. In this study, we aim to measure intergenerational transfers in Taiwan in 1981 and 2015, and then examine how and why these transfers change over time. To construct the necessary data, we use National Transfer Accounting method (NTA

hereafter), transforming micro data into age-specific statistics, and calibrating to macro values from National Income account. Two modifications are made to help accomplishing our goal. First, data were lacking concerning who had paid social premiums and received social benefits for years 1981 to 1995. We solve the problem by trying to identify social insurants by their social economic characteristics. Second, employment-based retirement benefits are quite important in Taiwan, yet they are not measured in NTA, probably because these are not a standard system everywhere. We solve the problem by introducing a new variable in the basic NTA identity. After explaining the research motivation, the second section review the NTA method and our modifications. The third section reports the estimation results, and then discusses and compares the two years. The last section concludes the study by summarizing the findings and remarks on future studies.

• 2. Data and Methodology

2.1 Brief review of the National Transfer Accounts In a nutshell, the NTA method introduces an extra dimension of age to the National Income data. The innovative NTA data serve to open up new opportunities for re-examining many economic theories, such as life-cycle hypothesis. In addition, they make possible the re-assessment of major social policies, such as the financial sustainability of the National Pension Program in the very long run, and hence carry rich and important policy implications (e.g., Hsieh and Tung, 2016). At the center of the NTA framework is the economic lifecycle, which refers to the variation over our lifetime of our needs and abilities. The economic lifecycle expresses itself in age variation in what we consume and what we produce (United Nations, 2012). Figure 1 shows the age profiles of labor income and consumption, with age shown by the horizontal axis, and per capita value by the vertical axis. The per capita labor income YL, exhibits a bell shape, with highest average income around age 40. Note that not all individuals of each age group are employed, so the average value is determined by both how many people earn labor income and how much they earn. The per capita consumption C, is relatively flat when compared with labor income, as both babies and the elderly consume. Here C includes both private and public consumption, which will be elaborated later.

Figure 1. Per capita labor income and consumption The Lifecycle Deficit (LCD), defined as the difference between consumption and labor income, is positive for people in both very young and very old age groups, but is negative for working-age population, as shown in Figure 2. Figure 2. Per capita Lifecycle Deficit The rise and fall of LCD can be interpreted in at least two ways. First, in every society, children and the elderly consume far more than they produce, while people at working age earn more labor income than they consume. Alternatively, in the lifetime

• f every individual, there are extended periods in childhood and old-age when

consumption exceeds labor income, while there are also periods when labor income is higher than consumption.

100,000 200,000 300,000 400,000 500,000 600,000 700,000 10 20 30 40 50 60 70 80 90 NT age 100,000 200,000 300,000 400,000 500,000 600,000 700,000 10 20 30 40 50 60 70 80 90 NT age

LCD<0 LCD>0 LCD>0 YL C

The lifecycle deficits and surpluses are sustainable only because a complex system of institutions and economic mechanisms enable flows of economic resources from surplus to deficit age. In NTA, this age reallocation system is the counterpart of the economic lifecycle. Economic flows across age are mediated by three main sources of reallocation, private transfers (TF), public transfers (TG), and asset-based reallocation (AR), and the last term can be divided by public and private elements. Figure 3. Channels to finance LCD. Transfers refer to reallocations which involve no quid pro quo, thus are different from consumption or labor income. Private transfers, TF, are governed by voluntary contracts and social conventions. Examples are familial support of children and parents, bequests, and charitable contributions. Children, and in many cases the elderly, are net receivers of private transfers, and working-age adults are net

• providers. TF consists of both intra-familial (TFW) and inter-familial (TFB) transfers.

Public transfers, TG, are governed by involuntarily resource flows. The public sector reallocates resources relying on social mandates embodied in laws and regulations and implemented by local, regional and national governments. Outflows

• f TG, or government receipts from the private sector, include all taxes, direct and

indirect, and social contributions. Inflows of TG, or government outlays to the private sector, include social benefits as well as government consumption. Asset-based reallocations, AR, realize inter-age flows through inter-temporal

• exchange. They involve two kinds of flows, asset income and savings. When

individuals accumulate pension funds or personal saving during their working years

• 500,000
• 400,000
• 300,000
• 200,000
• 100,000

100,000 200,000 300,000 400,000 500,000 10 20 30 40 50 60 70 80 90 NT age

LCD

TF TG AR

and rely on asset income and/or dissaving of those assets during their retirement years, they are relying on asset-based reallocations. The above variables can be organized into the following flow identity: C(x) – YL(x) = RA(x) +TG (x) +TF (x) = YAF(x) –SF(x) +YAG(x) –SG(x) +TG(x) +TFB(x) +TFW(x) (1) where x= 0, 1, 2, …, 89, 90+ The left-hand side is the lifecycle deficit, LCD of age x. The right-hand side consists of three transfers, namely RA, TG and TF. RA can be divided into net private asset income YAF, net public asset income YAG, private saving SF and public saving

• SG. TF can be divided into inter-familial transfer TFB and intra-familial transfer

TFW. 2.2 Our modification to adapt to the Taiwan pension systems As mentioned already, two modifications are necessary in constructing the NTA

• f Taiwan. First, there are no data on social insurances, namely, Government

Employee Insurance and Labor Insurance, for years 1981 to 1992,1 which would distort the estimated TG. The inflow of TG by the elderly, for instance, would be heavily underestimated. We solve the problem by first trying to identify who the social insurants are based on their social economic characteristics. For example, soldiers can be identified by their job type. Once those who are covered by various social insurances are identified, we use the age profile per insurant of 1993 year, the first year when social insurance status is available, and adjust by employment ratio in 1981. This is not the perfect solution, but is a reasonable first approximation. Second, employment-based retirement benefits (EBR) are now singled out from RA and listed under the retired, so that the age distribution would alter, but the economy total remains unchanged. This is necessary because if these items are huge,2

1 The Farmers’ Health Insurance began in 1985. 2 In 1981, the one-time old-age benefits from GEI was 2.093 million NT, the benefits from old PSP

was 4.487 billion, and the receipt from the high-yield accounts were 3.998 million. That is, the latter two were about 4 times as high as the first item. It was about 10 times in 2015, when both the old and new PSP benefits were included. As for those covered by LI, labor retirement pension other than the one-time payment from LI started in 1987 under the Labor Standard Law, and later evolved into the Labor Pension Fund. By 2015, the amount of retirement receipt from the old and new labor pensions were about 2.5 times as high as the amount of the old-age transfers from LI.

and if they are mixed with other items in RA, the resources at the disposal of the elderly would be unduly underestimated, as asset income is attributed to household heads in NTA, and yet the retired may or may not be heads. In estimation, three items are singled out. The first two are benefits from Labor Pension (EBl) and Public Servant Pension(EBg), and were implicitly included in private savings, SF. Take EBl for example. An employee covered by LI is usually

• bliged to pay into the Labor Pension Fund (LP). Later at retirement, she received an
• ld-age benefit from LI (which is one-time payment before 2009), as well as a

retirement benefit from LPF (since 1987, and is one-time payment before 2009). The premium and benefits of LI are already included in NTA as TG, but the payment and benefits of LPF are not, as these are not transfers.3 In fact, the payment into LPF is a “forced saving”, and is part of wage income, hence the benefit can be seen as a “dis- saving” from an invisible, or notional, account, and is implicitly part of private saving, SF, which is part of RA. Similarly, when a person covered by GEI retires, she would receive an old-age benefit from GEI (basically still a one-time payment), as well as a retirement benefit from Public Servant Pension (PSP), to which she and her employer (the government) have paid during the employment period. What complicates the matter further is the fact that a retired public employee is entitled to interests from a high-yield savings account, if she deposits the one-time

• ld-age benefit from GEI, at an annual interest rate of 18%.4 Although the high-yield

saving account is related to GEI, and the over-paid interests are subsidized by the government and the Bank of Taiwan, a public owned bank, it is not a transfer income, but rather an interest income. These interest receipt, EBi would be singled from other interest income. The three new variables can be incorporated into the basic NTA equation: C(x) – YL(x) = YAF(x) –SF(x) +YAG(x) –SG(x) +TG(x) +TFB(x) +TFW(x) = YAF’(x) +EBl(x) + EBg(x) –SF’(x) +EBi(x) +YAG(x) –SG(x) +TG(x) +TFB(x) +TFW’(x) = RA’(x) +EB(x) +TG(x) +TFB(x) +TFW’(x) (2) where

3 The government not only subsidizes about 10% of LI premium (the employee pays 20%, and the

employer 70%), it also takes ultimate financial responsibility of LI. In contrast, the government does not subsidize the LPF, nor take the financial responsibility.

4 The high-yield accounts will be closed in two years, in the 2017 Pension Reform.

YAF’(x) = YAF(x) – EBl(x) + EBg(x) SF’(x) = SF(x) +EBi(x) +TFW’(x) –TFW(x) RA’(x) = YAF’(x) –SF’(x) +YAG(x) –SG(x) EB(x) = +EBl(x) + EBg(x) +EBi(x)

Sx TFW(x) = 0 =SxTFW’

To begin with, EBl+EBg are singled out from YAF, and EBi is singled out from

• SF. However, one needs to take special heed that SF has to absorb whatever changes

that takes place in intra-familial transfer TFW, as it is the balancing item in NTA. And the new TFW’ differs from TFW, because once the three EB’s are considered to the asset income of the elderly, rather than that of the household head, the age profile changes, while the household total remains as zero. A diagram below indicates clearly why the separate estimation is necessary. The two bars at the left are the channels of how consumption is financed for those aged 65 to 74 in 2015, and the two at the right are for those aged 75 and more. It is clear that the separate listing of EB reduces the relative size of RA (the striped box). Moreover, TFW (the white box) also reduces for both age groups, as the EBs are the receipts of the elderly, rather than intra-familial transfer income from the household head. Note that the differences are less significant in 1981, not drawn here, as the EBs were smaller in magnitude. 2.3 Steps of estimation We first estimate the age patterns of each variable, then smooth the age profiles, and calibrate by macro values to stay consistent with the National Income.

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 65-74, with EB 65-74, w/o EB 75+, with EB 75+, w/o EB YL EB RA TG TFB TFW

In estimating the age profiles, consumption, labor income and public transfers are estimated first. Once these are estimated, smoothed and macro adjusted, TFW can be calculated, by checking the disposal income of each member within the household, which decides the resource flow among members, while household total of TFW would be zero, by definition. The last step would then be calculating SF, which is the balancing item in the NTA framework. The following is a brief classification of actual estimation method of the

• variables. For those variables when individual micro data are available, such as

employee labor income (YLE), the individual values are recorded directly. For those variables when only household data, rather than individual data are available, such as private education consumption, CFE, we derive individual data by regression household total on individual age or other social economic trait, such as social insurance status. For variables when no micro data are available, such as public consumption, we allocate the economy total equally to all citizens. 2.4 Data Sources We construct the NTA data of Taiwan in 1981 and 2015 in this study, as 1981 is the first year when both micro and macro data are available, and 2015 is most recent year we have. A comparison of these two years not only shows the impact of time, but also what happened for various cohorts over time. Three types of data are needed to construct NTA. To begin with, the population data are provided by the National Development Council. Second, Family Income and Expenditure Survey (FIES), also compiled by the DGBAS, is our main data source. Third, National Income (NI hereafter) statistics are compiled by the Directorate General of Budget, Accounting and Statistics (DGBAS). There are other supplementary data sources from official statistics, for example, the education statistics provided by the Ministry of Education, and statistics of public insurances are from various authorities, such as the Bureau of Labor Insurance. Various official actuarial reports on different social insurances and social programs also provide necessary information. Details of the methodology are reported in our earlier article (Hsieh and Tung, 2016), a book (UN, 2013) and the website (www.ntaccounts.org).

• 3. Results (to be strengthened)

We first report the national transfer accounts of 1981 and 2015 in Section 3.1, and then compare and analyze these findings in Section 3.2. Although the accounts are

compiled by single age, we present the results in six broad age groups (0-19, 20-29, 30-49, 50-64, 65-74, 75+) to facilitate the discussion. Among the six groups, 30-49 is the prime-age group, and the average labor income will be used to standardize the results of that year for easy cross-year comparison. 3.1 1981 vs 2015 Figure 1 presents the results, 1981 on the left and 2015 on the right. For the youngest group, public transfers TG increased, and private transfers TF increased even more. For the those aged 20-29, YL reduced but all other variables increased. In particular, TG and TFW turned from negative to positive. As for the prime-age group, RA has increased. For those aged 50-64, RA, TG and EB increased, while YL and TFB decreased. For the young elders (aged 65-74), YL and TFW reduced, TG, RA and EB increased. Finally, for the old elders (aged 75+), YL reduced, all other variables increased. 1981 2015 Figure 1 Channels to finance consumption, 1981 and 2015

• 0.6
• 0.4
• 0.2

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 0-19 20-29 30-49 50-64 65-74 75-90

YL EB RA' TG TFB TFW

• 0.6
• 0.4
• 0.2

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 0-19 20-29 30-49 50-64 65-74 75-90

YL EB RA' TG TFB TFW

3.2 analysis There are many changes between these two years. One of the most striking result is that those aged 20-29 have turned from net providers of resources through public transfers and intra-familial transfers to net recipients. Also, both the young elders or the old elders receive relatively less intra-familial transfers from co-resident family members, while new public transfer inflows, asset reallocation, and employment- based retirement benefits became the major sources to finance their consumption. These changes are consistent with macro changes in the society. To begin with, the rising importance of TG to the elderly is a result of the implementation of new social programs, such as National Health Insurance in 1995 and National Pension Program by late 2008. Second, the reversal of the role of those aged 20-29, and the significant decline of YL to those aged 50-64 is consistent with what has happened in the labor market, namely the late entry and the early exit. Third, the rise of importance of RA, asset reallocation is consistent with the shrinkage of labor share in production factor income. Fourth, the rise of the importance of inter-familial transfers TFB, though still small, is most likely consistent with family nuclearization, as some resource flows are now classified as inter-familial flows when large families break into smaller units. Most of these findings are consistent with what we already know. To begin with, Tung (2011) reported that the increase in consumption level is larger for younger and

• lder groups than the working-aged population for years 1985 and 2005. The same

phenomenon is observed in our comparison of 1981 and 2015. Second, Lai and Tung (2015) reported that the elders rely more on public transfers than intra-familial

• transfers. The same is observed here.

However, as employment-based retirement benefits were not considered in previous studies, the role of EB becomes clear only in this study. In fact, the importance of EB is probably mixed improperly into TFW, RA and YL.5 With our new data, the decrease of the role of TFW would be more pronounced, and the increase of RA is still large, but less.

5 Specifically, in 1981, there was no distinction of labor status in wage income in the raw data, and no

information of social insurance premiums and benefits at all. Therefore, results of from earlier studies need modification.

• 4. Discussion and concluding remarks (to be completed)

This section contains four parts: a summary of our findings, the implications, and discussion of future works. These will be completed soon. Some of the implications: First, it is an old tradition for the elders to expect supports from their children, which is described as “raising children to support ones

• ld-age” (Lai and Tung, 2015). But our results indicate that, the elders would need to

rely more on the government (for public transfers), past voluntary savings (into RA), and past forced savings (through the payment into the EBs). This is an on-going project. Future works include a more detail decomposition of intra-familial flows by broad age groups, rather than by the entire economy. If this is done, we can show more clearly who is supporting whom, as is done in Ogawa et al (2009) for Japan. Next, we shall compare and discuss the total burdens, private and public, born by different cohorts over their life cycle, for a better comparison of intergenerational equity. This require a longer time series data, and will be done next. References 1. Hsieh, Yu-Ching and An-Chi Tung, 2016, “Taiwan’s National Pension Program: A Remedy for Rapid Population Aging? The Journal of the Economics of Ageing, 8, 52-66 (http://dx.doi.org/10.1016/j.jeoa.2016.05.001) 2. Lai, Mun Sim and An-Chi Tung, 2015, “Who Supports the Elderly? The Changing Economic Lifecycle Reallocation in Taiwan, 1985 and 2005”, The Journal of the Economic of Ageing, 5, 63-68. 3. Ogawa, Naohiro, Andrew Mason, Amonthep Chawla, Rikiya Matsukura & An- Chi Tung, 2009, “Declining Fertility and the Rising Cost of Children: What Can NTA Say about Low Fertility in Japan and Other Asian Countries?” Asian Population Studies,5(3), pp.289-307. 4. Parker, Kim, and Eileen Patten, 2013, The Sandwich Generation: Rising Financial Burdens for Middle-Aged Americans. Washington, DC: Pew Research Center, Social and Demographic Trends. 5. Tung, An-Chi, 2011, “Are Children Consuming More Over Time? An Application of the National Transfer Accounts Method”, Taiwan Economic Forecast and Policy,42(1),119-148. (In Chinese)