INTER-GENERATIONAL IMPACTS OF SINGAPORE’S BUDGETARY RESPONSES TO COVID-19: GENERATIONAL ACCOUNTING FRAMEWORK

1. Introduction

Due to the combination of low fertility rates and increasing life expectancy, Singapore’s population is aging at an unprecedented rate. In 2017, Singapore officially became an aged society, with 14.8% of its population aged 65 and above (Department of Statistics, 2020e). This demographic transition presents major challenges to Singapore’s economic growth and fiscal sustainability. Currently, around 1/3 of Singapore government’s revenue comes from income taxes (Ministry of Finance, 2021b), for which a shrinking labor force implies a shrinking tax base. In contrast, age-based expenditures such as healthcare subsidies are expected to increase, as evident from the doubling of total social spending within the last decade. Will Singapore’s current fiscal system be sustainable in the long run?

Amidst the changing demographic landscape, the COVID-19 pandemic (COVID in short) in early 2020 triggered the largest recession that the world has faced in recent years and resulted in a 5.4% contraction of Singapore’s economy (Ministry of Trade and Industry, 2021). To safeguard lives and livelihood, the government pledged a total of S$193 billion across four Budgets in 2020, which is more than double the size of previous annual Budgets (Heng, 2020a), consequently incurring an overall deficit of S$65 billion (MOF, 2021a). Since these deficits are expected to be offset by future surpluses, the question that naturally arises is: What are the impacts of COVID-19 on Singapore’s long-term fiscal sustainability and inter-generational fiscal balances?

This paper adapted the well-established methodology by Auerbach et al. (1991) to construct a set of generational accounts for Singapore residents by projecting Singapore’s future demographics and tax and benefit profiles, taking into account the unique way of defining budgetary balances in Singapore. By comparing current and future generations’ accounts, we evaluate the effect of COVID-19 on fiscal sustainability. As posited by Herrmann (2011), demography is not destiny, we examine how policy interventions could help avert the challenges of aging population while addressing intergenerational inequity.

The rest of the paper is organized as follows. Section 2 summarizes the existing literature on generational accounting and the use of the residual approach to estimate generation accounts. Section 3 covers actuarial projections of Singapore’s population and calibrations of per capita tax/benefit age profiles, government net purchases and net debt. Section 4 analyzes the pre-COVID intergenerational equity in Singapore by examining the pre-COVID generational accounts constructed and conducting sensitivity analyses. Section 5 discusses the effect of COVID-19 on the assumptions and projections. Section 6 presents the post-COVID results and relevant policy recommendations to achieve fiscal sustainability. Section 7 summarizes the findings and limitations in using GA analyses.

2. Intergeneratonal Links and Aging Population

The relationship between aging population, old age dependency and intergenerational links was brought into attention by Samuelson (1958). As articulated by Weil (2008), population aging “has economic effects whenever some economic interaction…brings together people whose participation is a function of their age” such as pension systems. Other interactions include the dynamics between the elderly’s wealth and the human capital of the young (Kotlikoff and Spivak, 1981) and bequests and caregiving (Horioka et al., 2018). Generations are thus intricately linked to one another, and differences in their consumption opportunities affect the “intergenerational equity”.

Government policies impact consumption and after-tax income of generations differently. Typically, when assessing government’s fiscal position, conventional deficit accounting will be based on receipts and expenditures within the fiscal year. Due to its omission of demographic inputs, deficit accounting cannot reflect the intergenerational redistributive effect of fiscal policies which may affect consumption opportunities of different generations¹ unevenly (Bonin, 2001). Additionally, deficit accounting figures do not elicit the long-term significance of current fiscal policies given the likely changes in demographic landscape and are thus of little value in facilitating long-term economic planning by agents from the life-cycle perspective (Bonin, 2001).

Auerbach et al. (1991) proposed an alternative tool by incorporating population cohorts in the government’s intertemporal budget constraint and dubbed it as “generational accounting” (GA). This method fills the gap in deficit accounting by deriving a set of “generational accounts”, that is, the present value of the expected payments by each generation during its remaining lifetime to the government, less its expected future receipts from the government. It thus enables the analysis of current fiscal policies’ intergenerational redistributive effects and long-term sustainability.

Within seven years of its inception, generational accounts were constructed for 23 countries around the world (Auerbach et al., 1991); more recently, there was renewed interest from economists who wished to understand the long-term implications of fiscal policies implemented during the Eurozone debt crisis (Hacıibrahimoğlu and Derin-Güre, 2015). In Singapore’s case, Cardarelli et al. (2000) evaluated if the country’s fiscal system was prepared for the demographic transition using GA. However, their analysis was based on International Monetary Fund (IMF)’s definition of budget balance as gross operating surplus. Consequently, they omitted key items such as special transfers, Net Investment Return Contributions² (NIRC) and Singapore’s use of different budgetary accounts to reflect fiscal positions across different time horizons. Kong (2008) conducted a follow-up study by tailoring inputs to Singapore’s fiscal system, but we deem her results as too optimistic. This was because past reserves were modeled as government net assets and could be utilized.³ However, the Singapore Constitution prevents easy withdrawals from past reserves as requires the consent of the Singapore President. Thus far, in Singapore’s economic history, there were two withdrawals from the national reserves. The first utilisation of Past Reserves happened in the financial year (FY) 2009 when S$4 billion was drawn to support and mitigate the impacts of the global financial crisis on the economy. The withdrawn amount was return two years later when the Singapore economy recovered, and the fiscal position strengthened. Toh (2012) updated Singapore’s generational accounts by modeling mortality using the Lee–Carter model. This paper contributes to the existing literature by incorporating the changes in Singapore’s fiscal policies, such as the inclusion of Temasek Holdings in the net investment returns utilization framework in 2016 and the second significant drawdown of reserves due to COVID-19 and hence evaluate the impacts of COVID-19 on inter-generational equity.

GA adopts the neo-classical model of the intertemporal budget constraint from the government’s perspective. Under the model, total future net tax payments that the government will receive from current and future generations must be sufficient to finance its future net purchases and pay off its existing debts. This constraint can be summarized as follows:

where $N_{t,k}$ is the account (net future lifetime tax payment) for the generation born in year $k$ evaluated at year $t$. The base year $t$ is set at 2019 (pre-COVID), and $L$ is the maximum possible age of 110. Hence, $A$ is the sum of the accounts of all generations that are currently alive, from those aged 110 ($k=1909$) to the current newborns ($k=2019$). Similarly, $B$ is the sum of the accounts of all future generations who are born after the base year ($k=2020$ onwards). $G_s$ refers to the government’s net purchases in year $s$, making $C$ the sum of all future government net purchases discounted to year $t$. $W_t^g$ refers to the government net debt in the base year. $N_{t,k}$ is further defined as follows : where $T_{s,k}$ is the estimated taxes paid in year $s$ by a resident born in year $k$, $B_{s,k}$ is the estimated benefits received in year $s$ by a resident born in year $k$, $P_{s,k}$ is the estimated number of residents born in year $k$ that are alive in year $s$, and $r$ is the discount rate.

Notably, the accounts for the generations that were born before the base year only include their remaining lifetime’s net taxes. Hence, fiscal sustainability is usually gauged by comparing the accounts for the current newborns against those for future generations, both of which cover the residents’ entire lifetime. In general, if the future generations’ accounts are at least as negative⁴ as that of the current newborns, current fiscal policies are deemed to be sustainable in the long run; otherwise, the future generations would be worse off than current newborns in terms of total lifetime net transfers with the government.

To compute the accounts of future generations, we used the “residual approach” as proposed by Auerbach et al. (1991). Using Equation (1), we first compute the accounts for all current generations ($A$) by projecting demographics, taxes and benefits forward by the current residents’ maximum remaining lifetime (110 years). $C$ and $D$ are also projected, leaving $B$ as the equation’s only “residual”. This residual is then divided across all future populations after factoring in productivity growth,⁵ thereby arriving at the generational account for a “representative” future individual, denoted by ${GA}_{t,t+1}$.

Each current generation’s account is divided by the number of currently surviving residents from that generation, to facilitate the comparison between “representative individuals” of different generations :

Using the account for future individuals ${GA}_{t,t+1}$ and that for current newborns ${GA}_{t,t}$, we assess the degree of fiscal sustainability and intergenerational equity via the absolute intergenerational gap, defined as follows : A positive $\mathrm{\Delta }$ implies that future generations will pay higher net lifetime taxes compared to current newborns.⁶

3. Numerical Implementation of Generational Accounts (GA)

The GA for Singapore is constructed using the base year set at 2019, to facilitate comparison between pre-COVID and post-COVID scenarios. First, we project Singapore’s demographics till 2130 by forecasting mortality, fertility and net migration which are the key drivers of population changes. Next, we calibrate the age-sex-specific per capita tax and benefit profiles using aggregate government revenue and expenditure data in fiscal year (FY) 2019. Assuming that these are steady state values, we combine the per capita profiles with the projected demographic landscape to obtain aggregate projections of future taxes and benefits. Finally, we estimate the present value of government net purchases and debt.

3.1. Projecting net tax benefits for current generation and net government position

While the principles behind GA using the residual approach are simple and elegant, its implementation requires the projection of components $A$, $C$ and $D$ of Equation (1).

3.1.1. Accounts for current generation (A)

We estimate the accounts for current generations ($A$) using the actuarial projections of future demographics and the calibrated per capita tax and benefit profiles. To obtain ($A$), we first need to project Singapore’s resident population annually by age and sex until 2130, covering the lifetime for all cohorts alive in 2019.⁷ The following sub-sections detail the calibration of assumptions on mortality, fertility and net migration, which are the key drivers of population changes.

Mortality projection

The well-established Lee–Carter stochastic demographic model (Lee & Carter, 1992) is used to forecast mortality rates by age and sex. The model can be described as follows :

where $m_{x,t}$ is the central rate of mortality for age $x$ in year $t;a_x$ is the average log-mortality for age $x$; $b_x$ is the sensitivity of mortality at age $x$ to mortality trend over time; $e_{x,t}$ is a white noise process; $k_t$ is the general level of mortality in year $t$ and is subsequently modeled as a random walk with drift : where ${\epsilon }_t$ is another white noise process.

We used historical mortality rates for ages 0–99 from the Complete Life Tables for Singapore Male/Female Residents from 2003 to 2018⁸ (DOS, 2020b) as our inputs. Assuming uniform distribution of death (UDD) within each age, the central rate of mortality for age $x$ in year $t$ can be obtained from the mortality rates $q_{x,t}$ in the life tables as follows :

The outputs from the Lee–Carter model are then smoothed using cubic splines to ensure consistency across ages. Subsequently, the Coale–Kisker method (Coale and Kisker, as cited in DOS, 2008) is used to extrapolate the mortality rates for ages 100–110, with mortality rate assumed to be 1 (certain death) at age 110. We thus obtain the forecasted mortality rates by sex for ages 0–110 up to year 2130.

Fertility projection

To incorporate the trend of delayed marriage and childbearing in Singapore (The Strategy Group, 2012) in our projections, we project age-specific fertility rates for females rather than total fertility rate (TFR). Historical fertility rates per female aged 15–49 for years 1980⁹–2019 are extracted from DOS (2020a). Since the data is available only in 5-year age bands, we smooth it by fitting scaled beta density functions, following the methodology by Tham (2014). This method allows us to capture the bell-shaped birth distribution with respect to mother’s age. For a particular year, letting $f(x)$ denote the fertility rate per female aged $x$, the model is expressed as follows:

where $\alpha$, $\beta$ and $K$ are estimated via nonlinear least-squares. Here, $\alpha$ and $\beta$ are the shape and scale parameters of the beta density function, while $K$ is a “scaling” parameter that reflects the overall fertility level in the year. To estimate future age-specific fertility rates, we need to project $\alpha$, $\beta$ and $K$ forward. As the United Nations (2019d) has classified Singapore amongst countries whose TFR would remain steady in the future, $K$ is assumed to stay at its fitted value for the year 2019. Next, we compute the expectation and mode of the fitted beta distribution each year from the fitted values of $\alpha$ and $\beta$ in that year as follows:

$$E(X)=\frac{\alpha }{\alpha +\beta },$$(10)

$$M(X)=\frac{\alpha -1}{\alpha +\beta -2}.$$(11)

We assume that the most likely childbearing age will steadily increase from the current value of 29.4 (author’s computations) to 35, consistent with the trend of delayed childbearing. This is done by modeling the mode $M$ via a random walk with drift. Then, we assume that $E$ is linearly dependent on $M$ (R${}^2=0.999$, authors’ computations) subjected to constraints.¹⁰ We then substitute the projected values of $E$ and $M$ into Equations (10) and (11) to “back out” the projected values of $\alpha$ and $\beta$ for each future year. Finally, with the projected values of $\alpha$, $\beta$ and K, we forecast age-specific fertility rates from 2019 to 2130. Figure 1 shows the projected future age-specific fertility rates¹¹ for selected years.

Additionally, the proportion of newborns each year who are expected to be males is set at 51.7%, which is the average of the historical newborns data (Government of Singapore, 2020c).

Net migration projection

The Singapore government sets a target number of migrants each year to achieve similar effects on the population as a TFR of 2.1. Hence, net migration is another key determinant of population structure, since migrants tend to be of prime working age. We model net migration using the projected net migration rates from World Population Prospects (WPP) up to 2100 (UN, 2019c). From 2101 to 2130, we extend UN’s projection by applying an ARIMA model to it. The projected net migrants in each year are obtained by multiplying the previous year’s resident population by the projected net migration rate in that year. Migrants are assumed to be evenly distributed across both sexes between ages 10–49 based on the profiles of new citizens and permanent residents in 2019 (National Population and Talent Division, 2020). Due to data constraints, we do not differentiate between Singapore citizens and permanent residents, and both are regarded as “residents” in our analysis.

Population projection results

Figure 2 displays the projected life expectancies by year of birth and sex from the Lee–Carter model’s outputs, compared with projections¹² in World Population Prospects 2019 (UN, 2019a; UN, 2019b). The projection for female’s life expectancy aligns quite closely to the projections in WWP. However, our model predicts higher life expectancy for males. However, our model predicts that life expectancies will increase at a decreasing rate, which is arguably more realistic compared to the relatively uniform increments in the WWP.

Combining the mortality, fertility and net migration projections, we forecast Singapore’s resident population to gradually increase from 4.03 million in 2019, before peaking at 4.49 million in 2048 and subsequently decreasing to 3.40 million in 2130. The projected population pyramids are displayed in Figure 3. The change in the population pyramid’s shape from the current beehive to an inverted triangle in 2130 indicates significant upcoming demographic changes as a result of aging population. The total dependency ratio, defined as the ratio of the number of residents aged below 15 or above 64 to the number of residents aged between 15 and 64, will increase significantly from 41.1% in 2019 to 89.1% in 2070 and 100.9% in 2130. With a shrinking tax base and growing social and healthcare expenditures implied by the rising dependency ratios, the government’s fiscal position is expected to face considerable challenges in the years ahead if current trends persist.

3.1.2. Pre-COVID taxes and benefits profiles

Salient features of Singapore’s fiscal structure

Unlike the IMF which uses a single fiscal headline indicator to summarize fiscal positions, the Singapore government assesses its fiscal sustainability over the short- and long-term using 3 different metrics (Chia, 2014). Firstly, the primary balance is obtained by subtracting operating and development expenditure from operating revenue, and helps assess the sufficiency of the government’s receipts in financing its key spending items within the fiscal year. Next, the basic balance is obtained by subtracting special transfers (excluding top-ups to endowment and trust funds) from the primary balance, and indicates the government’s short-term fiscal position after accounting for non-compulsory short-term payouts. Finally, the overall balance is obtained by subtracting top-ups to funds from and adding NIRC to the basic balance. It indicates Singapore’s long-term fiscal position by considering the long-term investment returns that we expect to receive from our reserves. Therefore, we will use the overall balance to implement GA for Singapore.

Singapore government’s operating revenue and total expenditure (excluding special transfers) have both been approximately 15% of GDP in the past 5 years up to 2019 (DOS, 2020d), which is lower than most OECD countries. A small and lean government is the testimony of Singapore’s philosophy of fiscal prudence, with its Constitutional guidelines governing the utilization of reserves accumulated in previous government terms. After three constitutional amendments in 2001, 2008 and 2016, the government can now spend up to 50% of the returns each year (as NIRC) from past reserves invested by the Monetary Authority of Singapore (MAS), Temasek Holdings and GIC Pte Ltd. As of 2019, NIRC has become the largest single-source government income at S$17.0 billion, at around 3.3 % of GDP. Hence, it is important to incorporate NIRC in the GA framework.

A salient feature of GA is to compute per-capita taxes and benefits by age and sex. We calibrate these via a top-down approach using aggregate annual Budget data. The following sections present the calibrated profiles. Details on calibration of taxes and benefits are in Appendices A and B, respectively.

Taxes per capita

We have identified 8 groups¹³ of taxes which in total account for 77.0% of the total operating revenue in 2019. The remaining revenues are apportioned uniformly to each resident under “Other taxes”. Figures 4(a) and 4(b) display the average tax per capita payable by male and female residents, respectively, in 2019.

The figures show that on average, male residents would be paying considerably more taxes than female residents of equivalent ages, with the difference arising mainly from income taxes and GST. Since both PIT and CIT are apportioned using income profiles by age and sex (Appendix A), the current income disparity results in males aged 30 onwards paying approximately twice as much income taxes as females of the same age groups.¹⁴ Furthermore, as GST is assumed to be paid by only heads of households,¹⁵ the fact that 72% of head of households in 2019 are males (Ministry of Social and Family Development, 2020b) means that males are apportioned significantly more GST than females.

Benefits per capita

Singapore government spending can be classified into operating expenditure, development expenditure and special transfers. We have identified 6 age/sex-specific government spending items, which together account for 34.2% of the total operating and development expenditures in 2019. All other expenditures are treated as implicit transfers and apportioned uniformly to all residents, under “Other benefits”. For special transfers, instead of using 2019’s value, we use the average of 2015–2019’s values due to its potential cyclicity with the electoral cycle.¹⁶ Figure 5 displays the average benefits per capita in 2019 for both genders.

Compared to the age-specific tax profiles, most benefits (e.g., Education, Family Development Programme) are effectively gender-neutral¹⁷ since they do not directly depend on income level. Furthermore, while the tax burden mostly falls on middle-aged residents (Figure 5), most age-specific benefits are received by children or elderly. This is unsurprising as children are the main beneficiaries of early childhood and education-related services, while elderly are the main beneficiaries of government healthcare-related expenditures. Thus, an aging population could put considerable financial strain on the government by increasing the healthcare-related expenditures.

Growth rate of future taxes and benefits

For projection of future taxes and benefits, it is common to assume that per capita taxes and benefits will grow at the same rate as productivity (Auerbach et al., 1994), as shown as follows:

The annual productivity growth rate $g$ is taken as the average labor productivity growth in Singapore between 2007 and 2019¹⁸ (CEIC Data, 2021) of 1.235%. Hence, all taxes and benefits per capita are assumed to grow at 1.235% from 2019 to 2130. This value is lower than the target productivity rate of 2–3% set by the Economic Strategies Committee in 2010, possibly because the slowdown of population growth and aging population would also reduce the rate of technological developments (Prettner, 2012) and growth in entrepreneurship (Liang et al., 2018).

Discount rate

To reflect the riskiness of future cashflows, a discount rate is needed to convert future monetary amounts to present values (Auerbach et al., 1991) in Equation (1). We assume the discount rate $r$ to be the average historical yield for 30-year government bonds (MAS, 2021) from 2012 to 2019,¹⁹ at 2.635%.

3.2. Net purchases and debt (C, D)

Next, we discuss the projection of government net purchases ($C$) and net debt ($D$).

Net purchases

In the previous GA studies, non-age-specific government expenditures are typically treated either as government net purchases ($C$) or implicit transfers as part of $A$ (Bonin, 2001). In our implementation, all such expenditures are treated as implicit transfers and apportioned uniformly to each resident. We thus assume that all residents would benefit evenly from them (e.g., national defense) and the benefits are valued at cost. Hence, the net government purchase ($C$) reduces to zero (Raffelhüschen, as cited in Bonin, 2001).

Net debt

Unlike most other governments, Singapore government’s assets exceed its liabilities due to its fiscal prudence. These net assets form the government’s past reserves and help “protect the lives and livelihoods of our people” (Heng, 2021) during periods of adversity. However, the Constitution only permits withdrawals from past reserves during exceptional circumstances. Furthermore, in the only time (pre-COVID) when S$4.9 billion of past reserves were drawn to support wage subsidies during the 2008 GFC, the amount withdrawn was returned in 2011. Hence, although the Constitution does not stipulate that any withdrawn past reserves need to be returned, we do not regard past reserves as “government net asset”.

Instead, one major innovation in our implementation of GA is our computation of the present value of future NIRC, which is treated as government net debt. NIRC is affected firstly by the reserve’s size, which is expected to increase since at least 50% of past reserves’ net returns must be reinvested. It is also affected by investment returns, which we expect to fall based on the trends of 20-year returns published by Temasek Holdings (2021) and GIC (2021). Based on our simulation results after factoring in these 2 opposing factors, we predict a compound annual growth rate (CAGR) of approximately −1.5%. Thus, NIRC is modeled as a perpetual cashflow that decreases at 1.5% each year. The government net debt ($D$) is the (negative of the) present value of this cashflow discounted at $r=2.635$%.

4. Generational Accouts Pre-COVID

With the constructed profiles and projections, we implement the residual approach (coded in R) to construct GA for the pre-COVID benchmark scenario. Under this scenario, the base year is set at 2019, $r=2.635$% and $g=1.235$%. Table 1 displays the generational accounts (expressed in SGD thousands) for current generations by sex in 5-year age intervals and for future generations.

Considerable differences can be observed between the accounts for males and females. While females of all ages enjoy net future lifetime benefits (negative accounts), males aged 20–50 are expected to pay net future lifetime taxes. Additionally, while both male and female current newborns enjoy net benefits, female newborns’ net benefits are more than twice as large as the males’. These results are consistent with the tax and benefit profiles in Figures 4 and 5, where middle-aged males bear significantly larger tax burdens than females while benefits are similar across sexes. This disparity diminishes when residents reach retirement, as shown by the convergence of the accounts at older ages.

It must be noted that the proportion of females amongst heads of resident households has been increasing considerably in the past decade, from 21.6% in 2010 to 28.0% in 2019 (MSF, 2019). This implies that taxes and benefits apportioned to heads of households will shift from males towards females in the future if this trend continues, thereby potentially decreasing the gender gap between the accounts. However, we will not extrapolate this trend in our projections, since it is unlikely to be exogenous; instead, it would likely happen in tandem with shifts in the relative income levels of males and females (which we use to apportion other taxes/benefits), amongst other things that are difficult to forecast. Thus, we will focus on each generation’s overall account rather than sex-specific accounts in subsequent analyses.

While a current newborn expects to enjoy a net lifetime benefit of S$465 thousand, future generations are expected to pay a net lifetime tax of S$47 thousand. This amounts to an absolute intergenerational gap ($\mathrm{\Delta }$) of S$512 thousand, indicating that the current fiscal policies are not sustainable in the long run since future generations would be significantly worse off than current newborns. Our conclusion of the fiscal system being unsustainable is consistent with the findings by Toh (2012), although the extent of intergenerational inequity in this study is considerably more severe. Toh estimated that although future generations would be worse off than the newborns in 2012, they would nevertheless be enjoying net lifetime benefits of S$223 thousand (compared to the newborns’ S$306 thousand). The intergenerational gap $\mathrm{\Delta }$ has therefore increased significantly from S$83 thousand in 2012 to S$512 thousand in 2019.

The variation in results might stem from how NRIC is modeled. While (Toh, 2012) modeled NIRC as an increasing cashflow at the rate of $g$, in our benchmark scenario, we model it as a decreasing cashflow after considering the opposing effects of increasing reserve size and decreasing returns under the current investment climate. Consequently, the present value of NIRC that can be distributed to future generations is much smaller. The difference could also be attributed to the different methodologies in our demographic projections, for examples using different fertility and immigration assumptions.

4.1. Sensitivity analyses

Collard (2004) noted that generational accounts can be highly sensitive to key parameters which affect future cashflows and/or demographics. To investigate the robustness of our results and the conclusion of significant intergenerational inequity, we conduct sensitivity analysis for $\mathrm{\Delta }$ on 5 variables: discount rate ($r$), productivity growth ($g$), NIRC growth rate, fertility rate and net migration rate.

Economic assumptions

First, the discount rate used directly affects the present value of future cashflows. Table 2 displays the sensitivity of the generational accounts under different discount rates.

As shown above, increasing $r$ by 0.5% causes an approximately 15% fall in $\mathrm{\Delta }$ relative to the benchmark scenario, and vice versa. This is expected as increasing $r$ would decrease the present value of all future cashflows, thereby diminishing the magnitude of both the current newborns’ net future lifetime benefits and the future generations’ net lifetime taxes. Nevertheless, we see that reasonable volatility in discount rates does not alter our conclusion of intergenerational inequity within current fiscal policies.

Second, the productivity growth rate $g$ is essentially a compounding factor as it is the rate at which taxes and benefits per capita are projected to grow in the future. Table 3 displays the generational accounts under different productivity growth rates.

Unlike $r$, increasing $g$ by 0.5% would cause an approximately 22% increase in $\mathrm{\Delta }$, and vice versa. This is due to $g$ affecting future cashflows in the “opposite way” as $r$; while $r$ reduces (the present value of) future cashflows, $g$ increases them. Hence, reasonable volatility in $g$ does not alter our conclusion that future generations are worse off than current newborns. However, we do not mean that productivity rates are inherently “undesirable”, since it affects the country’s overall standard of living and competitiveness.

Third, given that NIRC is modeled separately from the other cashflows, we want to assess the impact of its annual growth rate on the generational accounts, and the result is shown in Table 4.

As shown above, NIRCs year-on-year growth (currently at −1.5%) changes only the accounts of future generations, since NIRC is treated as government net debt in Equation (1). An increase in its growth rate of 0.5% decreases $\mathrm{\Delta }$ by about 7%, while a decrease of 0.5% increases $\mathrm{\Delta }$ by about 5%. As such, investment returns would have to be significantly higher than our projections if the government were to rely on NIRC to fund future expenditures and close the intergenerational gap.

Demographic assumptions

Generational accounts are also affected by future population demographics, since government taxes and benefits are age-based. Table 5 shows the results when we change future age-specific fertility rates.

It appears that increasing fertility rates would marginally decrease $\mathrm{\Delta }$, potentially because there would be more residents to share the total future net tax burden. However, we acknowledge that in general, the effect of fertility on generational accounts is indeterminate (Bonin, 2001) as it depends on whether the residual ($B$) of the intertemporal budget constraint implies a net tax or net benefit.

Next, while the Singapore government had historically adopted an open-door policy to attract foreign talents, it now carefully monitors the pace of immigration to achieve a balance between sustaining the labor force and maintaining cohesion in the society (NPTD, 2020). Table 6 shows the sensitivity of generational accounts to changes in net migration rates.

As shown above, when net migration increases, $\mathrm{\Delta }$ decreases marginally. This is because immigrants are assumed to be aged between 10 and 45 and thus would generally pay net future lifetime taxes (Table 6), since tax burdens mostly fall on residents of working ages. Nevertheless, the effect of the changes on the accounts is minor. This result agrees is consistent with conclusions drawn from previous studies (Bonin, 2001) that achieving intergenerational balance through immigration alone would be challenging.

5. Generational Accounts – Post COVID

In response to the significant impact of COVID-19 on the economy and livelihoods of Singaporeans, the Singapore government rolled out a total of four budgets from February to May 2020. These budgets, termed “Unity”, “Resilience”, “Solidarity” and “Fortitude”, summed up to an unprecedented S$193 billion (Heng, 2020a). They contained numerous stimulating measures such as the Job Support Scheme, under which the government would effectively pay wages on behalf of employers (subjected to conditions, as discussed subsequently). Some of these measures continued in Budget 2021, and the government approved to withdraw up to S$53.7 billion from the past reserves to support Covid measures across 2020 and 2021 (Heng, 2021).

These unprecedented events could significantly impact the generational accounts constructed pre-COVID. Firstly, the per capita taxes and benefits in 2020–2021 would be very different from the projections based on 2019’s profiles and need to be updated. Similarly, the present value of future NIRC would have to be recalculated with 2021 figures as the starting point. There has also been evidence of a “baby bust” in the form of fewer births (Kearney & Levine, 2020) partly due to the social distancing measures that many countries have enforced to limit the spread of COVID-19. As such, we would be updating our pre-COVID fertility rate assumptions. Sections 5.1–5.4 detail the changes to the pre-COVID assumptions.

To ensure comparability with the pre-COVID results, the base year in our model will continue to be 2019, even after updating the data and assumptions using 2020–2021 figures. In doing so, we aim to quantify the effects of COVID-19 and the government’s response measures on the intergenerational gap.

While the government is not obliged by the Constitution to return the past reserves drawn, in our model, we assume that past reserves for COVID-19 measures would be returned.²⁰ As mentioned by DPM Heng (2020b), it would be the “moral obligation and sense of duty to current and future generations” to return the reserves, even without such obligation. Since the government has not set an exact date, we assume that the government would strive to achieve overall budget neutrality from 2020 to 2029,²¹ which would return all withdrawn reserves by 2029. Section 6 discusses the key measures that the government are likely to implement to restore fiscal balance and their effects on the generational accounts. If such measures are projected to be insufficient in achieving budget neutrality from 2020 to 2029, we would evaluate additional changes required to achieve the said goal as well as to ultimately bridge the intergenerational gap in the long run.

5.1. Impacts on GA components

a. Impact on demographic assumptions

As mentioned by Minister Indranee Rajah (The Strategy Group, 2021) during the Committee of Supply debate, total marriages fell by around 10% from 2019 to 2020 and total fertility rate fell to a historic low of 1.1. We assume decline in births to continue and assume fertility rates to drop annually by 10% for 2020 to 2022.

b. Impact on taxes

Total government operating revenue in 2020 was S$64.4 billion, 13% lower than 2019. However, since taxes per capita are apportioned using the actual taxes collected by the government, it is difficult to directly incorporate specific measures such as tax rebates and exemptions in our model. Hence, we assume that the distribution of the tax burdens by age and sex will be identical to the pre-COVID situation, and the actual amount of taxes collected in 2020–2021²² would already incorporate the effects of tax-related measures. Therefore, taxes in 2020–2021 are apportioned the same way as before.

c. Impact on benefits

COVID-19 has significantly increased total government expenditures (including special transfers); they sum up to S$147.7 billion in 2020, 63% higher than the previous year. We firstly assume that the actual recurrent and development expenditures in 2020–2021 would have reflected the relevant measures, and apportion them the same way as before.²³ Next, we apportion five key special transfers that aim to protect Singaporeans’ livelihood, totaling S$34.3 billion. They are: Job Support Scheme (JSS), Workfare Special Payment (WSP), Wage Credit Scheme (WCS), Self-Employed Person Income Relief Scheme (SIRS) and Care and Support Package Cash Payout (CSC). All other special transfers will be apportioned in the same way as pre-COVID special transfers. Details on the apportioning can be found in Appendix C.

Figures 6(a) and 6(b) display the benefits per capita by age in 2020 and 2021, respectively. Only male profiles are displayed since the benefits are similar across both sexes, as discussed above.

Most of the COVID-19-related special transfers in 2020 were made to workers in the prime working age of 25–54, who were most impacted by retrenchments and business closures due to the pandemic. For example, under the Resilience Budget in 2020, 80% of the $17 billion requested from Past Reserves was directed towards JSS to help the severely affected sectors. This demonstrates that the government has deemed direct cash distribution in the form of wage subsidies as the most effective way to assist Singaporeans. Notably, all COVID-19-related special transfers are either discontinued or significantly reduced²⁴ in 2021, as shown in Figure 6(b) where the only remaining item is JSS. This could reflect the government’s optimism regarding the economy’s speed of recovery from the pandemic, as well as the “unsustainability” of transfers of such magnitudes.

d. Impact on government net purchases and net debt

As government net purchases are assumed to be zero, it is not affected by COVID-19. For net debt (NIRC), we use the actual 2019–2020 amounts and project the 2021 amount to perpetuity using the method outlined in Section 3.2. All cashflows are then discounted back to 2019 using the discount rate $r$. Notably, NIRC has increased from S$17.0 billion in 2019 to S$18.1 billion in 2020 and S$19.6 billion in 2021.

5.2. Impacts of COVID-19 on GA

To summarize, in the post-COVID model, the per capita tax and benefit profiles for 2020-21 are apportioned from actual data (rather than projected from 2019 profiles as in the pre-COVID model). From 2022 onwards, the tax and benefit profiles are the same as the pre-COVID benchmark scenario. Demographic projections are revised after adjusting for fertility changes; NIRC is based on actual FY2020/21 data. The assumptions regarding $r$, $g$ and NIRC growth rate are the same as the pre-COVID benchmark scenario. The base year is still set at 2019 to facilitate result comparison.

Table 7 shows the post-COVID generational accounts after making the various changes. For comparison, the first row pertains to the pre-COVID benchmark scenario.

It is clear that future generations would have to bear the burden of the additional transfers made to current generations, reflected by the increase in the intergenerational gap ($\mathrm{\Delta }$) from S$512 thousand to S$579 thousand. Remarkably though, the revised NIRC has considerably lowered the gap by S$42 thousand, due to the compounded effect of a higher starting point on the present value of the perpetual cashflow.

a. Expanding taxable capacity

GST hike

To meet the needs of an ageing population, there were plans to increase GST to raise revenues. (Singapore Budget 2018) However, the increase was delayed due to COVID-19. The GA calculations in this paper assume that GST hike occurs in 2022 and that the GST per capita increase from the pre-COVID estimates by 14.3% (1/7, assuming that 50% of the increase is offset) for 5 years until 2026. From 2027 onwards, GST per capita is assumed to increase by the full 28.6% (2/7) compared to pre-COVID estimates.²⁵

Carbon tax

Carbon tax was introduced in Budget 2018 to reduce the country’s carbon footprint and promote sustainable growth. It is levied to all industrial facilities above an emission threshold at the rate of S$5/tonne from 2020–2023 and is expected to increase to between S$10–15 per tonne by 2030 (National Environment Agency, 2021). Even if we assume that the carbon tax is eventually passed onto consumers, it is difficult to determine its burden by age and sex. Hence, we treat carbon tax as a proportion of “Other taxes” which are uniformly apportioned to all residents. Carbon tax rates are assumed to increase to S$10/tonne in 2024 and S$15/tonne from 2027 onwards, effectively doubling and tripling the revenue²⁶ as compared to 2021. This is reflected by increasing “Other taxes” by 1 time and 2 times of said proportion, respectively.

Table 8 displays the post-COVID generational accounts when we account for the hikes in GST and carbon taxes. The GST hike is expected to significantly reduce $\mathrm{\Delta }$ by S$109 thousand, putting the future generations in a net benefit instead of net tax position. The carbon tax increase will reduce $\mathrm{\Delta }$ by a further S$20 thousand. As such, actual NIRC 2020/21, GST hike and carbon tax increase reduce $\mathrm{\Delta }$ by a total of $170 thousand. However, future generations would still be worse off than the current newborns, as shown by the final intergenerational gap of $408 thousand.

b. Other policy recommendations – special transfers and top-ups

For the government to return the past reserves withdrawn by the end of its following term, it will need to achieve overall budget neutrality across the period of 2020–2029. In our estimation, even after incorporating the GST hike and carbon tax increases, there would be an overall deficit of S$14.4 billion (discounted to 2019 dollars) during this period.²⁷

The government could potentially make up this deficit via a reduction in special transfers. This is because special transfers are non-recurrent by design, and the government has set up and made top-ups to endowment and trust funds during periods of fiscal strength (Chia, 2014). These funds can continue to fund social protection programs, albeit at reduced amounts. Such reductions are already evident in FY2021 where special transfers are reduced to S$4.9 billion, compared to S$15.1 billion in FY2019.

We thus conducted simulations by applying percentage decreases in special transfers from 2022 onwards²⁸ with the goal of achieving overall fiscal neutrality from 2020 to 2029. On top of that, we also conducted simulations to identify the percentage decrease in “Other benefits”²⁹ needed to fully close the intergenerational gap and achieve intergenerational equity in the long run.

As shown in Table 9, a 19.9% decrease in special transfers from pre-COVID levels is required to return the past reserves by 2029. This is equivalent to an annual decrease of S$1.9 billion in special transfers in 2019 dollars. However, the intergenerational gap ($\mathrm{\Delta })$ would still amount to S$338 billion after such reductions, and a further 18.2% decrease in “Other benefits” (annual decrease of S$9.0 billion in 2019 dollars) starting in 2022 would be required in theory to bring $\mathrm{\Delta }$ to 0.

A multi-pronged approach involving multiple policies would be a more feasible way to achieve the aforementioned goals of fiscal balance and intergenerational equity, since each individual instrument would need to be changed by a smaller extent (Kotlikoff & Leibfritz, as cited in Toh, 2012). That being said, sensitivity analyses in Section 4.1 highlight that generational accounts are significantly more sensitive to changes in fiscal policies which directly affect taxes and benefits than demographic policies. Hence, while increasing fertility rates and net migration may be the only ways to sustain the population size, they do not prevent future generations from having to bear a larger net lifetime tax burden than the current newborns.

The Singapore government has conscientiously strive to maintain an overall budget balance over its term of government, with NIRC contributing around 3% of GDP to its annual budget. Not only does NIRC play a significant role in the annual budgetary process, the growth rate of NIRC also impacts the intergeneratioal gap. Our numerical simulations show that when taxes and benefits increases at the rate of g but if NIRC has negative growth rate, intergenerational gap will worsen. As long as NIRC growth rate is positive, intergenerational inequity will be reduced.

6. Conclusions

Singapore’s rapid transition toward a super-aged society puts tremendous pressure on the government to maintain fiscal sustainability. In response, the Singapore government has adopted a unique social protection scheme by earmarking and topping up various funds to fund long-term payment schemes. It has also been tapping on larger proportions of returns from past reserves by adding Temasek’s returns to NIRC since 2015. The generational accounting framework for Singapore shows that due to ageing population and falling fertility, pre-COVID, the fiscal system has an intergenerational gap of S$512 thousand. There is significant disparity across genders: female newborns are expected to enjoy more than twice the amount of net lifetime benefits from the government as compared to male newborns (S$681 thousand to S$259 thousand). However, it does not mean that males in general have lower expected standards of living than females; instead, this disparity precisely reflects the government’s role in allocating resources to those in greater need, given the current empirical differences in income distribution between the sexes.

Additionally, our simulations show that intergenerational inequity cannot be addressed merely through pro-natalist or immigration policies alone. Pro-natalist policies do not have significant impacts on generational accounts since they redistribute the fixed burden over a larger future population. The government also has to strike a balance between expanding the labor force through increasing the pace of immigration and ensuring social cohesion (NTPD, 2020). Hence, Singapore’s fiscal policies require significant changes to be sustainable.

Our computations also show that fiscal measures used to deal with COVID-19 and the impact of COVID-19 on short-term fertility rates would increase the intergenerational gap by S$67 thousand, implying that future generations indeed have to bear the consequences of increased fiscal spending during COVID-19. However, higher-than-expected NIRC, GST hike and implementation of carbon tax would decrease the gap by S$170 thousand in total. Fiscal policies are thus more effective than demographic policies in changing intergenerational redistribution.

Nevertheless, even after the aforementioned policy changes, the government would still incur an overall deficit of S$14.4 billion from 2020 to 2029 and the intergenerational gap remains. Fiscal balance the could be restored by substantially by reducing special transfers by 19.9% from pre-COVID levels. A further 18.2% reduction in non-age-specific benefits from pre-COVID levels would fully eliminate intergenerational imbalance by reducing the intergenerational gap to zero.

One limitation of this paper is its inability to account for the funds that the government had set up before COVID-19. Historically, the government has earmarked capital for various endowment and trust funds (e.g., Pioneer Generation Fund) to finance social protection payments in the long term. While future top-ups to the funds are subsumed under special transfers, we do not have data on their current balances which would eventually be distributed to the beneficiaries. Hence, we could have underestimated the government’s fiscal strength and thus overestimated the extent of intergenerational inequity in Singapore.

There have been various discussions on the residual approach’s limitations. For one, the notion of a “representative” future account relies on the assumption of every future generation’s account being equal after adjusting for growth, which is unlikely to hold. In response, Boll (as cited in Bonin, 2001) introduced the “sustainability approach” under which current and future generation’s accounts are both calculated by continuing current fiscal policies in perpetuity. However, this approach is not used as it is arguably less useful in quantifying the intergenerational redistributive effect of current fiscal policies (Bonin, 2001).

Another limitation of conventional GA would be the potential bias induced by the selection of the base year. Due to the cyclical nature of short-term fiscal policies (Bonin et al., 2013), the base year’s numbers may not constitute a representative basis for future fiscal outlook. As such, de-trending procedures such as the Hodrick and Prescott filters (Bonin et al., 2013) have been used to remove the cyclicity from budget balances before making projections. However, such methods are less applicable when apportioning specific taxes and benefits which are affected by structural policy changes over time. Hence, most taxes and benefits are apportioned using actual (non-smoothed) data with the exception of special transfers.³⁰

Notes

¹ A generation in GA refers to either males or females born in a particular year.

² NIRC refers to contributions to the government revenue from Singapore’s past reserves, in the form of income and capital returns. See Ministry of Finance, What is NIRC? Available at http://ask.gov.sg/mof/questions.

³ See Chia (2014) for details; the treatment of past reserves will be discussed in more details in Section 3.

⁴ Since generational accounts are defined from the government’s perspective, a more negative value indicates lower net lifetime taxes (higher benefits).

⁵ All cashflows are assumed to grow at the rate of the country’s productivity growth.

⁶ Alternative measures of fiscal sustainability include the “standard sustainability index” (Bonin, 2001) which is the ratio of GA${}_{t,t+1}$ and GA${}_{t,t}$ factoring in productivity growth. However, we do not use it since it increases exponentially by design as GA${}_{t,t}$ approaches zero and can be misleading due to its sign.

⁷ The maximum age of a Singapore resident is assumed to be 110 years.

⁸ The 2019 life tables are not used as they are preliminary figures, as of March 2021.

⁹ 1980 was the first year that total fertility rate for females aged 45–49 became available.

¹⁰ For a beta distribution to be unimodal, $\alpha$, $\beta$ must be positive; furthermore, its mean must be greater than its mode when mean $<0.5$, and its mean must be smaller than its mode when mean $>0.5$.

¹¹ Projections stabilize after year 2070 (author’s computations).

¹² Medium variant, up to the year 2100.

¹³ Personal income tax (PIT) and corporate income tax (CIT) are treated as one group.

¹⁴ In reality, the tax structure also gives preferential treatments such as tax exemptions to working mothers.

¹⁵ Appendix A gives the detailed apportioning methodology for GST.

¹⁶ Appendix B gives the detailed rationale behind the apportioning methodology for special transfers.

¹⁷ The “dip” in the educational benefits per capita for males at age 18–19 is due to most males of that age having to go through national service and thus not being classified as students.

¹⁸ Data from 2007 to 2019 are used because these 2 years, on hindsight, are the years immediately preceding a major economic downturn (Global Financial Crisis in 2008 and COVID-19 in 2020). As such, the data within this period arguably cover exactly one business cycle, eliminating the cyclical nature of productivity growth.

¹⁹ Data are only available from 2012 onwards.

²⁰ In response to whether the amount withdrawn from the reserves would be returned, the then Finance Minister Mr Heng Swee Keat explained that “there is no obligation under the Constitution for the Government to put back the amount withdrawn… rather it is about having the moral obligation and sense of duty to current and future generations…” (Heng, 2020b)

²¹ The condition for budget neutrality is the net present value of all taxes, benefits and NIRC from 2020 to 2029 summing up to zero.

²² Data for 2020 are revised figures in Budget 2021, while data for 2021 are estimated figures.

²³ There have been minor post-COVID changes to a few of the benefits, namely: FDP, where the government introduced a new Baby Support Grant in 2020 to encourage childbearing during COVID-19; GST vouchers, where the government increased the cash and U-Save special payments in 2020 and 2021; WIS, where the government increased the maximum eligible income from 2020 onwards and increased the maximum payment. However, these changes do not affect the overall apportioning mechanism of the benefits themselves when they are incorporated in the calculations.

²⁴ In Budget 2021, the other 4 COVID-19-related special transfers are subsumed under “Other special transfers”.

²⁵ GST was increased in two steps. It was hiked to 8% in 1 Jan 2023 and then to 9% in 1 Jan 2024. However, the GA calculations are completed before the actual implementation of the GST hikes and assume that GST hike occurs in 2022.

²⁶ Similar to GST hike, we acknowledge that the increase in carbon taxes may result in emission reductions, thereby causing the tax revenue to be less than the projected amounts.

²⁷ Computed by summing up all taxes, payments and NIRCs from 2020 to 2029 and discounting to 2019.

²⁸ We assume that such decreases would be non-reversible and extended beyond 2029.

²⁹ Includes all other non-age/sex specific expenditures which are uniformly distributed to residents.

³⁰ See Appendix B for its treatment, due to its potential cyclicity with the electoral term rather than business cycle.

³¹ Minimum working age assumed to be 17, when the resident is no longer under the “Employment of Children and Young Persons Regulations” by MOM. Maximum working age assumed to be 79 based labor force participation rate comparisons for the age category “above 70”.

³² For consistency, this ratio is applied to all taxes and benefits apportioned by age group of head of household.

³³ Defined as the property’s estimated gross annual rent should it be rented out.

³⁴ It is assumed that the average betting amount per gambler does not change with age group.

³⁵ For the purpose of the survey by MOH, binge drinking was defined as consumption of at least 5 or 4 alcoholic drinks for males and females, respectively, in a single session within the month before the survey.

³⁶ Part of the total cash gift for babies born in 2018 would be distributed in 2019, since the cash gift is given in 5 instalments up to when the baby turns 18 months old.

³⁷ We considered combining MediShield Life Premiums with hospital admission rates to estimate the average size of hospital bills by age. However, this would be difficult due to scheme’s various terms and conditions. Furthermore, the premiums are structured such that residents pay more than their expected claims while young to cross-subsidize those in retirement ages, and therefore may not give an accurate picture of the hospital bill sizes.

³⁸ Employees and self-employed workers having the same income and age receive different WIS amounts.

³⁹ Upon satisfying the income eligibility condition, the cash payout amount is dependent on the annual value of the resident’s house. Additionally, residents are unemployed or not in labor force are added to the employed residents since they may also receive the payments, unlike the case of WIS.

⁴⁰ Medisave payout amount is dependent on age group and annual value of the resident’s house.

⁴¹ This may not be realistic in the case of endowment and trust funds, which are designed to finance long-term payments. However, due to the difficulty in modeling the amount, duration and pattern of such payments, we instead assume instant receipt regardless of the form of the transfer.

⁴² The maximum qualifying gross monthly income is S$2,300 as of 2020.