2.1 Earnings paths

In the base model, we assume that male workers enter the workforce at age 25 and female workers at age 23; and they retire at age 65. The starting wages and starting CPF balances are based on CPF administrative data of workers at that age. Table 1 reports the starting wages of male and female workers, which are used to anchor the lifetime earnings path for different income percentiles.

Table 1. Starting wages of male and female workers (in 2012 dollars)

Source: CPF administrative data.

With regards to wage growth rates, both OECD (2012) and Hui (Reference Hui2012) assumed constant annual growth rates. We use data from the Ministry of Manpower (MOM) Labour Force Survey from 2001 to 2011 to simulate real wage growth paths for individuals since wage growth is a key factor in determining how much CPF savings are accumulated.

As an illustration, for each income percentile level, to calculate the real wage growth of a worker aged 30–34, we first rebase the wage series to the same reference year before computing the annualized real wage growth for: (i) workers who moved from age 25–29 (in 2001) to age 30–34 (in 2006) and (ii) workers who moved from age 30–34 (in 2001) to age 35–39 (in 2006). The real growth rates from (i) and (ii) are then averaged to avoid under/overestimation of the real wage growth. The same procedure is repeated using data from 2006 to 2011. To account for cyclical effects, we take the average of the annualized real wage growths over the two 5-year periods (i.e.2001–2006 and 2006–2011) to obtain the real wage growth for workers aged 30–34. The same process is repeated to calibrate the real wage growth of workers in other 5-year age bands.

With detailed empirical data from the Labor Force Survey, we map out the age-earnings profile of workers over their working life for each income percentile, thereby providing more realistic wage growth data for the model. Figures 1a and 1b show the resulting empirical age-earnings profile for workers at different income percentiles.Footnote ⁸ It is a hump-shaped distribution of earnings by age where wage growth is faster when the worker is young and tapers off as he gets older. This profile could be attributed to the effects of human capital. In the initial years of work, average earnings rise as workers accumulate human capital (skills and experience). It falls around age 45–50 possibly due to changes in labor-leisure preferences as workers reduce their work hours and phase into retirement.Footnote ⁹

Figure 1. (Colour online) (a) Empirical age-earnings profile (male workers). (b) Empirical age-earnings profile (female workers). Source: Authors’ calculations.

Savings accumulation depends on the earnings profile. The hump-shaped age-earnings profile may be indicative that seniority-based wage structure is no longer a norm in Singapore. A hump-shaped profile also implies that workers will save more during the early work cycle, and the savings will be compounded for a longer period. We also assume that workers remain in their respective income percentiles throughout their working life. Workers, in reality, can possibly move across income percentiles through their working careers. For sensitivity study, we allow workers to move to higher income percentiles.

2.2 CPF policy parameters

The CPF parameters used in the simulation model are based on policies in September 2012. The contribution rates and the allocation of contributions into the different accounts are tabulated in Table 2.

Table 2. Rates of contribution and allocation as of 1 September 2012

Source: CPF website at http://www.cpf.gov.sg.

Notes: All figures are in percent of wage. Figures above are for monthly wages of $1,500 and above.

Interest rates for the OA are at the statutory minimum of 2.5% per annum. Returns to savings in the SA, MA, and the Retirement Account (RA) are at the floor rate of 4%. For simplicity, we use these rates to project accumulated CPF savings over the work cycle. Furthermore, we assume that the CPF Board will continue to give an extra 1% interest on the first $60,000 of CPF savings including up to $20,000 in the OA. These assumed long-run rates are equivalent to a real interest rate of 0.7% per annum for the OA and 2.2% for the SA and RAs, if we factor in a long-run inflation rate of 1.8%, which is the 20-year annualized inflation rate using the consumer price index for the period 1991 to 2011. These assumed rates are lower than the historical real rate of return for these accounts.Footnote ¹⁰

Contributions to CPF are subject to a salary cap. The salary ceiling was $5,000 in 2012. However, in anticipation that CPF will adopt the recommendation by the Economic Review Committee 2003 to peg salary ceiling to the income of the 80th percentileFootnote ¹¹, we set the salary ceiling to the wage growth of the 80th income percentile, i.e. at a constant real rate of 1.6% (MTI, 2003). The growth rate is based on the historical compounded annual growth rate of the 80th percentile gross monthly income from work among full-time workers from 2001 to 2011. The data were provided by the MOM.

2.3 Housing consumption

About 85% of Singaporeans live in subsidized public housing built by the Singapore Housing Development Board (HDB). To encourage home ownership, HDB homebuyers enjoy explicit subsidies and these flats are priced affordably. There is a range of HDB flats in terms of size, quality, and finishing to meet different housing needs and aspirations.Footnote ¹² HDB flat types are categorized by the number of rooms.Footnote ¹³

We assume that married couples purchase HDB flats directly from the HDB under the build to order (BTO) scheme.Footnote ¹⁴ The marriage age of male workers are set at 30. His spouse, who belongs to the same income percentile, is aged 28.Footnote ¹⁵ We assume that they buy a flat type within their financial means. The 30th income percentile (30P) buys a three-room flat; 50th income percentile (50P) buys a four-room flat and 70th income percentile (70P) buys a five-room flat. Eligible grants are included in the study to compute the effective home mortgage loans.

Besides market price subsidies, the government also provides CPF Housing Grant to all eligible HDB homebuyers. These grants are in terms of CPF monies. The grants are credited to the buyer's CPF OA and can be used to pay for the CPF portion of the initial down payment. The balance, if any, must be used to reduce the mortgage loan. Grants are tiered according to the socio-economic characteristics of HDB buyers (earned incomes, singles, couple, or singles/couples with parents staying them) and according to the types of flats purchased (new flats purchased directly from HDB, flats from the resale secondary market or premium flats purchased from private developers). Higher grants reduce the effective housing loans and thus lead to smaller withdrawals from the OA to pay for the reduced mortgage payments.

In addition to the CPF Housing Grant, there are additional housing subsidies for low- and middle-income families to buy subsidized HDB flats. These include the Additional CPF Housing Grant (AHG) and Special CPF Housing Grant (SHG). AHG was introduced in 2006 and SHG was introduced in 2011. Under AHG, first-time homebuyers in Singapore with a monthly household income below $5,000 are eligible for additional housing grants.Footnote ¹⁶ The AHG amount ranges from $5,000 to 40,000 depending on the buyers’ average gross monthly household income. As can be gleaned from Table 3, low-income households with $1,500 or less can get an AHG of $40,000. AHGs are included in the calibration of the effective housing loans.

Table 3. Additional CPF housing grant

Source: CPF website.

Special CPF Housing Grant (SHG) targets low-income families, with average household income below $2,250, to help them buy new HDB flats. They are however limited to buying smaller (two-room or three-room) standard flats in non-mature estates. SHG, which ranges from $5,000 to 20,000, is given on top of AHG. In this study, only 30P workers buying HDB are eligible to receive both SHG and AHG.

Table 4 summarizes the HDB types bought by the different income percentiles, the housing price, housing loan, and mortgage. The computed monthly mortgage payment is based on the effective housing loan. This is the flat price net of the housing grants and initial down payment. The HDB flat prices used are the median BTO selling price by flat type in 2011. The prices used are $233,000 for three-room flat, $372,000 for four-room flat and $468,000 for five-room flat.

Table 4. Housing consumption and financing

Source: Authors’ calculations.

Note: The effective housing loan is net of housing grant and house down-payment.

Since we assume that an entrant male worker at age 25 will buy a BTO flat when he turns 30, the BTO housing prices in 2011 dollars need to be adjusted to account for housing appreciation over 5 years. Based on the sale price of new flats reported in HDB Annual Reports, the real annualized growth rate of new flat prices between 2008/09 and 2010/11 is about 3%. However, we choose to use a lower real housing appreciation rate of 2%. This is because the government has articulated its housing policy to moderate the price increase of new flats to make public housing affordable. To achieve this policy objective, HDB has also increased the supply of BTO flats.

We assume that the couple will fully utilize their accumulated CPF savings in the OA for down payments, which is a requirement set by the HDB. The remaining outstanding amount, net of housing grants, is financed using HDB loans, up to a maximum of 90% of the purchase price. The HDB mortgage-financing rate is at 2.6% per annum and mortgage duration is up to 30 years. The monthly mortgage will be paid from monthly mandatory contributions to the CPF OA. CPF monies, however, cannot be used to pay for house-related expenses such as property tax, maintenance fees, and home improvement. In our base case, we assume joint financing. The monthly housing mortgage payment is shared equally between the couple and is paid out from their CPF monies in their OAs.

The computed monthly mortgage payments for the different percentiles are presented in Table 4. The last column shows the ratio of the monthly mortgage payment as a percentage of salary at the time of house purchase. The monthly mortgage cost for Singaporean household buying subsidized public housing is <15% of household income. Housing cost burden is much lower than most OECD countries. In the USA, housing costs were 20% of household incomes for above-median households and 32% for moderate-income households (Haas et al., Reference Haas, Hickey, Lubell and Morse2012). Not only is the housing cost burden lower in Singapore, there is a financing mechanism set in place to pay for the housing mortgage. Homebuyers can use their monthly contributions to pay for the monthly housing mortgage. The ratio is smaller than the mandated total contributions into the OA. For example, for workers under 35 years old, CPF contributions from both employer and employee amounted to 36% of the monthly salary. Out of this 36%, 23% is credited to the OA (see Table 2). This means that the monthly CPF contributions in the OA can meet the entire monthly mortgage payments. Unutilized savings in the OA will accumulate as retirement saving. As income increases over the work-cycle, excess savings in the OA net of mortgage payments will also grow.

2.4 Retirement income

The model also incorporates the features of the CPF payout phase.Footnote ¹⁷ Although CPF members can withdraw part of their savings at age 55 by pledging their property, they cannot cash out the entire amount of savings. They have to set aside a decreed retirement amount in their RA. This account is created when a member reaches 55. The decreed sum is cohort-specific, set by the CPF Board. Since 2009, CPF members have to annuitize the monies in their RA under the national annuity scheme known as the CPF LIFE (CPF Lifelong Income for the Elderly). There are two CPF LIFE plans – the Standard and Basic Plans. Both plans allow for a bequest. The Standard Plan, which is also the default plan, annuitizes a larger proportion of accumulated savings, provides higher monthly payouts but leaves less as a bequest.Footnote ¹⁸ The Standard Plan, annuitizes a smaller proportion of accumulated savings and gives smaller monthly payouts, in return for a higher bequest.

In the benchmark model, we assume that the entire accumulated CPF savings in the SA and OAs net of withdrawals for housing are fully annuitized. Premiums for CPF LIFE are assumed to be paid in two tranches. The first tranche is paid at age 55, when the net accumulated savings up to the full retirement sum is committed to buying the CPF LIFE product. Amounts above the full retirement sum, if any, will remain in the OA and SAs. This amount with interest accrued, together with new CPF contributions from work at age 55 to age 64, will be used to pay the second tranche of CPF LIFE premiums at age 65.

The first monthly payout begins at the drawdown age of 65. Monthly payouts from the simulated CPF balances are provided by the CPF Board. The formulae used to calculate the payouts are not publicly accessible. The payouts are different for males and females as females have a longer life expectancy. No income tax is imposed on CPF payouts.

Although CPF LIFE redresses longevity risks, it does not address inflation risk. CPF LIFE monthly payouts are fixed; without cost of living adjustments and not linked to the consumer price index. Thus, the replacement rate computed using annuity payout at the drawdown age of 65 for the first year of retirement ($y_{65}^g $) does not reflect retirement adequacy for subsequent years. We, therefore, indexed the CPF LIFE payouts to inflation to reflect ‘real annuity, which is more consistent with the goal of smoothening real consumption during retirement.’ (Brady, Reference Brady2010, p. 242). Following Brady, we compute the average retirement consumptions, weighted by survival probabilities, so that retirement income received in the earlier year of retirement is given greater weight.

The expected real annual retirement income at age t is given by ($Y_t^g $) and defined as follows:

(1)$$Y_t^g = \displaystyle{{y_{65}^g} \over {(1 + \pi )^{t - 65}}} \times {}_t^g p_{65}, $$

where $y_{65}^g $ is the nominal annuity payouts starting at age 65 and g indicates the gender. ${}_t^g P_{65} $ denotes the probability of survival of an individual at age 65 up to time t. The time t indexes the age of the individual. The maximum age is 100.Footnote ¹⁹ To obtain the real value of the retirement income, the CPF LIFE payouts are discounted using core inflation. Compared with headline inflation, core inflation, which excludes private transport and housing rentals, is a better measurement of the cost of living adjustment since the majority of Singaporeans are homeowners and use public transport. Using the core inflation data series from 1990 to 2013 published by the Monetary Authority of Singapore (MAS), we set the average core inflation to 1.8%.Footnote ²⁰

The average real annual payouts for different genders (AY ^g) is defined as:

(2)$$AY^g = \; \; \; \; \displaystyle{{\mathop \sum \nolimits_{t = 65}^{100} Y_t^g} \over {\mathop \sum \nolimits_{t = 65}^{100} {}_t^g p_{65}}}. $$

3.1 Retirement adequacy of workers under base case scenario

We examine the retirement adequacy of lower-middle (30P), median (50P) and upper-middle income (70P) male workers at age 65. Table 6 summarizes the average lifetime earnings, retirement incomes and the net replacement rates under the base case scenario. Net replacement rate in the first year of retirement is the ratio of CPF LIFE annuity payouts at age 65 to average lifetime income. The net replacement rates under the Standard Plan for lower-middle, median, and upper-middle income male workers are 72.6%, 63.2%, and 59.5%, respectively. See Table 6. The net replacement rates for lower-middle, median, and upper-middle income female workers are higher, at 69.1%, 60.6%, and 60.3%, respectively. The corresponding replacement rates under the Basic Plan for male and female workers are relatively lower by about 4–6 percentage points.

Table 6. Average lifetime earnings, annuity payouts, and net replacement rates under base case

Source: Authors’ calculations.

The net replacement rates in the first year of retirement are comparable with the OECD for full career workers entering the labor market in 2012. The net replacement rate for median earners in Singapore is lower than the net pension replacement rates for median earner in some OECD countries, for example, Australia (75.6%), Austria (89.9%), Canada (64.4%), and Switzerland (77.8%). It is comparable with Belgium (63.9%) and Norway (63.8%). It is higher than some OECD countries, for example, Germany (57.8%), Japan (49.1%), Korea (49.1%), Sweden (55.3%), the UK (48%), and the USA (49.9%). The net replacement rate for median income earners in the OECD countries is 69%. See OECD (2013b, p.143).

Table 6 also presents the average real annuity monthly payouts, calculated by using equation (2), using a core inflation rate of 1.8% per annum. The average net income replacement rate uses average real annuity payouts in the numerator and average real lifetime net earnings in the denominator. Accounting for inflation, the average net replacement rates are lower than replacement rates in the first year of retirement. Under the Standard Plan, average replacement rates are 52.6% and 49.3% for median male and female workers, respectively. The simulation results show that average net replacement rates for median male and female workers would be 10.6 and 11.3 percentage points lower if CPF LIFE annuity payouts are not inflation indexed. Replacement rates without inflation adjustments are lower than the OECD average, but are comparable with Germany, Japan, Sweden, the UK, and the USA. Our findings indicate that if CPF LIFE annuity payouts were inflation-indexed, then the CPF system would deliver better retirement support than some OECD countries.Footnote ²³

Next, we perform sensitivity analyses of the average replacement rates using different core inflation rates. The core inflation rates used are 0.5%, 1.0%, 1.5%, 2.0%, and 2.5%, whereas 1.8% is used in the base case scenario. Table 7 displays the computed average replacement rates for various core inflation rates under the Standard and Basic Plans.

Table 7. Sensitivity of average net replacement rate to inflation

Source: Authors’ calculations.

Table 7 shows that there is a negative relationship between average replacement rate and core inflation across different income percentiles. For example under the Standard Plan, the average replacement rate for median male workers decreases from 52.6% to 49.3% as the core inflation increases from 1.8% to 2.5%, which further increases to 59.9% when the core inflation drops to 0.5%.

3.2 Average net replacement rate with imputed rent

The homeownership rate in Singapore is about 90.5%. Since the mid-1980s, home ownership in most OECD regions has increased. OECD (2013b) opined that the upward trend in home ownership could be attributed, in part, to population aging as older people are generally more likely to be homeowners. Andrews et al. (Reference Andrews, Caldera Sánchez and Johansson2011) found that up to 1 percentage point of the growth in owner-occupancy homes is due to aging. The effect is most pronounced in Canada, Denmark, Germany, and Switzerland. With high elderly homeownership, it is imperative to include imputed rents in the replacement rates calculations to better reflect the state of retirement adequacy among home-owning pensioners. Even if they do not monetize their housing assets because of strong bequest motive and/or strong aging-in-place preference, being homeowners can make a big difference to retirees as they need not pay rents and thus have more resources available for consumption. For Singapore, rental costs would have been the largest expenditure item for most healthy retirees if they were renters. Imputed rental would come out to about 37% of the household budget. See DOS (2014c, Table 46).

According to Munnell and Soto (Reference Munnell and Soto2005), imputed rent is the amount that the owner would have to pay to rent an equivalent dwelling and is treated as an amount that the homeowners pay to themselves. It is included in both the numerator and denominator of replacement ratio. Imputed rent is included in the numerator because it is considered as part of the retirement income, which supports housing consumption during retirement. It is also included in the denominator as worker receives imputed rent as part of his income before retirement.

Incorporating imputed rent into the replacement rate formula is particularly challenging as there are various measurements for imputed rent. We perform two separate sets of income replacement calculations that factor in imputed rent. We will first report replacement rates, using imputed rent for the retiree households published in Department of Statistics (DOS, 2014c, Table 46). This uses the rental equivalent approach based on the housing asset. This approach, however, did not consider the asset cost burden and the tight link between housing and retirement policies. Section 3.3 will study these considerations in details.

In benchmarking the imputed rent, DOS uses the estimated market rent of the flat if it is rented out an unfurnished net of maintenance cost. DOS's estimation of imputed rent is based on the AAV (annual assessed values that the Inland Revenue Authority of Singapore (IRAS) uses to compute the property tax payable. The imputed rents per retiree household member are $486, $541, and $560 for three-room, four-room, and five-room HDB flats, respectively. In computing replacement rates to account for homeownership, we added imputed rents to both the numerator and denominator since mortgage payments are paid from the CPF OA.

Table 8 compares the average net replacement rates in the base case with and without imputed rent. Including imputed rent would augment the replacement rates by percentage points ranging from 4.6 to 8.1 for male workers, and from 5.7 to 9.3 for female workers, depending on the annuity plans and the income percentiles.

Table 8. Average net replacement rates with and without imputed rent, using published data from DOS

Source: Authors’ calculations.

3.3 Pre- and post-retirement net imputed rent

Because of the asset cost burden and the tight link between housing and retirement policies, there are differences in the net imputed rent before and after retirement.Footnote ²⁴ According to Cronin (Reference Cronin1999, p. 12), net imputed rent is gross rent minus the costs of homeownership. Homeownership costs include mortgage interest payments, property taxes, depreciation, maintenance, and repairs. DOS's computation of net imputed rent considered only maintenance costs but not mortgage interest payments. Furthermore, the imputed rent concept did not account for maximum house tenure allowable under housing loans. The loan tenure rule is to ensure that most homeowners no longer need to make mortgage payments by the time they retire. This implies that the amount of imputed rent added to the numerator (post-retirement) should not be the same as that in the denominator (prior retirement).

We adopt Cronin's approach to compute the pre- and post-retirement net imputed rents. The median rental rates for different flat types for the period 2007Q3 to 2012Q2 are culled from the HDB website. The subletting rental values are then re-based to 2012 dollars before averaging across the period. Column 2 of Table 9 shows the subletting HDB rental rates for different flat types. In addition, property tax rates are based on the IRAS's property tax schedule. The first $8,000 of the assessed value is zero-rated. The remaining assessed value up to $59,000 is taxed at 4%. Based on the subletting rate in the HDB rental market for 2010 to 2012, the assessed value of the flats do not exceed $59,000 and are therefore taxed at 4%. The calculated average monthly property taxes are $42, $57, and $66 for three-room, four-room, and five-room, respectively.Footnote ²⁵ Moreover, data on house maintenance monthly expenditure for different housing types and for different household types including general and retirees are obtained from DOS (2014c). We use this data to proxy the home maintenance expense prior retirement and during retirement. These expenses are given in columns 5 and 6 in Table 9.

Table 9. Net imputed rent pre- and post- retirements by flat type (in 2012 dollars)

Source: Authors’ calculations.

Note: ¹Assuming a two-person household.

Mortgage payments consist of the principal and interest payments. The principal is considered as savings and is excluded from net imputed rent. Interest payment is part of the cost of homeownership. However, since interest payments stop when the mortgage is paid off, these are not expenses that need to be funded in retirement. In the base model, the interest payments are amortized over 30 years using a housing mortgage rate of 2.6%. We discount the present value of all interest payments at the time of home purchase at age 30 and then compute the average payments over the loan period. Finally, we convert the average interest payments to 2012 dollars, the time when the workers enter the workforce. As reported in column 4 of Table 9, the computed present values of the average monthly interest payments under the base case housing consumption are $134, $241, and $308 for 30P, 50P, and 70P, respectively.

Pre- and post-retirement net imputed rents are tabulated in the last two columns of Table 9. Pre-retirement net imputed rent is obtained by deducting property tax, house maintenance, and interest mortgage payment from gross rent. For post-retirement imputed net rent we deduct property tax, home maintenance but not interest mortgage expenses. Hence, net imputed rent prior to retirement is smaller than post-retirement net imputed rent. The average net replacement rate is obtained by adding post-retirement net imputed rents to the average annuity payouts in the numerator, and adding pre-retirement net imputed rent to the average net lifetime income in the denominator. Table 10 reports the adjusted average net replacement rates using pre- and post-retirement net imputed rents under the Standard and Basic Plans for male and female workers. As can be observed, including net imputed rent in the calculation raises the average net replacement rate under both plans. For example, under the Standard Plan, the increases of average net replacement rates are 12.4, 12.6, and 10.1 percentage points for male workers in 30P, 50P, and 70P; and the increases are 14.8, 14.8, and 12.2 percentage points for female workers in 30P, 50P, and 70P, respectively.

Table 10. Average net replacement rates using pre- and post-retirement net imputed rent

Source: Authors’ calculations.

4.1 Housing consumption

In the base case model, we assume that workers buy an HDB housing type within their financial means. Consuming bigger and more expensive house type would mean utilizing more CPF monies to pay for higher housing mortgages; thus leaving less CPF savings for retirement. For sensitivity analysis, we consider the case when workers upsize their housing consumption and buy a flat type one size larger than in the base case. We also consider the situation when workers downsize to smaller flats.

The average net replacement rates using payouts from the Standard Plan for the male worker are presented in Figure 2.Footnote ²⁶ There is a tradeoff between housing consumption and retirement adequacy. Net replacement rates fall when workers upsize their flat type. The fall in replacement rates is more pronounced for lower income groups than for higher income groups. For example, housing loan quantum increases by about $100,000 when a median worker buys a five-room instead of a four-room HDB. With a larger loan, he would have to use the entire monthly contribution in his OA and possibly out-of-pocket cash to pay the monthly mortgage payment. There will be no accumulation in the OA and retirement savings accumulate entirely from CPF contributions in the SA. Annuitizing will then replace about 37–45% of the pre-retirement average lifetime income, depending on the types of annuity products.

Figure 2. (Colour online) Replacement rates with different housing consumption for male workers. Source: Authors’ calculations. Notes: 1. Under downsizing, 30P buys a two-room, 50P buys three-room, and 70P buys four-room. 2. Under upsizing, 30P buys four-room, 50P buys five-room, and 70P buys more epensive BTO flat which costs $550,000 (in 2012 dollars).

On the other hand, replacement rates would be higher when households buy a flat which is one size smaller. For example, under the Standard Plan, the replacement rate for male median workers increases by about 10 percentage points from 53% in the benchmark to 63% by downsizing, which is comparable with several OECD countries’ replacement rate for median earner. This implies that higher replacement rates can be achieved by lowering household's spending on housing by tweaking housing policies such as lowering housing prices or by giving larger housing grant/subsidy.

4.2 Single income earner and sole financing of housing mortgage

In the base case, we assume households to be dual wage earners. The accumulated CPF savings of both husband and wife from the time they enter the labor force are used for down-payment of the flat. Monthly mortgage payments are shared equally between the couple and are paid out from their respective CPF OAs. For sensitivity analysis, we examine the case when only the male (or female) is the sole earner. Since only his CPF savings in the OA are used for down-payment and the monthly mortgage payments, more CPF monies are utilized for housing and less savings for retirement. We run simulations assuming two different types of housing consumption. First, sole-financing but keeping the same amount of housing consumption as specified in the base case. Second, sole-financing but buying a smaller flat, which reduces housing cost by about $100,000. We assume 30P households downsize from three-room to two-room, 50P households from four room to three-room and 70P households from five-room to four-room, respectively.

Under the housing grant scheme, lower-income households will receive higher housing grants. A 30P earner, being the sole earner would have lower total income, will be eligible for higher housing grants. The larger housing subsidy and lower housing cost lead to smaller mortgage payments and higher CPF accumulation in the OA and hence higher payouts at retirement. The replacement rate is slightly lower for sole-financing with downsizing. Under the Standard Plan, the replacement rate is 57% compared with 60% under the base case with joint-financing. However, if the sole earner household does not downsize and buys the same flat type as in the base case, the replacement rate will fall to 40% under the Standard Plan as shown in Figure 3.

Figure 3. (Colour online) Replacement rates for male workers with joint-financing versus sole-financing of housing mortgage. Source: Authors’ calculations. Note: Under the downsizing scenario, 30P buys a two-room, 50P buys three-room, and 70P buys four-room.

As can be gleaned from Figure 3, similar patterns are observed for the 50P and 70P households. Single income earner household, even with downsizing, would have lower replacement rates. The fall in replacement rates are respectively, 16 and 11 percentage points for sole earners in the 50P and 70P households under the Standard Plan. The 70P worker would have to use the entire amount of the CPF savings to pay the monthly housing mortgage and hence have smaller accumulation.

Our findings show that single income families have to be more prudent in making their housing consumption decisions. However, if they wish to accumulate more housing wealth, they could consider other available monetizing options to unlock the housing wealth when they retire to enhance their retirement income. They could, for instance, sublet a room or downgrade to a studio apartment to unlock housing equity. In a survey on HDB residents’ housing aspirations and intention to move within the next 5 years, younger residents or those living in smaller flat types had greater intention to upgrade to a bigger flat. In contrast, a lower proportion of older residents had the intention to move, reflecting a greater desire to age in place. (HDB, 2014, p. 131). About 80% of elderly households indicated that they would want to live in their existing flat. The survey also shows that if older households have the intention to move, about 26% of these ones planned to either rent a room/housing unit or live in family members’/friends’ place. Presumably, these households could then monetize their flats for retirement needs and pay reduced rents or have free accommodation.