Household Saving Behaviour in New Zealand: Why do Cohorts Behave Differently?

Scobie, Grant M; Gibson, John K Working Paper Household Saving Behaviour in New Zealand: Why do Cohorts Behave Differently? New Zealand Treasury Working Paper, No. 03/32 Provided in Cooperation with: The Treasury, New Zealand Government Suggested Citation: Scobie, Grant M; Gibson, John K (2003) : Household Saving Behaviour in New Zealand: Why do Cohorts Behave Differently?, New Zealand Treasury Working Paper, No. 03/32, New Zealand Government, The Treasury, Wellington This Version is available at: https://hdl.handle.net/10419/205537 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/4.0/ Household Saving Behaviour in New Zealand: Why do Cohorts Behave Differently? Grant M Scobie and John K Gibson N EW Z EALAND T REASURY W ORKING P APER 03/32 D ECEMBER 2003 563411 NZ TREASURY WORKING PAPER 03/32 Household Saving Behaviour in New Zealand: Why do Cohorts Behave Differently? MONTH / YEAR December 2003 AUTHORS Grant M Scobie New Zealand Treasury PO Box 3724 Wellington New Zealand Email Telephone Fax [email protected] 64 4 471 5005 64 4 473-0982 John Gibson University of Waikato Private Bag Hamilton New Zealand Email Telephone Fax [email protected] 64 7 838-4466 64 7 838-4331 ACKNOWLEDGEMENTS The authors thank Ivan Tuckwell for his help in preparing the data files and participants in seminars at the University of Waikato and the Treasury. In particular, Bob Buckle and John Creedy have made helpful suggestions. NZ TREASURY New Zealand Treasury PO Box 3724 Wellington 6000 NEW ZEALAND Email Telephone Website [email protected] 64-4-472 2733 www.treasury.govt.nz DISCLAIMER The views expressed in this Working Paper are those of the author(s) and do not necessarily reflect the views of the New Zealand Treasury. The paper is presented not as policy, but with a view to inform and stimulate wider debate. WP 03/32 | HOUSEHOLD SAVING BEHAVIOUR IN NZ: WHY DO COHORTS BEHAVE DIFFERENTLY I Abstract The aim of this paper is to add to the understanding of saving decisions by households. The saving behaviour of households is found to differ depending on the birth cohort of the household head. This paper seeks to explain why this pattern might exist. It is based on an analysis of synthetic cohorts derived from unit record data taken from the Household Economic Survey (HES) for the March years 1984 to 1998. The need to use synthetic cohorts arises as the HES is not a longitudinal panel survey, but rather a time series of independent cross-sectional samples. We use a range of regression models to separate out the effect of age, birth-year cohort and year on saving rates. The typical saving rates for the cohorts born between 1920 and 1939 are found to be significantly lower relative to the younger and older cohorts studied. This pattern of cohort effects is robust to the inclusion of conditioning variables; to the trimming from the sample of households with either negative or very large ratios of savings to consumption, and to different definitions of saving. Some exploratory investigation supports the hypothesis that changes in the economic and policy environment help explain the different saving behaviour of different birth cohorts. Tentative results suggest that more “favourable environments” are associated with lower rates of lifetime saving. JEL CLASSIFICATION E21 Consumption; Saving J26 Retirement KEYWORDS Household saving rates; cohort effects; New Zealand; economic and social policies WP 03/32 | HOUSEHOLD SAVING BEHAVIOUR IN NZ: WHY DO COHORTS BEHAVE DIFFERENTLY II Table of Contents Abstract ...............................................................................................................................i Table of Contents ..............................................................................................................ii List of Tables......................................................................................................................ii List of Figures....................................................................................................................ii 1 Introduction ..............................................................................................................1 2 Construction of the data and the cohort approach...............................................2 2.1 The Data 2.2 The Use of Synthetic Cohorts.........................................................................................4 3 The Model and Results ............................................................................................9 3.1 The Basic Model .............................................................................................................9 3.2 The Initial Results .........................................................................................................10 3.3 Adding Conditioning Variables .....................................................................................12 3.4 Alternative Definitions of Saving...................................................................................16 4 Exploring the Cohort Patterns ..............................................................................16 5 Conclusions............................................................................................................25 References .......................................................................................................................28 List of Tables Table 1 – Sample size and saving rates by survey year.....................................................................4 Table 2 – Cohort definitions, cell sizes and saving rates....................................................................6 Table 3 — Mean and median saving rates, averages over overlapping ages....................................7 Table 4 – Cohort effects in individual saving rates, controlling for age and year effects..................11 List of Figures Figure 1 – Household Saving Rates by Five-Year Birth Cohort .........................................................8 Figure 2 – Cohort Effects with Different Sets of Controls .................................................................14 Figure 3 – Smoothed mean savings rate by Cohort .........................................................................14 Figure 4 – Smoothed median savings rate by Cohort: quantile regression......................................15 Figure 5 – Cohort effects with different estimation samples .............................................................16 Figure 6 – Cohort effects with different definitions of consumption and saving................................17 Figure 7Union density and membership in New Zealand: 1936-1999 ...........................................21 Figure 8 – Real Payments under New Zealand Superannuation: 1970-2001..................................23 WP 03/32 | HOUSEHOLD SAVING BEHAVIOUR IN NZ: WHY DO COHORTS BEHAVE DIFFERENTLY 1 Household Saving Behaviour in New Zealand: Why do Cohorts Behave Differently? 1 Introduction This paper has two major objectives. The first is to present estimates of the household saving patterns for different aged cohorts. The second is to offer an explanation of why saving behaviour might be different for different aged cohorts. It is extremely difficult to glean the implications of saving for say retirement income from aggregate data on the household sector. Aside from difficulties of measurement, a low overall level of saving in ageing population could be consistent with high saving by those in their working years offset by dissaving among an expanding older population of retirees. In short, a better understanding of saving by households requires an analysis of micro data based on individual household records. This study uses individual records from the Household Expenditure Survey (HES) for a 15-year period to construct synthetic cohorts (Section 2). Regression models estimate the average saving rates for each five-year birth cohort, after allowing for age and year effects together with a set of conditioning variables (Section 3). Saving rates are found to differ markedly across cohorts. Section 4 presents some tentative findings, which suggest that household saving behaviour may well be influenced by economic conditions and social policies. It is argued that the cohort patterns of saving may reflect different conditions which faced different cohorts as they moved through their working ages, especially those existing during their peak saving years. Conclusions follow in Section 5. The study finds that different cohorts do display different saving patterns. Those born from 1920 to 1939 are found to have significantly lower saving rates than older or younger cohorts. The paper finds that these differences are consistent with the fact that each cohort faced a different set of economic and social policies. The environment that prevailed especially during peak earning and saving periods was different for the different cohorts. In particular a more “favourable” environment that prevailed in the period 19501980 seems to explain why certain cohorts had lower saving rates. The implication is that extent of public provision of social welfare and retirement benefits together with conditions in labour markets, do influence the rate at which households will save. WP 03/32 | HOUSEHOLD SAVING BEHAVIOUR IN NZ: WHY DO COHORTS BEHAVE DIFFERENTLY 2 2 Construction of the data and the cohort approach In this section we first describe the data, and how we have constructed measures of saving and synthetic cohorts. We then present a first glimpse of household saving patterns based on these cohorts. 2.1 The Data We have used the income and expenditure data from the HES to estimate saving as a residual. We do this fully cognizant of the limitations of the HES.1 Our defence rests largely on the fact that there is no other source of micro-data for examining household saving behaviour. Moreover, analysts in other countries have used similar data sources, particularly the Family Expenditure Survey in the United Kingdom (Attanasio and Banks, 1998) and the Consumer Expenditure Survey in the United States (Attanasio, 1998). We have tried to eliminate some outliers, and we have a large sample, which might arguably compensate for the underlying deficiencies. But we accept that our results are only as good as the survey data from which they are derived. The definition of consumption that we have used when deriving the saving rate excludes items that are more properly considered as forms of investment and hence are a type of saving. In particular, to obtain the estimate of “current” consumption expenditure we removed from HES total expenditure, expenses on education, life and health insurance, purchases of durable goods, medical expenses, repayments of mortgage principal, and contributions to savings. In other words we attempt to construct a measure of expenditure, which would result in an "economic” view of saving.2 Consequently, our consumption and saving variables differ from those that may have been available for previous studies and from the definitions used for national aggregates. The data cover the years 1983-48 to 1997-98. We refer to these years by the latter year; ie, 1984 and 1998. A total sample of 50,624 households was available over these 15 years. We have removed all observations where household disposable income was reported as negative (some 330 households)3. Further we have truncated the sample to remove all observations where the age of the household head was reported as less than 19 or greater than 74 at the time of the survey. This left us with a sample of 46,269 households. 1 "For several reasons, care is required in making comparisons of expenditure with income from the Household Economic Survey, as the method of surveying income and expenditure does not provide for consistency at an individual respondent level....Consequently, comparisons of total expenditure against total income are not valid at the household level. It follows that any comparisons of average expenditure statistics against average income statistics for groups of households, to estimate savings, for example, could lead to spurious results". Background Notes to the Household Economic Survey, Statistics New Zealand (1998), p.17. 2 For details of the adjustments see Gibson and Scobie (2001). 3 We recognise that these households might include some self-employed unincorporated businesses, whose business expenditures result in negative reported incomes. WP 03/32 | HOUSEHOLD SAVING BEHAVIOUR IN NZ: WHY DO COHORTS BEHAVE DIFFERENTLY 3 Our justification for restricting the sample to the age range 19-74 is that those less than 19 were not considered important for studying lifetime saving patterns, while amongst the elderly, the HES does not cover institutions so those living in rest homes are not included. This means that in the upper age groups we have an incomplete sample based only on those living on their own or as part of another household and this group may not be representative of the full population of the elderly.4 Rather than taking the ratio of saving to disposable income, we have chosen to follow Attanasio (1998) and calculate saving rates by the ratio of saving to consumption. This has the advantage of being defined even when reported disposable income is zero5. Ideally, a complete accounting for income and saving requires both flow measures and a household balance sheet to track stock changes. The HES provides no measures of the stock of household wealth. As a consequence, any contributions made by an employer to a private pension fund are not recorded as saving; and any withdrawals from a pension fund are counted as income in the year received rather than “dissaving”. Fortunately, pension schemes (outside the public superannuation scheme) play a relatively minor role in New Zealand6. Should the unit of analysis it be the individual or the household? There is no clear answer to this; both have advantages and drawbacks. The HES reports income for each individual in the household, and expenditure on a household-wide basis. This means that to compute saving, one needs to either: a. allocate expenditure to individuals and then subtract from reported incomes to find individual saving levels; or b. combine the incomes of individuals to a total household income and subtract reported expenditure. We have chosen the second option, believing that many saving decisions are taken on a household basis, and considering that allocating expenditure to individuals would have created some spurious saving estimates, especially for those household members who are not participating in the labour force. It must however be recognised that in multigenerational households the saving of working age members could be offset by the dissaving of younger and elderly members of the same household. The net saving rate in such a household could then differ quite significantly from the saving rates of individual members. We have defined the age, gender, labour market status and ethnicity of the household based on the reported characteristics of the head. However we also report results based on household shares (eg the share who are working, who are male, etc). 4 Another reason for eliminating the oldest households is that pension income may not be distinguished from other income. Failure to recognise the running down of the underlying pension assets will lead overstating saving by the elderly. See Deaton and Paxson (2000), who report a more hump shaped pattern of saving with respect to age when flows into and from pension funds are included. 5 Denote saving by S, consumption expenditure by X and disposable income by YD. Then: S/YD =(S/X) .(X/YD); ie, the ratio of saving to income is a monotonic transform of the ratio of saving to consumption and the saving ratios reported in this study can be converted by multiplying by the average propensity to consume. 6 In 1997-98, 6 percent of all households received income from a private pension, and this accounted for just over 1 percent of their gross income. A little under one half of the household receiving a pension had a head aged 65-74. Among this group, pensions made up 8.5 percent of gross income. 76 percent of these households had no pension income. WP 03/32 | HOUSEHOLD SAVING BEHAVIOUR IN NZ: WHY DO COHORTS BEHAVE DIFFERENTLY 4 Table 1 reports the number of observations in each year together with the saving rates at the mean, median, 25th and 75th percentiles. The final column reports the ratio of the averages of saving to consumption, as distinct from the average (or quantile) of the ratios. The mean is clearly influenced by extreme values, so frequently it will be helpful to focus on the median saving rate. Table 1 – Sample size and saving rates by survey year Savings Rate (S/X) Total Sample Mean 25th Percentile Median 75th Percentile X S 1984 3331 0.376 -0.058 0.227 0.613 0.189 1985 3295 0.287 -0.106 0.168 0.498 0.139 1986 3174 0.318 -0.078 0.201 0.551 0.177 1987 3210 0.341 -0.084 0.209 0.581 0.204 1988 4021 0.347 -0.043 0.212 0.563 0.200 1989 3142 0.358 -0.080 0.207 0.601 0.200 1990 3047 0.313 -0.110 0.188 0.560 0.166 1991 2674 0.340 -0.099 0.193 0.575 0.227 1992 2712 0.380 -0.062 0.217 0.609 0.251 1993 4244 0.415 -0.057 0.222 0.621 0.270 1994 2839 0.338 -0.096 0.197 0.546 0.235 1995 2695 0.336 -0.110 0.191 0.607 0.234 1996 2621 0.355 -0.111 0.186 0.572 0.246 1997 2642 0.359 -0.091 0.193 0.578 0.259 1998 2622 0.428 -0.081 0.238 0.676 0.320 Total 46269 0.353 -0.086 0.202 0.584 0.222 It should be stressed that there is enormous underlying variability in the data. Household savings vary between -$1.56 million and +$0.78m. Among any one group with the same rate of saving (e.g. 0.20 to 0.29) consumption expenditures vary from $2,950 to $125,474, and their absolute level of saving varies from $879 to $32,241. Nearly 32 percent of the sample report negative saving. Inevitably some of this may be due to under-reporting of income. Evidence from the USA suggests that when the differential under-reporting of both income and consumption is allowed for, up to one third of the apparent fall in the saving rate between 1972-73 and 1983-84 may be due to misreporting (Bosworth, Burtless and Sabelhaus (1991)). 2.2 The Use of Synthetic Cohorts To study the lifecycle profiles of saving we would ideally have panel data, where the same people are tracked over time. However, the available panel surveys in New Zealand are restricted to cohorts of young people who were born in the 1970s, and so are unsuitable for studying lifecycle phenomena.7 But the availability of a time-series of cross-sectional Household Economic Surveys allows us to construct synthetic panels following methods described by Shorrocks (1975) and Deaton (1985) 7 The “snapshot” offered by a single cross-section is also unsuitable for observing life-cycle patterns because although a variety of ages are observed in a cross-section, they also represent different birth cohorts. If there are strong cohort effects, a cross-section age profile may be very different from the age profile of any individual, as noted by Shorrocks (1975). WP 03/32 | HOUSEHOLD SAVING BEHAVIOUR IN NZ: WHY DO COHORTS BEHAVE DIFFERENTLY 11 Table 4 – Cohort effects in individual saving rates, controlling for age and year effects Mean 25th Percentile Median 75th Percentile Cohort 2 (b. 1915-19) -0.045 -0.007 -0.036 -0.136 (1.06) (0.25) (1.14) (3.05)** Cohort 3 (b. 1920-24) -0.063 -0.014 -0.051 -0.165 (1.41) (0.47) (1.51) (3.42)** Cohort 4 (b. 1925-29) -0.122 -0.030 -0.084 -0.225 (2.55)* (0.90) (2.30)* (4.31)** Cohort 5 (b. 1930-34) -0.142 -0.046 -0.106 -0.279 (2.65)** (1.24) (2.62)** (4.84)** Cohort 6 (b. 1935-39) -0.066 -0.033 -0.081 -0.190 (1.05) (0.83) (1.85)+ (3.03)** Cohort 7 (b. 1940-44) 0.054 -0.021 -0.051 -0.074 (0.72) (0.48) (1.07) (1.10) Cohort 8 (b. 1945-49) 0.106 0.002 -0.004 0.003 (1.42) (0.05) (0.08) (0.04) Cohort 9 (b. 1950-54) 0.168 -0.017 0.020 0.102 (2.08)* (0.35) (0.37) (1.31) Cohort 10 (b. 1955-59) 0.198 0.003 0.029 0.121 (2.32)* (0.05) (0.50) (1.48) Cohort 11 (b. 1960-64) 0.223 0.011 0.040 0.122 (2.48)* (0.20) (0.66) (1.41) Cohort 12 (b. 1965-69) 0.245 0.019 0.054 0.178 (2.61)** (0.32) (0.84) (1.93)+ Cohort 13 (b. 1970-74) 0.281 0.035 0.074 0.250 (2.89)** (0.55) (1.08) (2.56)* Cohort 14 (b. 1975-79) 0.263 -0.081 -0.013 0.211 (2.39)* (1.12) (0.16) (1.87)+ Age 1.332 0.759 0.998 1.644 (8.61)** (7.32)** (8.78)** (10.04)** Age2 -0.067 -0.038 -0.051 -0.083 (8.90)** (7.64)** (9.29)** (10.52)** Age3 0.002 0.001 0.001 0.002 (9.18)** (7.95)** (9.76)** (10.93)** Age4 0.000 0.000 0.000 0.000 (9.35)** (8.21)** (10.09)** (11.16)** Age5 0.000 0.000 0.000 0.000 (9.40)** (8.41)** (10.29)** (11.21)** Constant -10.098 -5.910 -7.390 -12.190 (8.18)** (7.10)** (8.11)** (9.26)** R2 0.0135 0.0042 0.0064 0.0144 Cohort effects = 0 P < 0.000 P < 0.003 P < 0.000 P < 0.000 Age effects = 0 P < 0.000 P < 0.000 P < 0.000 P < 0.000 Year effects = 0 P < 0.001 P < 0.000 P < 0.005 P < 0.000 Note: Coefficients weighted by population sampling weights. Absolute value of robust t-statistics in parentheses; + significant at 10% level * significant at 5% level; ** significant at 1% level. The sample has N=46269 observations. Each regression also includes 13 time dummies, whose coefficients are constrained to sum up to zero and to be orthogonal to a linear trend. The cohort effects in the mean saving rate are reported in the first column of Table 4 and follow a somewhat ‘V’ shaped pattern. Relative to the reference group, which is households headed by someone born in 1910-14, saving rates fall across later born cohorts, reaching their lowest point for households headed by someone from the 1930-34 birth cohort, where the saving rate is 14 percentage points below the reference group. Thereafter, the mean saving rate increases monotonically across the more recent cohorts, until it peaks amongst those households headed by someone from the 1970-74 birth cohort, where it is 28 percentage points above the reference group. This result carries the possible implication that downward trends in aggregate saving rates might be temporary, WP 03/32 | HOUSEHOLD SAVING BEHAVIOUR IN NZ: WHY DO COHORTS BEHAVE DIFFERENTLY 12 as middle-aged cohorts with low saving rates will eventually be replaced by younger cohorts with higher saving rates. However, there are fewer grounds for such optimism when considering median saving rates, which exhibit the same ‘dip’ for cohorts born ca. 1925-39 but do not show any statistically significant rise in saving rates across the more recently born cohorts. It appears that the results for mean saving rates are being caused mainly by the behaviour of households in the upper end of the distribution; at the 25th percentile there are no significant cohort effects, whereas at the 75th percentile the ‘V’ shape is accentuated. Amongst these households with high saving rates, the results in the final column of Table 4 show that the saving rate for the 1930-34 birth cohort is 28 percent lower than for the 1910-14 cohort. While the cohort effect appears to be largely restricted to the upper end of the distribution, it must be stressed that these are the households who contribute the bulk of aggregate saving.18 It is therefore, important to describe and understand these cohort effects if one is to make any inferences about the future path of aggregate savings. 3.3 Adding Conditioning Variables 19 To check whether within-cell heterogeneity can explain the marked pattern of cohort effects, the regression models are augmented with various conditioning variables, controlling for demographics, education, employment, family structure and dwelling tenure. For example, one possible cause of the cohort effects in Table 4 is that there are differences in family structure across birth-year cohorts due, say, to the impact of changing social conditions and welfare policies on the prevalence of sole parenthood. By checking to see if the pattern of cohort effects changes when these conditioning variables are introduced, we test if the shifts in lifecycle saving profiles can be explained by these demographic, education and family structure effects, rather than by pure cohort effects. Some of the conditioning variables, such as employment status, are likely to change over the lifecycle, whereas others, such as ethnicity and gender obviously remain fixed. In both cases, the conditioning variables are allowed to shift the intercept of the estimated age profile of saving but because of the small sample sizes we do not consider interaction effects where the shape of the age profile can differ between, say, education groups20. Therefore, the specification of the regression models is: ( ) ch t ch t hhch tuwcafs + ′ += β , (2) where: ch t s = the saving rate for household h, observed in year t and belonging to (five-year) birth-cohort c; ch t w = a vector of conditioning variables; and ch t u = the residual term. 18 Over the fifteen-year sample period, total household savings were estimated at $93bn of which $87bn was accounted for by the households whose disposable incomes fell in the top three deciles on the income distribution. 19 This section presents only abbreviated results of the findings of the robustness of the basic model as different sets of conditioning variables were added. A complete set is available from the authors on request. 20 See Attanasio (1998) for an example of interacting education with cohort dummies. WP 03/32 | HOUSEHOLD SAVING BEHAVIOUR IN NZ: WHY DO COHORTS BEHAVE DIFFERENTLY 13 When the gender and ethnicity21 of the household head are included, there is a strong effect of gender, with male-headed households having saving rates approximately nine percentage points higher; but there is no apparent effect of ethnicity on savings. The addition of these demographic controls did not change the basic pattern of cohort intercepts. As in the previous results (Table 4), households whose head was born ca. 1925-1939 show lower than average saving rates, while the rise in savings rates for the more recently born cohorts is even more apparent than when the demographic controls were absent. The next model included variables for whether the household head is either employed or unemployed, and another variable for whether the head receives self-employment income. In comparison with the reference category, which is households whose head is out of the labour force, average saving rates are seven percent lower if the head is unemployed and 17 percent higher if the head is working. The saving rate appears about 13 percent higher when the household head receives self-employment income. The difference in income levels between the self-employed and other households may be too small to explain this large jump in saving rates,22 so it may be evidence for theoretical arguments that uninsurable income risk, which is likely to be greater for the self-employed, raises the level of wealth accumulation (Caballero, 1991). The addition of these three employment variables reinforces the basic cohort pattern in saving rates that was reported in Table 4, with higher saving rates amongst the later born cohorts and lower saving rates amongst the households whose head was born ca. 19201939. Adding the employment variables also affects the results for the other demographic controls, halving the coefficient on gender and producing a significantly positive coefficient on ethnicity. Hence, the lower saving rates of female-headed households are partly because of their lower employment rates, while households headed by Maori and Pacific Islanders would have higher than average saving rates if their household heads had employment probabilities that were the same as the rest of the population.23 Figure 2 plots the cohort intercepts estimated at the mean, median, 25th and 75th percentiles of the distribution of saving rates, along with the intercepts from the models that include conditioning variables. It is evident that the introduction of controls for withincell heterogeneity does not greatly modify the relative magnitude of the cohort dummies, tending to cause variation only for the most recently born cohorts. There is rather more variation in the patterns of cohort effects estimated at different points in the distribution, so we return to that point below, in considering whether the results are robust to increasingly severe trimming of outliers from the estimation sample. The other notable feature of Figure 2 is its similarity to results reported by Attanasio (1998, Figure 9). For the U.S., Attanasio finds that the household saving rate falls for the first four five-year birth-cohorts from 1910-14 to 1925-29, and then rises for each of the younger cohorts. With the exception of the later dating of the turning point in New Zealand (the 1930-34 cohort) and the inclusion of cohorts born post-1959, the patterns in the two countries are strikingly similar. 21 Both these are fixed over the lifecycle. 22 Using all 15 surveys from 1984-98, the average disposable income of households headed by someone receiving self-employment income was $41,900 (in December 1993 prices), while the average for other households where the head is employed is $39,600. 23 One hypothesis, untested in this study, is that because of lower accumulated wealth and more erratic employment history, Maori and Pacific Island households do not enjoy the same access to credit, inducing a higher level of savings, other factors constant. WP 03/32 | HOUSEHOLD SAVING BEHAVIOUR IN NZ: WHY DO COHORTS BEHAVE DIFFERENTLY 14 Figure 2 – Cohort Effects with Different Sets of Controls -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 1910-14 1915-19 1920-24 1925-29 1930-34 1935-39 1940-44 1945-49 1950-54 1955-59 1960-64 1965-69 1970-74 1975-79 Cohort (Birth Year) Mean 25%ile Median 75%ile Mean + demographic controls Adding employment controls Adding dwelling tenure dummies When the basic model of five-year birth-cohorts, is augmented with demographic, employment, family type and tenure status variables, we obtain a reasonably robust description of the underlying data. We therefore use the predictions from this model in to illustrate the shape of the age profile in median household saving rates and to show how that profile has shifted up and down across birth cohorts (Figure 3). Figure 3 – Smoothed mean savings rate by Cohort Mean (S/X) Age of Household Head 20 40 60 80 -.2 0 .2 .4 The typical age profile for the average saving rate is somewhat hump shaped with a peak around age 57 but does not become negative at older ages. While the hump shape is consistent with the lifecycle hypothesis, the apparent increase in the saving rate beyond WP 03/32 | HOUSEHOLD SAVING BEHAVIOUR IN NZ: WHY DO COHORTS BEHAVE DIFFERENTLY 15 age 70 is not.24 The downward shift in the saving profile for earlier born cohorts, up until the fifth oldest one, is also apparent (earlier born cohorts are shown by the start and end points for their graph occurring at an older age, which is the age at the time of the first survey in 1984). In Figure 2 the pattern of cohort effects is more apparent at the mean and 75th percentile of the distribution of saving rates than it is at median and 25th percentile. The quantile regression at the median is based on least absolute deviations of the residuals, rather than least squares, and so is less sensitive to the presence of outliers. To investigate whether the pattern of cohort effects is just due to some of the extreme values of saving rates that are present in the data various “trimmed samples” that removed extreme values of saving rates were used. In all cases, the models include the demographic, employment, family type and tenure status variables. The results again showed that the median saving rate falls from the earliest born cohorts until those born in 1930-34 and then rises across the cohorts born in later years. Hence, the pattern is the same as for the mean saving rates, although the rise in saving rates for the most recently born cohorts is not as marked. The predictions from this quantile regression model give the smoothed median saving rates in Figure 4. The pattern is similar to that for the mean saving rate, except that median saving rates are everywhere lower so that there is negative saving at the start of the lifecycle and around age 65, and the downward shifts in the profile when moving from later to earlier birth cohorts are smaller. Figure 4 – Smoothed median savings rate by Cohort: quantile regression Figure 6.3: Smoothed Savings Rate by Cohort: Quantile Regression Median (S/X) Age of HH Head 20 40 60 80 -.1 0 .1 .2 .3 Figure 5 plots the cohort intercepts for each of the models estimated with reduced samples and for the model of mean saving rates estimated on the full sample. Although the magnitude of the cohort effects vary as the sample or estimation method is altered, the relative ranking of cohorts does not change. In all cases, saving rates fall from earlier to later born cohorts between the 1910-14 and 1930-34 cohorts and then the pattern 24 The use of a fifth-order polynomial does mean that the smoothed saving rate will exhibit four turning points but the predicted rise in the saving rate beyond age 70 does not appear to be a result of over-fitting the data. See Gibson and Scobie (2001) where this same feature was evident in the unrestricted estimates of the mean saving rate. WP 03/32 | HOUSEHOLD SAVING BEHAVIOUR IN NZ: WHY DO COHORTS BEHAVE DIFFERENTLY 16 reverses with later born cohorts exhibiting higher saving rates. These patterns seem sufficiently robust to warrant further investigation. Figure 5 – Cohort effects with different estimation samples -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5 1910-14 1915-19 1920-24 1925-29 1930-34 1935-39 1940-44 1945-49 1950-54 1955-59 1960-64 1965-69 1970-74 1975-79 Cohort (Birth Year) Mean Median (S/X)<10 (S/X)<5 (S/X)<2 (S/X)>=0 3.4 Alternative Definitions of Saving As noted in the introduction, consumption expenditures were adjusted to remove those items reasonably viewed as “investment” so as to give a truer picture of the underlying saving rate. In this section, we assess whether the cohort effects that we have found previously are sensitive to the reinclusion of some of these items in the household consumption variable. The cohort effects estimated under these different definitions of consumption and saving are plotted in Figure 6. These graphs illustrate the robustness of the relative cohort effects and especially the location of lower saving rates amongst those born ca. 19251939. The pattern of cohort effects in Figure 6 is consistent with our other sensitivity checks, including trimming the sample (Figure 5) and controlling for within-cell heterogeneity (Figure 2). Hence, we are confident that this lower saving rate for those born ca. 1925-1939 is a genuine feature of the saving behaviour of New Zealand households rather than just some artefact of the data or of our econometric procedures. The remaining task is to explain this cohort pattern of saving rates. 4 Exploring the Cohort Patterns Section 3 reported on the pattern of coefficients for the birth year cohorts. A range of these estimates for different quantiles and with different sets of controls was reported in Figure 2. They display a distinct V-shaped pattern. In short, saving rates appeared to differ significantly for different cohorts. WP 03/32 | HOUSEHOLD SAVING BEHAVIOUR IN NZ: WHY DO COHORTS BEHAVE DIFFERENTLY 17 Figure 6 – Cohort effects with different definitions of consumption and saving -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 1910-14 1915-19 1920-24 1925-29 1930-34 1935-39 1940-44 1945-49 1950-54 1955-59 1960-64 1965-69 1970-74 1975-79 Cohort (Birth Year) Current consumption Adding medical, education, insurance Adding durables Adding mortgage and contributions Total Consumption The cohort coefficients are estimated with reference to the 1910-14 birth cohort. Those coefficients for cohorts 3, 4 and 5 are all significantly lower than the reference cohort. In contrast, the coefficients for cohorts 9 through 14 are all significantly greater. In other words, those born from 1920 to the mid-1930s have demonstrably lower lifetime saving rates, while those born after 1950 have significantly higher rates of saving. At first glance this result may seem surprising. Anecdotal evidence might have suggested that those born in the early inter-war period would have been conditioned by wars and the Great Depression, which could have lead to higher saving rates, at the least that part of saving driven by a precautionary motive. In contrast, the post WWII cohorts facing greater economic growth and security, together with liberalised financial markets after 1986, might have been expected to have displayed greater profligacy, and have had lower, not higher, rates of saving. While these effects may be responsible for some influence on the estimated coefficients, clearly some other forces have operated to override them and produce a ‘V’ rather than an ‘inverted-V’ shaped pattern of saving by birth cohorts. It is important to explore further the cohort pattern of saving. As cohorts with different patterns of saving move through their life cycle, they may influence the aggregate pattern of saving. The cohorts with significantly lower saving rates were aged between 45 and 60 in 1980, and entering their peak saving years. This is precisely the time that aggregate household saving was observed to start declining (Claus and Scobie, 2002). Attanasio (1998) finds a similar ‘V-shaped’ pattern of cohort saving behaviour. He notes that the lower saving rates of the group aged in their 40s and 50s in the 1980s “are those mainly responsible for the decline in aggregate saving…because those cohorts were in the part of their lifecycle when saving are highest” (p.604). While he adds that even in for the USA where the data are much more consistent, it is not possible to precisely match the aggregate and micro level data. Nevertheless, the estimated cohort effects “explain a substantial part of the decline in the aggregate saving rate”. WP 03/32 | HOUSEHOLD SAVING BEHAVIOUR IN NZ: WHY DO COHORTS BEHAVE DIFFERENTLY 18 Attanasio continues, noting that: “The main deficiency of the analysis is its failure to explain why those particular cohorts did not save ‘enough’. A plausible hypothesis, that I have not tested explicitly, is that the negative cohort effects for the middle cohorts are linked to increases in social security entitlements that the same cohorts have enjoyed” (p.604). Attanasio (1998) posited that more generous public pensions might have explained the different lifetime saving patterns of different cohorts. If over an individual’s lifetime there is increasingly generous provision by the state for public pensions, it seems entirely plausible that this would shape expectations about the level of the publicly subsidised pension that one might receive. Those expectations could then in turn influence an individual’s decisions about the optimal allocation of consumption over their lifetime. Knowing (or at least predicting) that there would be a generous state pension to underpin consumption levels after retirement may lead to higher consumption prior to retirement from a given level of lifetime wealth. The consequence would be that the lifetime saving rate would be lower than in the absence of the public pension scheme, or with a less generous scheme. If this were the case one would expect to see lower saving rates among those cohorts whose expectations were for the receipt of a more generous pension and higher rates among those who expected to receive lower real pension payments. The impact of the social and economic environment over the lifetime of an individual (household) could be called the direct effect. There is also the indirect effect transmitted through family and those close to a person. These are the values and norms that are transmitted to them by their parents and others, and which in turn were formed by the conditions prevailing in an earlier time and shaping the values of their parents. In that sense, the behaviour patterns observed at any one time are a function of the prevailing climate, the expectations of the future climate, together with the effects of all previous environments. Those in the most immediate past could be expected to have the greatest effect, with the impact tailing off the further back we go25. We have not tried to explicitly allow for these indirect influences of past conditions, which might shape the behaviour of a particular cohort. To proceed further with this hypothesis, it is necessary to posit some mechanism of how and when expectations are formed. Clearly this is a complex process, and one that would reflect the person’s perception of their economic and social environment. In addition, the experiences of their parents and that of their childhood could well condition their perceptions and the need for saving. The environment prevailing during their working lives will affect their labour market experience and earnings (the ability to save) while the provision of social services (health, education, housing, pensions) and welfare (sickness, disability, unemployment, family and single mother benefits) will influence the need to save. Other factors including capital gains on housing, real interest rates, rates of income growth, access to credit and life expectancy could all be expected to impinge on the decision to allocate income to current or future consumption. In addition, the desire to make bequests may also influence the saving rate. 25 Counter examples can be found where some traditions are handed on virtually unchanged through many generations. WP 03/32 | HOUSEHOLD SAVING BEHAVIOUR IN NZ: WHY DO COHORTS BEHAVE DIFFERENTLY 19 In short there are many possible indicators that might affect the decisions of individuals with regard to their saving rate. Some of these will operate throughout their working lives. Given that the peak saving years are typically between ages 45 and 60, an individual’s perceptions and expectations during this period, would arguably have a significant influence on their saving behaviour. Essentially there are two steps in the argument: the first, that different cohorts operated in different environments; and the second that these different environments shaped the responses of different cohorts, particularly in the present case, with respect to their household saving behaviour26. Thomson (1991) documents changes that would support the first of these steps. He argues that over “the last 50 years welfare states have been very uneven in the benefits they provide for successive generations, that is for people born in different decades” (p.1). “The prizes and penalties of living in a welfare state are distributed more on the basis of birth date than of need, justice or desert. In New Zealand the big winners in this have been the ‘welfare generation’ – those born between about 1920 and 1945” (1991, p.1). Testing these hypotheses is clearly a challenge on at least four counts. First we have little theoretical guidance as to how expectations are formed; in particular what relative weight should be assigned to each of the three critical periods: the experience of the previous generations particularly parents and grandparents which through norms and values could be expected to shape the saving behaviour of a particular cohort; the conditions prevailing during their working life and in particular applying during the peak saving years; and their expectations throughout their working lives about the level and eligibility for state support of retirement income. Second, we need long time series, arguably covering the last 70 to 100 years to provide a quantitative assessment of the different policy environments enjoyed by different cohorts. Third, we only have 14 observations of the “dependent” variable. ie, the cohort dummies that relate lifetime saving behaviour to year of birth27, meaning the scope for any statistical tests is limited; and fourth, as the working lives and saving periods of the later cohorts extend into the future, some forecasts of future conditions will be involved in order to compare their behaviour with that of individuals from the older cohorts who are either retired or dead. In what follows we “test” the hypotheses in a very elementary way. Basically we look at a series of indicators for which we can obtain at least partial data. Often we focus only on selected years or periods that might be “typical or representative”, or occur at that time of peak saving28. In effect we are conducting in a loose manner a non-parametric sign test as an initial step. Are changes in the indicators broadly consistent with the hypothesis that the savings patterns of different birth cohorts could have been influenced by the economic and social policy environment prevailing at key points in the lifetimes? The objective is to make a preliminary foray to establish whether the patterns of some key variables that arguably affect the saving decisions of individuals are consistent with the 26 We refer to these effects as cohort effects for convenience. In fact it possible that the effect is really a “time” effect, so that in the absence of certain changes both younger and older cohorts would have behaved similarily. We are grateful to John Creedy for pointing out this difference. 27 It is true that we could estimate the saving rate models with many more cohort dummies; in fact, potentially one for each birth year of all the individuals (or household heads) in the sample, as we have done in Sections 4 and 5. This would span some 80 years. However these estimates would tend to be noisy; making it difficult to estimate relationships with the policy variables, which show much less year-to-year variance. 28 Clearly this approach could be enhanced with more continuous time-series data from the 1930s to the present, but that is not an insignificant task, and one we assign to the category of “future research directions”. WP 03/32 | HOUSEHOLD SAVING BEHAVIOUR IN NZ: WHY DO COHORTS BEHAVE DIFFERENTLY 20 hypothesis that the cohort differences reflect the external environment. In particular, we examine both labour market indicators (the ability to save) and some measures of public pensions and welfare (affecting the incentive to save). Table 5 summarises the working life (assumed to be 40 years from age 20) and the peak earning years (assumed to be from ages 45 to 60) for each of the key birth year cohorts selected for this section. The first cohort (1910-14) is the reference cohort in the sense that the regression coefficients for cohort presented in Section 3 are referenced to this cohort, which has a value of zero. The lifetime saving rate of the adjoining cohort (191519) is not significantly different. The next three cohorts covering birth years from 1920 to 1934 are the group that typically have demonstrated significantly lower lifetime saving rates, while the last two (covering 1950 to 1959) are representative of those showing a significantly greater level of lifetime saving rate. Table 5 – Lifetime patterns for different cohorts Cohort Number Birth years Saving Rate (a) Working life Peak Saving Years 1 1910-1914 0 (b) 1930-1974 1955-1969 2 1915-1919 0 (b) 1935-1980 1960-1974 3 !920 -1924 Negative 1940-1984 1965-1979 4 1925-1929 Negative 1945-1989 1970-1984 5 1930-1934 Negative 1950-1994 1975-1990 9 !950-1954 Positive 1970-2014 1995-2010 10 1955-1959 Positive 1975-2019 2000-2014 Notes: Refers to lifetime saving rate relative to the 1910-14 reference group (by definition zero). See pattern of cohort dummies in the regressions presented in Section 3. Negative and positive refer to the cohorts that were typically significantly lower or higher in their lifetime saving rates. In what follows we examine some selected aspects of the economic and social environment facing the different cohorts both over their working lives as a whole, and in particular during their peak saving years. The question posed is the following: do those indicators vary in a way that is consistent with the observed cohort patterns in saving rates? We would expect to find that the proxies chosen for the economic and social environment adopted values less favourable for household saving rates during the critical years of the low saving cohorts, while the same indicator should be more favourable in years corresponding to the high saving cohorts. Because of the magnitude of the task of assembling annual data on a wide range of variables in a consistent manner for 70 years, we have chosen to use selected years to illustrate the results. We focus on three cohorts: 1 (born 1910-14), 4 (born 1925-29) and 9 (born 1950-54), and will refer to these as the reference, early and late cohorts. Typically we will look at the values of the indicator variable prevailing during their peak saving years (given in Table 5). We start with some key indicators relating to the labour market. The extent of unemployment is a critical factor affecting the expected flow of earnings. Those cohorts facing a lower probability of unemployment would be expected to have less incentive for precautionary saving. The unemployment rates (based on the average of the census years) facing the reference and early cohorts were 1.2 and 3.3 percent respectively, while based largely on projections the late cohort could face an average of 6 percent29. Prior to the major reforms of the late 1980’s, the New Zealand labour markets were characterised by central wage fixing, limited flexibility and a high degree of unionisation (see Figure 7). Strong national unions were able to bargain particularly with state sector 29 Unless otherwise noted, all the values of the economic and social policy variables are taken from various editions of the New Zealand Official Yearbook. WP 03/32 | HOUSEHOLD SAVING BEHAVIOUR IN NZ: WHY DO COHORTS BEHAVE DIFFERENTLY 27 aggregate trends. But until we have a reconciliation and can explain the divergent series, then this task remains in the category of unfinished business. While the results of the cohort saving behaviour (Section 3) seem both significant and robust, our attempts to provide an explanation (Section 4) are partial and tentative. This is a complex area; the saving rates we observe are the resolution of a set of forces encompassing social and cultural norms shaped by the experience of earlier generations, economic conditions over the working life, expectations of future incomes, health status and life expectancy, and myriad state interventions. Arguably, the provision of higher state benefits, or more certainty would be expected to dampen the incentive for private saving.43 Our preliminary examination of some snippets of evidence is at least consistent with that argument. Further testing of this relationship awaits the development of longterm data series for at least the last one hundred years, together with richer models about how cultural norms and values together with expectations, blend to shape consumption and saving decisions. 43 Cross-sectionally countries with pay-as-you-go pension schemes funded from general or payroll taxes tend to have lower household saving rates. See Samwick (2000). WP 03/32 | HOUSEHOLD SAVING BEHAVIOUR IN NZ: WHY DO COHORTS BEHAVE DIFFERENTLY 28 References Attanasio, O., A Cohort Analysis of Saving Behaviour by U.S. Households. Journal of Human Resources 33(3): 575-609, 1998. Attanasio, O. and J. Banks., Trends in Household Saving: A Tale of Two Countries. Working Paper No 98/15, The Institute of Fiscal Studies, London, 1998. Bosworth, B., G. Burtless and J. Sabelhaus, The decline in saving: evidence from household surveys. Brookings Papers on Economic Activity 1: 183-256, 1991. Caballero, R., Earnings Uncertainty and Aggregate Wealth Accumulation. American Economic Review 76(3): 676-691, 1991. Claus, I. and G. Scobie, Saving in New Zealand: Measurement and Trends, Working Paper 02/02, New Zealand Treasury, 2002. http://www.treasury.govt.nz/workingpapers/2002/02-2.asp Deaton, A., Panel Data from Time Series of Cross Sections. Journal of Econometrics 30(1): 109-126, 1985. Deaton, A., The Analysis of Household Surveys: A Microeconomic Approach to Development Policy, Johns Hopkins University Press: Baltimore, 1997. Deaton, A. and C. Paxson, Growth and Saving Among Individuals and Households. Review of Economics and Statistics 82(2): 212-225, 2000. Feldstein, M., Social Security, Induced Retirement and Aggregate Capital Accumulation. Journal of Political Economy 82:905-926, 1974. Feldstein, M., Social Security and Savings: New Time Series Evidence. National Tax Journal 49:151-163, 1996. Gibson, J. and G. Scobie, Household Saving Behaviour in New Zealand: A Cohort Analysis. Working Paper 01/18, New Zealand Treasury, http://www.treasury.govt.nz/workingpapers/2001/01-18.asp 2001. Gokhale, J., L. Kotlikoff, and J. Sabelhaus, Understanding the Post-war Decline in U.S. Saving: A Cohort Analysis. Working Paper 5571, National Bureau of Economic Research, 1996. Jappelli, T., The Age-wealth Profile and the Life-cycle Hypothesis: A Cohort Analysis with a Time Series of Cross-sections of Italian Households. Review of Income and Wealth 45(1): 57-75, 1999. Lumsdaine, R.L. and D.A.Wise, Aging and Labor Force Participation: A Review of Trends and Explanations. Working Paper No. 3420, National Bureau of Economic Research, 1990. Meguire, P., Social Security and Personal Saving: 1971 and Beyond. Paper presented to the Annual Conference of the New Zealand Association of Economists, Christchurch, New Zealand, June 27-29, 2001. Preston, D., Retirement Income in New Zealand: the historical context. Wellington: Office of the Retirement Commissioner, 1999. WP 03/32 | HOUSEHOLD SAVING BEHAVIOUR IN NZ: WHY DO COHORTS BEHAVE DIFFERENTLY 29 Rankin, K., New Zealand’s Labour Supply in a Long Term Perspective. Proceedings of the Sesquicentennial Conference of the New Zealand Association of Economists, Auckland August 20-22 : pp. 519-539, 1990. Shorrocks, A., The Age-wealth Relationship: A Cross-section and Cohort Analysis. Review of Economics and Statistics 57(1): 155-163, 1975. Thomson, D., Selfish Generations? The Ageing of New Zealand’s Welfare State (Wellington: Bridget Williams Books), 1991. UMR Insight Ltd, New Zealand Retirement IncomeA Tracking Study. Prepared for The Super 2000Taskforce. Wellington, 1999. Venti, S. and D. Wise, The Wealth of Cohorts: Retirement Saving and the Changing Assets of Older Americans. Working Paper 5609. National Bureau of Economic Research, 1996. Verbecek, M. and T. Nijman, Can Cohort Data be Treated as Genuine Panel Data? Empirical Economics 17:9-23, 1992.