Two forms of questions were used to try to quantify the losses in savings that resulted from the decline in equity markets. The first asked respondents to describe what had happened to the value of their savings in the last three years:
Thinking of all moneys you had set aside as savings before 2000 (such as pensions, bonds, ISAs, stocks and shares) have they increased or decreased in value over the last 3 years?
Responses were qualitative, with categories: ‘increased a lot’, ‘increased a little’, ‘remained about the same’, ‘decreased a little’, and ‘decreased a lot’.
Whilst we do not know the previous exposure to equities, it seems likely that those who have suffered the greatest declines in savings are also those who had greater investments in equity products. This question will reflect both the absolute decline in savings and subjective factors, which will influence how an individual reacts to a given savings loss. This subjectivity, nevertheless, has some advantages. First, workers’ perceptions are likely to be an important
Older workers and scary markets 107
determinant of behaviour. Second, the question may capture the relative impact of the decline in savings more effectively than attempting to compare an individual’s savings to their other asset wealth – on which individuals are reticent. These subjective measures may then allow us to infer, albeit with some error, the relative declines in savings.
The responses to this question are shown in Figure 6.4. Some 48.6 per cent of individuals responded that their savings ‘declined a lot’, with 20.1 per cent reporting they ‘declined a little’. A majority then saw their savings decline over the period in question. In 11.2 per cent of cases, the value of savings has remained largely flat, whilst for 18.8 per cent there has been a small increase in savings. Only 1.3 per cent saw a large increase in savings. For those whose savings increased in value, we can hypothesize that they were either largely invested in bonds or other guaranteed-return products, or they increased their savings rate, or they may even have received an inheritance. It is important to note that we capture the change in the stock of savings, which is the relevant issue for our analysis, not the decline in moneys invested in the stock market, though these may be closely related.
A second question, more quantitative in tone, was subsequently asked.6For those individuals who had responded that their savings had increased or decreased, we then asked:
Source data: 2003 Watson Wyatt YouGov Scary Market Survey.
Figure 6.4 The change in the value of savings, 2000–2003 Increased a lot
Increasd a little About the same Decreased a little Decreased a lot
%
0 10 20 30 40 50
48.6 20.1
11.2
18.8 1.3
By approximately how much have all the moneys you had set aside as savings before 2000 increased (decreased) in the last 3 years?
Response categories were: ‘less than 5 per cent’, ‘between 5 and 10 per cent’,
‘between 11 and 25 per cent’, ‘between 26 and 50 per cent’, and ‘more than 50 per cent’. Sample responses are shown in Figure 6.5. Some 8.2 per cent report losses of greater than 50 per cent, 24.9 per cent a fall in savings between 26 per cent and 50 per cent, and 20.7 per cent losses of between 11 and 25 per cent. Around a quarter (25.7 per cent) report their savings have changed by less than 5 per cent, only 2.3 per cent respond that their savings have increased by more than 10 per cent.7
Using responses from these banded categories we can estimate the mean decline in savings for individuals with different characteristics, using the inter- val regression technique. Assuming the change in savings is normally distrib- uted, this technique maps the true change in savings on to the ordered bands described above. Estimation then maximizes the probability of observing a response within a band given the characteristics of the individual and the band cut-points.
Table 6.1 reports the estimated mean change in savings for different sub- samples of individuals. In all cases the mean change is negative. The average change in the value of savings, in the sample, is estimated to be –17.0 per cent.
Older workers and scary markets 109
Source data: 2003 Watson Wyatt YouGov Scary Market Survey.
Figure 6.5 The proportionate decline in savings, 2000–2003 (%) Fall more than 50%
Fall 26–50%
Fall 11–25%
Fall 5–10%
Fall less than 5%
About the same Rise less than 5%
Rise 5–10%
Rise 11–25%
Rise 26–50%
Rise more than 50%
% 0
8.2
24.9 20.7
11.4 3.4
12.2 10.1 6.9 1.4
0.3 0.6
5 10 15 20 25
This figure is slightly higher for men than for women, but not statistically significantly different. To be clear, this does not imply that there is no differ- ence in the amounts lost, but rather that their proportionate decline is no differ- ent. Those with less education report losing less as a proportion of their income, and as they are, on average, likely to have lower absolute amounts of savings, we can infer that absolute losses are also lower. Those in the private sector lost more than those in the public sector, and professionals lost more than non-professionals. Those who are retired reported losing more than those currently working, though this may be explained by the fact that those employed were still in the accumulation phase – and so offset losses with further contributions. In all cases the differences are statistically significant.
Those with personal pensions lost more than those with occupational pensions, both DB and DC, and this difference is statistically robust. It may be Table 6.1 Summary statistics: proportionate change in savings
Sub-sample Mean Standard deviation
All –17.04 (20.94)
Male –17.24 (20.91)
Female –16.47 (20.99)
Low educationa –13.54 (19.69)
Medium educationb –16.12 (21.12)
High educationc –19.10 (21.11)
Main pension: employer DB –16.60 (21.22)
Main pension: employer DC –15.77 (19.01)
Main pension: personal pension –18.52 (20.61)
Private sector –18.71 (21.63)
Public sector –14.87 (19.35)
Professional occupationd –18.30 (21.18)
Non-professionale –14.60 (20.26)
Currently retired –18.19 (20.65)
Currently working –15.79 (20.61)
Self-employed –19.82 (21.40)
Flexible retirement date –16.59 (20.56)
Fixed retirement date –14.52 (20.62)
Notes:
a ‘Low education’ denotes no formal or lower vocational qualifications.
b ‘Medium education’ denotes intermediate educational qualifications.
c ‘High education’ denotes degree-level or professional qualifications.
d ‘Professional occupation’ denotes a manager or a professional in current or last occupation.
e Non-professional includes the remaining occupations.
that those who have chosen personal pensions, or work in occupations where they are the norm, are less risk averse, or more wealthy, or have greater flexi- bility with their retirement planning. In any case, they appear to have been more exposed to equities.
Therefore, in accordance with Bodie et al. (1992), those with more human capital, here taken to be those with greater education or in managerial or professional occupations, are found to have been more exposed to equities.
However, one of the interesting features in Table 6.1 is that individuals who have flexibility over their retirement plans appear to have lost no more, and hence be no more likely to hold equity, than those without flexibility. In Bodie et al. (1992), individuals with flexible retirement dates should be willing to risk greater losses in their retirement savings – as they have the option to make up losses through working longer. In contrast, we see no statistically signifi- cant difference in the change in savings between those with fixed and those with flexible retirement dates.
In Table 6.2 we examine whether the differences in sample means discussed above are robust to the addition of explanatory variables and report multivariate regression results. We model the proportionate decline in savings as a function of retirement flexibility, age, education, income and other char- acteristics. Again we see that flexibility over retirement date has little impact on the savings loss, whilst education remains a strong predictor of equity exposure. By contrast, the effects of occupation and income are not statisti- cally robust.