Do stocks match pensions liabilities as well as bonds match them?
The Universities Superannuation Scheme (USS) is committed to a costly de-risking of the assets in the pension scheme from stocks (equity) into government bonds (gilts), on the theory that bonds are less risky because the income they generate more closely matches pensions liabilities. According to USS, ‘pensions are similar to inflation-linked bonds, in that the level of payments is pre-determined to be subject to inflationary increases’ (‘An integrated approach to scheme funding’, July 2014, p. 12).
In a recently published article, however, John Woods has argued that, when measured by mean duration, the dividend payments on stocks are about as good a match to pensions liabilities as are the coupon payments on inflation-linked bonds. He writes:
Recognising that an equity can be regarded as an irredeemable bond with a variable coupon … an equity, or a portfolio of equities, has similar characteristics to index–linked [bonds] because there is typically a regular income payment (i.e. a dividend compared with a coupon, both generally paid semi–annually) with a growth component (i.e. a variable rate of dividend growth, compared with a variable rate of inflation). … If duration concepts can be applied to index–linked, we argue that they can also be applied to equities….At this point, we can appeal to the work of Dechow et al. (2004) and Schröder and Esterer (2012), who have provided a theoretical basis for, and empirical estimates of, equity duration: note, in particular, that their estimates of the mean equity duration were similar to the PPF’s estimate of the mean index–linked gilt duration. (‘On the political economy of UK pension scheme regulation’, Cambridge Journal of Economics, 41 (2017), p. 158)
Even if, contrary to Woods, income from gilts tracks pensions liabilities better than income from equities, this virtue is overwhelmed by the fact that the expected returns on equities are much greater than on gilts. According to USS’s February 2017 consultation document (see p. 30), Mercer and USSIM estimate the 20 year annual return on equities as CPI plus 3.1–3.5%, and they estimate the 20 year annual return on index-linked gilts as CPI minus roughly 1%. Therefore, even if income from equities tracks pensions liabilities less well than income from gilts, the fact that the expected total volume of income from £X of equities is much higher than the expected total volume of income from £X of gilts ensures that the liabilities will be covered. Why match pensions liabilities with a small stream of income that has the same shape as the liability rather than a less well matched but large river of income that will more than cover this liability, whatever the mismatch?
The above provide further grounds, beyond those I have discussed here, for resisting USS’s Test 1 and its associated gilts-based self-sufficiency valuation.