|Mr Pitzer?s contribution (19 August 2003) lays out some implications of not imputing liabilities and property income for employers? unfunded schemes. He points out that the costs of providing pension benefits to employees should not differ simply because of the means of financing. Paying unfunded pension benefits rather than contributions to a funded scheme means to pay the historical contributions plus the imputed property income accrued over the past decades. Hence, future benefits should be discounted to their present value. The author, Mr. Steurer, agrees with these statements. This note supplements the author?s contribution dated 6 June 2003 and attempts to correct some possible misperception or imprecision.
There are two potential sources of error when using the benefits-paid method: demographic bias and property income bias. The 1993 SNA mentions the first source of error, but not the second. In Section 1, this note tries to clarify the concept of the property income bias. Section 2 indicates possible orders of magnitude of these errors (based on some extremely heroic shortcuts). Section 3 looks at the implications for international comparisons of labour costs when estimation methods or pension systems differ across countries.
In Section 1, it is observed that (although this is a bit surprising), for a stable defined benefits scheme, when the rate of future salary increase is equal to the rate of discount (or the assumed future return on the assets of a funded scheme), then the benefits currently paid are exactly equal to the actuarial value of contributions. However, Section 1 shows the large sensitivity of results to the discount rate ? GDP growth rate differential.
National accountants have to judge the quality of their data sources. For imputed contributions, possible errors could be large. It cannot currently be assumed that actuarial estimates made by the employer are in all countries a better source than benefits paid or wage shares. If contributions to employers? unfunded schemes should be identical to those to an equivalent funded scheme, the benefits-paid method can only be used when the size of major potential errors can be assumed to be small. Subject to taking account of possibly large differences across countries, at the moment this seems to be the case. The EDG so far has not spent much time addressing practical problems of what should be done when actuarial estimates are not available or, if they are, are judged to be biased.
Anton Steurer is an economist at Eurostat (unit B1: National accounts methodology, statistics for own resources)
The author of this contribution to the discussion group on this site bears the sole responsibility for both the substance and the style of the contents. The purpose of the discussion group is to elicit comments and to promote debate on specific topics. As such, the views expressed on any of the issues raised are not to be attributed to the IMF.