An evaluation of the individual pension forecasts

Transcription

1 Summary 2015:16 An evaluation of the individual pension forecasts Do they differ from outcomes?

2 Swedish Social Insurance Inspectorate Stockholm 2015 Inspektionen för socialförsäkringen Authors: Marcela Cohen Birman and Christian Andersson 2

3 Summary An evaluation of the individual pension forecasts: do they differ from outcomes? The Swedish Social Insurance Inspectorate (Inspektionen för socialförsäkringen, ISF) is an independent supervisory agency for the Swedish social insurance system. The objectives of the agency are to strengthen compliance with legislation and other statutes, and to improve the efficiency of the social insurance system through system supervision and efficiency analysis and evaluation. The ISF s work is mainly conducted on a project basis and is commissioned by the Government or initiated autonomously by the agency. This report has been initiated by the agency. Background Individual pension forecasts are important instruments that individuals can use when planning for their retirement. This applies to decisions about private saving, labour supply, consumption, and the timing of their retirement. In Sweden, individual pension forecasts on public pension incomes, which for the majority are the main retirement income, have been produced by the Swedish Pensions Agency since The regulatory framework for the calculation of the public old-age pension is complicated, and it is likely that only individual-level pension forecasts can help individuals to get an idea of how much pension they can expect from the public pension system at retirement. The forecasts may, therefore, be expected to be the main tools of information for individuals when making decisions about savings and adjustment of labour supply at the end of their working life. Forecasting future pensions requires a number of assumptions concerning, among other things, economic growth, inflation and financial markets returns. These assumptions can be more or less accurate depending on the forecast horizon. Regardless of assumptions, forecasts will always be an estimate that can both underestimate and overestimate future pension outcomes. If large deviations between forecasts and outcomes exist, it could eventually affect the legitimacy of the forecasts. This is particularly problematic if the forecasts systematically underestimate or overestimate the outcome, since a consequence could be that the behaviour of individuals is affected in a detrimental way. 3

4 Objectives The objective is to analyse the accuracy of the public old-age pension forecasts for individuals that are close to 65 years of age, which is when the majority retire. The main objective is to analyse how well public pension forecasts function as a tool for individuals to plan their retirement. We study, therefore, the accuracy of the forecasts and not how this information is comprehended or used by individuals. Methods The analysis is based on register micro data from the Swedish Pensions Agency and the data cover public pension forecasts on the individual level for the years The data include public pension outcomes for individuals born between 1938 and 1948 who started withdrawing old-age public pension at the age of 65. In addition, for the years , we link these data to individual background information (age, gender, education level, sector, country of birth and income) from Statistics Sweden (SCB). A prerequisite for an individual to be included in the analysis is that the individual takes a full old-age public pension at the age of 65. Furthermore, disability pensioners, individuals without a complete record of forecast data in the study period, and individuals that died before the age of 65 are not included in the analysis. The main focus of the analysis is on the income-based old-age public pension (inkomstpension). We also focus on forecasts between 2006 and 2012 and on birth cohorts The period 0 7 years before age 65 is therefore covered. Findings and conclusions The results show that the public pension forecasts, on average, both overestimate and underestimate the pension outcome 1 to 7 years before retirement. Overestimations are, however, more frequent than underestimations. The difference between the public pension forecast and the outcome also varies relatively sharply from one year to the next. No major differences in the accuracy of the forecasts can be observed for different groups based on gender, country of birth, education level, sector and level of pension savings. The overestimations are, however, large for those who have a forecast containing guaranteed pension (garantipension), especially for individuals not born in Sweden. The analysis indicates that it is the income-based oldage pension part that is driving the results. The largest overestimation of the pension forecast was on average 13 percent, corresponding to approximately SEK 800 (85 EUR) per month deflated to 2013 year prices. The largest underestimation was 6 percent, corresponding to almost SEK 500 (53 EUR). Furthermore, the results show a relatively large variation over time in periods when the public pension forecasts and outcomes occur both before and during a balancing period, but also when the income index fluctuates sharply in individual years. The 4

5 balancing mechanism is an element of the pension system that ensures financial stability of the pension system, i.e., that pension liabilities never are greater than the system s assets. The analysis shows that the overestimations and underestimations and the variation over time are primarily a result of the recalculation of the pension balance, which occurs when the income index or the balance index is established. It is also a result of the growth assumption that has been applied to the forecasts up until The assumption of growth changed from 0 percent to 1.6 percent when the individual turned 60 years of age. The large overestimations that are observed for individuals with forecasts containing guaranteed pension is due to the fact that the Swedish Pensions Agency have had problems estimating the number of years of residence in Sweden. Work has, however, been initiated at the Swedish Pensions Agency to mitigate this problem. During the study period, the assumption was 40 years of residence for everyone and this led to a sizable and systematic overestimation of pension forecast for individuals immigrating to Sweden. What level of overestimates or underestimates that can be considered acceptable is hard to say, since a forecast can never be expected to have prefect precision. In this study it has been shown that, a couple of years before retirement age, it is not uncommon for there to be overestimations and underestimations of over five percent on average, corresponding to SEK 500 (53 EUR) or more. In relation to a public pension level of SEK 10,000 per month (1,065 EUR), which is not unusual, overestimates or underestimates of that magnitude can be considered problematic. If the variation in the difference between forecasts and outcomes is to decrease over time, it is necessary that forecasts are made with an assumption that growth is as close as possible to the development of future income or balance index, or alternatively with realistic growth assumptions possibly corresponding to different age groups. However, such forecasts will be much more complicated to calculate and possibly more difficult for the individual to understand. A trade-off between simplicity and an ambition to produce highly accurate forecasts will always exist. Recommendations To avoid unnecessary variation in public pension forecasts in relation to pension outcomes, the Swedish Pensions Agency and other stakeholders within this field should seek stability in assumptions over time. Since 2014, stakeholders within the pension area have agreed on a common standard for forecasts. In order to be able to develop public pension forecasts and pension information, the Swedish Pensions Agency should regularly evaluate their forecasts. Finally, an important prerequisite for this is that the Swedish Pensions Agency develops clear and transparent documentation about how forecasts actually are done. 5