Japan's Pension System -Evaluating the 2004 Reform and Establishing Clear Principles for Further Reforms-

RIETI Faculty Fellow FUKAO Mitsuhiro made the following points in a presentation titled "Enhancing the Financial Sustainability on a Basis of Clear Principles."

I have made various suggestions regarding the problem of Japan's public pension system since I was first involved in the white paper on the economy in 1985, yet many issues remain unresolved. Repeated amendments and protection of vested interests have made the system increasingly complex and the Ministry of Health, Labour and Welfare's interpretations have changed at each revision.

I would like to report on our examination of reform plans based on fundamental pension principles, aimed at making Japan's pension system more sustainable. We have developed a new pension funding model (the RIETI model) for quantitative analysis of reform proposals, such as integration of the National Pension System and the Employees' Pension System that could not be analyzed with existing simulation models. Based on the same economic assumptions as the standard case for the Ministry of Health, Labour and Welfare's fiscal recalculations, the RIETI model is designed to provide similar results.

As outlined in the previous session, the 2004 reform can be summed up in the following four points:
(1) Incremental increases in the contribution rate and legal stipulation of the ceiling for such increases.
(2) Control of benefit increases by means of a macroeconomic slide system.
(3) Drawing down of the pension reserve fund by means of a finite horizon balancing mechanism.
(4) Raising of national government subsidy for the basic pension.
However, opinions are divided on the effectiveness of the reforms.

Public pensions began in Japan with the establishment of the Employees' Pension System in 1942. This was a form of compulsory savings. The post-war employees' pension was characterized by low contributions and remained a very simple system until the introduction of the basic pension in 1986. That system was introduced because the high level of benefits under the National Pension System prior to 1986 was having a negative impact on public finances, but since 1986 the relationship between benefits and contributions has become obscure. Subsequently the system was repeatedly modified in attempts to stabilize the fiscal burden, but its basic framework has not changed.

Under the existing Employees' Pension System, contributions are earnings-related while benefits consist of two tiers: a fixed-rate basic pension and an earnings-related component. Since benefits are partially paid at a fixed rate, the system is advantageous for low income earners and disadvantageous for high income earners. There is also a differential between households with a fulltime homemaker and households where both partners work: if the internal rates of return are compared, the system favors households with fulltime homemakers. By contrast, the National Pension System has fixed contributions and fixed benefits.

The pension system combines the principles of insurance and assistance. For the insurance element, the burden is based on the principle that those who benefit from the system should contribute, and benefits are paid in proportion to contributions. For the assistance element, the burden is based on the principle of ability to pay and benefits are paid according to needs.

If we calculate the internal rate of return for the Employees' Pension System following the 2004 reform, we see that the earlier the birth year of the recipient, the larger the benefit received relative to contributions. For someone born in 2005, the internal rate of return will fall below the projected yield on investment of 3.2%, while only in model households will the rate exceed the projected wage increase rate of 2.1%. The system is disadvantageous to high income earners and advantageous to households with fulltime homemakers, and favors women over men. In the case of the National Pension System, the large government subsidy means that even for someone born in 2005 the internal rate of return is close to the projected yield on investment of 3.2%. Thus under the current system paying into the National Pension System provides a better return.

One reason trust in the existing system has been eroded is that it is difficult to see how much of the contribution an individual pays is reflected in the benefits received and how much is used in assistance. The current system is so complex that even the Social Insurance Agency makes calculation errors. In an opinion poll on the public pension system conducted by the Cabinet Office in February 2003, more than 80% of respondents agreed that a system with a clear relationship between contributions and benefits is preferable.

As the first step of our research we simulated paying basic pensions entirely from the national treasury. The reference year for the reform was set at 2010, with a reserve ratio of 1 in 2100. If the schedule for increasing contributions set out in the 2004 reform is maintained, the macroeconomic slide is applied, and survivors' pensions are excluded, the benefit multiplier works out to 10.45, equivalent to 1.91 times the current rate. A multiplier of 10 means that for each year of contributions paid annual pension income increases by 1%, so if contributions are made for 40 years the earnings-related portion of the pension will be 40% of annual income, in addition to the basic pension. However, raising the government subsidy for the basic pension would mean a bigger role for government, and in this scenario survivors' pensions would need to be covered from tax revenue.

When we calculated how far the contribution rate could be lowered while maintaining benefit levels, we found that if the macroeconomic slide is applied and survivors' pensions are maintained the contribution rate could be reduced to 11.938%, or to 8.442% if survivors' pensions were excluded.

If half of the basic pension were funded from consumption tax, with application of the macroeconomic slide, the consumption tax rate would need to be increased by approximately 4 percentage points by around 2010 and approximately 6 percentage points by around 2050. If this additional consumption rate were leveled out over the years from 2010-2100 it would amount to 5.5%. In this case a basic pension reserve would be built up, but the amount would be large enough to have an impact at the macroeconomic level.

Next we simulated a new pension system integrating the Employees' Pension System and National Pension System and combining a fully earnings-related pension with a guaranteed minimum pension (at the level of the basic pension). The internal return rate of such a new integrated pension would exceed the wage increase rate for women, but not for men. The problem with such a system is that compared to the existing regime it would disadvantage low income earners. The pension reserve would also increase above existing levels.

Another challenge for the future is that sufficient consideration has not yet been given to transitional measures at the time of system changeover.

After the presentation RIETI Consulting Fellow KANEKO Yoshihiro added the following comments from an actuarial perspective.

The approach of this research is to sort out the Japanese pension system based on fundamental pension principles.

Reports on pension systems in OECD countries all cite the role of such systems as sharing the risk of longevity and redistributing income. In that sense the existing structure of a basic pension plus an earnings-related component can be seen as far-sighted.

However, the Japanese pension system has become too complicated and cannot win public trust. It is essential to reform the pension system in line with structural and attitudinal changes relating to industry and work. Sweden has developed an NDC system and the U.S. has a tax-based social security system. The simulations we conducted in this study aim to complement debate on scenarios such as increasing the government subsidy or splitting off disability and survivors' pensions into a different system.

In response to the points made in the presentation, Olivia S. MITCHELL (Professor of Insurance & Risk Management, The Wharton School of the University of Pennsylvania) made the following comments.

A sustainable pension system needs to be transparent and resilient. It is also important not to trigger negative incentives. Personally I believe that in developing a model an infinite horizon for balancing should be a prerequisite.

To achieve sustainability, it is not sufficient simply to ensure adequate cash flow. Care must be taken if the need for actuarial solvency changes basic assumptions, since results will also change.

Management of the reserve also has an impact on the sustainability of pensions. Moreover, there are various ways of setting the minimum pension level.

Since individual circumstances alter due to factors such as change of employment status or divorce, there is a need to consider how to ensure full pension coverage, and to this end it is essential to amass microdata.

Care must be taken with the internal rate of return, since adjusting assumptions will dramatically affect return calculations.

An adjustment formula is required for revising the system if economic performance fails to live up to projections.

Pension finances need to be calculated on an infinite time horizon. It is also helpful to conduct simulations at the micro level.

Next, Ole SETTERGREN (Director of the Department of Pensions, Swedish Social Insurance Agency) offered the following comments.

It is important to consider what principles are essential and what kind of design will embody them.

Splitting off disability and survivors' pensions from the old age pension is an effective option. Such a split would simplify the problem.

In Sweden there is no great difference in pensions for white-collar and blue-collar workers.

Why would the reserve increase if the Employees' Pension System and the National Pension System were integrated?

Prof. Fukao gave the following responses to the comments above.

To Prof. Mitchell:
Concerning the resilience of the system to adjust to changes in employment status or divorce, the mechanism for splitting employees' pensions in the case of divorce that comes into force in April 2007 could be regarded as one formula for such adjustments. However, that mechanism is not compulsory and ideally it would be desirable to divide pension eligibility accrued during the period of the marriage at the time of divorce.

To Mr. Settergren:
Because Japan's demographic composition is distorted and the population will shrink in the future, the pension reserve will swell temporarily. The reason the reserve would increase further under a new integrated pension is because higher contributions are collected from the self employed.

Supplementary comment:
The basic pension for someone who has made contributions for 40 years is roughly the same as the level of public assistance (i.e. welfare benefits), but since obtaining public assistance involves means-testing and investigation of family circumstances we can assume that most people would prefer to pay contributions and receive a pension.

Prof. Kaneko then made the following response to Prof. Mitchell's comments:
Research undertaken by Prof. Higuchi, who will give a presentation in Session 4, relates to the question of the desirable level for benefits, but basically some form of criteria for deciding benefit levels should be established after ascertaining patterns of consumption, living conditions and standards of health based on microdata.

The current system should be modified to make it easier for divorced women who work part-time to join the Employees' Pension System. Another point relevant to the question of part-timers is that a fully earnings-related pension system is an option to be considered for making the system responsive to changes in employment formats.

Q: What happens if mutual aid pensions are integrated under the RIETI model?

A (Prof. Fukao): The problem with analyzing mutual aid pensions is that the data are not available. Discussion of integrating mutual aid pensions will be difficult unless information on mutual aid scheme members is disclosed, but public servants (i.e. the members of such schemes) would probably object.

Q (For Prof. Mitchell): I understand that in the U.S. pension benefit levels differ according to income levels. What discussion has there been about a fully earnings-related pension system?

A: Since 2001 there has been lively public discussion about social security in the U.S., but the Bush administration has lost its momentum and social security reform has stalled. Personally I think that a minimum guaranteed pension set at 1.2 times the level of public assistance is preferable to a fully earnings-related system.