Issues with Ghilarducci GRA proposal: diversification

On Monday I began outlining several issues I have with Teresa Ghilarducci's proposals for Guaranteed Retirement Accounts (GRAs). The issues discussed here is diversification, both financial and political. Financially, the GRA proposal would place most worker's entire retirement savings in low-risk, low-return assets. Politically, the GRA proposal would have most retirees derive their entire income from the government. In neither case do I think it's wise to put all your eggs in one basket.

Financial diversification: For the typical worker, money invested in Social Security's retirement program is akin to the purchase of bonds paying an interest rate equal to the rate of total wage growth in the economy. Since Social Security is a pay-as-you-go program that transfers money from workers to retirees, the implicit rate of return on the program is equal to the rate of growth of money that can be transferred: roughly speaking, this is the rate of labor force growth plus the rate of wage growth per worker. Going forward, these produce an implicit Social Security return of around 1.5% above inflation. While each person will earn a slightly different return, based on how their life circumstances interact with the complex Social Security benefit formula, for the average person the 10% or so of wages they pay to the Social Security retirement program will earn this 1.5% return.

To this, the GRA plan would effectively require that workers contribute an additional 5% of wages to individual accounts managed by the Social Security Administration. While I have no particular problem with this, what does worry me is account holders would have no choice regarding what their contributions would be invested in. Every worker's account would consist of special government bonds paying a fixed return of 3% above inflation.

The key insight of portfolio theory is that by combining different assets we can produce higher returns at lower risk. The reason isn't simply that assets have different levels of risk, it's that they don't always go up and down together. Mixing together assets that don't have perfect covariances can reduce the risk of the overall portfolio. So even people who want a low-risk, low-return portfolio might have stocks in it, while even those seeking higher returns might have some bonds in their portfolio.

The chart below uses historical data on the standard deviations of annual returns for stocks, corporate bonds, trust fund bonds (which serve as a stand-in for the 3% return paid to the GRA accounts) and a "pay-as-you-go bond" that pays the same average return as Social Security. In addition, the chart utilizes the covariances between each asset: how closely the asset returns tend to follow each other. From this data, I calculated the lowest-risk portfolio for each expected rate of return. The rate of return any individual seeks is a function of their own taste for risk, but this chart can tell you that if you seek an x% rate of return that there's a mix of assets that minimizes your risk in doing so.

What does this chart tell us? First, that even very risk-averse people – those seeking a return of only 3% above inflation, which is roughly what government bonds are projected to pay – would not invest solely in Social Security and a GRA account. Even these very low-risk investors would hold about 14% of their portfolio in stocks and around 21% in corporate bonds. As individuals' taste for risk increased, the percentage of their portfolios held in Social Security bonds and the 3% GRA-bond would change. Individuals seeking a 4% return would hold 42% of their portfolio in GRA bonds, but nothing in Social Security. Individuals seeking a 5% annual return would invest only 16% of their assets in GRA bonds and nothing in Social Security, but 32% in corporate bonds and 53% in stocks.

The point here is that, since few people save more than 15% of their income for retirement, the GRA proposal forces individuals to invest most or all of their retirement savings in a portfolio that may be lower-risk than they desire and doesn't maximize returns given the level of risk the portfolio does have.

Political diversification: As noted above, not many Americans save more than 15% of the income for retirement, for the good reason that 15% is a pretty adequate saving rate. As a result, were the GRA plan put in place, such that Americans saved roughly 10% of their wages through the Social Security retirement program and another 5% through GRAs, most people would choose to save nothing outside of these programs. In other words, their entire retirement income would derive from one source: the government.

You don't need to be a wild-eyed libertarian to think that complete dependence on government for retirement income is unwise. As noted yesterday in the Financial Report of the United States, the government faces very large long-term obligations for which it has not well prepared. For that reason, Social Security benefits are an at least somewhat risky proposition going forward. Adding increased retirement benefit obligations through GRA accounts, which may or may not be fully pre-funded, seems like asking for trouble.

Alternately, most people can help diversify against the solvency risk of Social Security benefits by investing in market assets whose returns aren't directly affected by the federal budget. In other words, as Social Security faces financial trouble, the rational solution for most individuals would be to save more outside of Social Security, not to start a new retirement program dependent on the same funding source. Some people like the idea of everyone getting their retirement income from the government, but I suspect that's more of a philosophical preference than a well thought-out policy conclusion.