I've been blogging here at TPMCafe a bit about my idea to have the US Gov't start up a public mutual fund, US Mutual.

This idea is not completely new with me. Here is an article that describes how a similar idea was considered during the Clinton Admin. The bottom line is that

The projected annual rate of return on U.S Treasury securities held in the Social Security trust funds is 2.7 percent, after inflation. In contrast, stocks generated an annual return of about 7 percent above the inflation rate from 1900 to 1995. If past serves as prologue and stocks continue to significantly outperform Treasuries in the future, diversification would bolster the trust funds.

Now my idea has some significant differences from the above idea. Differences that address some legitimate concerns about the politicization of the fund and the use of its market power. US Mutual would target a 6% control of the US Companies in the NYSE. According to Wikipedia, that would make the targeted holdings .84 Trillion or 840 billion, approximately half of the 1.7 trillion that was in the Social Security Trust Fund at the end of calendar year 2004. This would be seriously larger than other mutual funds, like the Vanguard Group with only .1 Trillion in holdings. However, as a public mutual fund, it would not compete with other mutual funds.

It would be a new form of index fund that would be low-cost and have a wide diversification and have a low turnover in its long holdings(it would also have short holdings.). It would determine its weekly holdings using a simple algorithm that would use 29 weeks of publicly available NYSE weekly valuations and embody a conservative investment strategy.

The key to the algorithm is that it would use more robust median AR(1) statistics to predict the weekly log-return(=ln(1+return)) and its standard deviations. Log-returns would be used because they are easier to predict in stock markets and usefully underweight large gains and overweight large losses. It will make predictions conditional on the past week's values and unconditional predictions. It will combine the unconditional and conditional predictions, giving the unconditional predictions twice as much weight since their values would vary less on a weekly basis. The result would be predicted values of the median weekly log-return and its median standard deviations for all the US stocks that have been traded publicly on the NYSE for at least 29 weeks. An index will be formed that would weight the predicted median standard deviations twice as much as the predicted median log-returns.

Then, using predicted values of the total market capitalizations for stocks as weights, US Mutual will use 2/3rds of its investments to buy long on the top 20% of stocks based on the index formed. This would result in holdings of an average of 20% in these stocks. The individual stock holdings would be determined by taking the index values for the stocks bought and standardizing their index values and using the standardized values as weights.

US Mutual would also capitalize on the information in its algorithm by using 1/3rd of its funds to sell short(meaning we make money when the value goes down)on the bottom 5% of stocks evaluated by the index, with the individual levels determined by weights formed as above. What this would do is have US Mutual profit some from exerting some market power on unstable underperforming stocks that are likely overvalued. It would also help US Mutual to have a low beta, as it would buy long in more stable stocks (that likely have lower betas like Berkshire Hathaway), and sell short in unstable stocks (that likely have higher betas). This would hedge well against a worse-case scenario of a general downturn. It would also affect a general change in stock-market behavior, discouraging high turnover strategies that tend to increase stocks' volatility.

The algorithm described above is designed with the goal to determine holdings mechanically in a way that would be stable and easy to replicate. This serves to both keep the costs of the fund down and to prevent its "managers" from being corrupted or influenced in their stock-holdings. Attempts at arbitrage would be limited by how the limited levels of weekly fund turnover would be staggered over the course of the week, using algorithms that would make use of the information in the conditional predictions.

So, so long as the buy long holdings do not vary significantly over time(as could be verified easily using historical data), the considerable market power of US Mutual would serve to help stabilize the market. As for the impact of its short interests, there would likely be a transition phase where some unstable stocks would leave the market, but eventually the fund would simply bleed regularly the more volatile and overvalued stocks in the market, perhaps in exchange for helping to reduce the general volatility of the market. A reduction in volatility would attract more capital to the market and increase the growth rate of our economy. It would also make it easier for smaller investors and corporations to trade and raise capital on the market.

So that's the idea. We set up protections around the SSTF and allow it to grow by investing half of it in the US Companies in the NYSE. We do this with low overhead, using a strategy that hedges well against a general downturn. All that remains is to make the idea simple and compelling. And in the worse case scenario, whenever a Republican brings up the privatization of Social Security, we can counter with this idea, making them wish that they'd never brought up the subject...

dlw