A Product of the MOSERS Investment Staff Value Added A Newsletter for the MOSERS Board of Trustees March 2000 Volume 2 Issue 5 I n this issue of Value Added, we will follow up on the discussion from the January board meeting about replacing some or all of our nominal bonds with real bonds or Treasury Inflation Protected Securities (). are U.S. Treasury debt obligations that provide an investor who holds the bonds to maturity a guaranteed real return in addition to the rate of inflation as measured by the consumer price index (CPI). The principal amount of the security is adjusted monthly by the rate of change in the monthly CPI. This allows the coupon amount and principal to retain the purchasing power of the original investment. A previous edition of Value Added was devoted to the mechanics of ; a copy of that issue is attached. Executive Summary At the conclusion of the last board meeting, our assignment was to analyze two fundamental questions. 1. Should replace a portion of the nominal bonds in the portfolio? (Drawing such a conclusion would require an expectation that, relative to nominal bonds, will exhibit less volatility and lower correlation to equities while having the same or a higher expected return.) 2. If the answer to one is yes, is it possible to increase the allocation to equities, enhance overall fund returns, and keep the overall risk of the portfolio the same? Through extensive discussions with experts from various consulting and money management firms, we are firmly convinced that the answer to question one is yes, if nothing more than from a common sense perspective. The answer to the second question is more difficult because it is extremely dependent upon the expected return, volatility, and correlation inputs used in the asset allocation model. As with most of the decisions made concerning long-term investment issues, the answers will only be known with 100 percent certainty long after we are gone. In the following material, you will find a condensed version of our research efforts with supporting rationale for the numbers ultimately used in the modeling work.
The Differences Between Nominal Bonds and In order to draw some conclusions about these questions, it is important to understand the differences between nominal bonds and. As you can see from the diagram at the right, nominal bonds derive their return from three components: real yield, some compensation for expected inflation, and a small premium to compensate the investor for the chance that inflation differs from what was expected at the time of the purchase. With, in contrast, you earn a real yield plus whatever inflation turns out to be as measured by the CPI. This is why are called real bonds; they pass through the change in the CPI, thus protecting the holder from inflation. Keeping that diagram in mind, it becomes easier to begin drawing some conclusions about expectations for returns, volatility, and correlation. Returns of Relative to Nominal Bonds Let s start with return since it is the driver of the asset allocation model we use. In a theoretical world, should have an expected return somewhat below that of a nominal bond. Why? Because inflation isn t a source of risk for ; therefore, their yields should not incorporate an inflation risk premium. The amount of additional return an investor should expect for taking on the risk of actual inflation being different from expected inflation is difficult to pinpoint. Most studies indicate that the premium is between 25 and 50 basis points. Simply put, in theory, should return 25 to 50 basis points less than a nominal bond of equal duration. Now let s move from theory to the real world. Today, an investor can purchase a 30-year nominal bond that yields a nominal return of 6.15 percent, or that same investor can purchase a bond that yields a real return of 4.15 percent. Going back to our discussion about the components of return, by subtracting the nominal bond s yield from the yield we can conclude that the 2 percent difference is what the market is currently paying an investor for Risk Premium Inflation Real Conventional Treasury Actual Inflation Real expected inflation and the risk premium. In other words, the market is telling us that over the next 30 years inflation will average somewhere in the range of 1.5 percent to 1.75 percent. If the market is right, then the real return on the nominal bond and the return on the will be exactly the same. If inflation averages less than the expected range of 1.5 percent to 1.75 percent, then the real return on the nominal bond will be higher than the return. The opposite will be true if inflation averages more than the expected range; the return will be higher. Given that consensus estimates for inflation over the next ten years are in the 2.5-3.0 percent range, it is our opinion that are very attractively priced relative to nominal bonds and that an investor in today might expect returns in excess of real returns on nominal bonds. Volatility of Relative to Nominal Bonds It s much tougher to estimate the risk (volatility) of. When interest rates change, there is no way of knowing the underlying cause whether it was real yield, inflation expectations, the risk premium, or some combination of the three that changed. All we know is that the total nominal yield changed; we don t know what happened to the
components. So we don t know how sensitive real yields or yields are to changes in nominal interest rates. So how can we estimate the risk of? One way is to assume that changes in nominal yields will come partly from changes in real yields and partly from changes in inflation expectations and the risk premium. Under this scenario, will exhibit less volatility just because they are only exposed to changes in real yields. Remember, they are essentially immune from changes in inflation expectations and the risk premium. Another way to explore this issue is to look at other markets. The U.K. first issued inflation indexed bonds in 1981, so we can look at the track records in this market to see if there is a pattern for how real yields move in relation to nominal yields. The chart below, for example, compares the historical yield of the U.K. s real bonds, or Linkers, with the yield of the U.K. s government and as it moves up or down, the portion of the liability tied to the retired lives changes. Nominal bonds do a poor job of tracking that liability due to the fact that as inflation increases interest rates generally will increase also. Increasing interest rates cause nominal bonds to fall in value at the same time that the value of the liability is increasing. The opposite is true with, as they will generally move in the same direction as the inflation sensitive liabilities. The conclusion here is that should have minimal volatility relative to the inflation sensitive liabilities while nominal bonds will be very volatile. All three of these examples point to intuitive reasons why should be less volatile than nominal bonds. Given that this is not an exact science, we have, in our modeling effort, made an educated guess that will have roughly 60 percent of the volatility of nominal bonds. Historical s Long Gilts (Nominal) Long Linkers (Real) Naive Inflation Expectations bonds, or Gilt. From the graph, it is clear that the Linkers have exhibited less volatility than Gilt. Over the past 18 years, the yield on these securities has ranged from 2 percent to 4 percent, while Gilt yields have ranged from 4 percent to 13 percent. What this tells us is that when nominal rates change in the U.K. real yields also change but to a much smaller degree. One final way to explore volatility relative to nominal bonds is to think about how react to changes in our liabilities. As you know, the MOSERS plan includes a COLA for retirees. The COLA is tied directly to the CPI, Correlation of to Nominal Bonds and Stocks E stimating the correlation of to nominal bonds is much like estimating the volatility. Intuition tells us certain things about how should act relative to other assets, yet pinning down numbers is more art than science. Remember the assumption that we made earlier; changes in nominal yields will come partly from changes in real yields and partly from changes in inflation expectations
and the risk premium. If changes in yields were 100 percent due to changes in real yields, you would expect to have a very high correlation to nominal bonds. On the other hand, if changes in yields were coming completely from changes in inflation expectations and the risk premium, you would expect to have no correlation to nominal bonds. Assuming that the truth lies somewhere in between, it seems reasonable to conclude that should have a correlation to nominal bonds in the range of.4 and.6. The correlation between returns and equity returns should be low. This is because over long periods of time the correlation will result from the average of two very different economic scenarios. In the first, which seems to be our recent experience, equity values will be increasing because interest rates are falling in response to policies that are causing inflation to fall. In this environment, and equities will have positive correlation. That is, they will have correlation above 0. In the second scenario, interest rates fall because the economy slows, as in a recession, and the demand for capital declines. The positive effect of a decline in interest rates is more than offset by the decline in expected future earnings for equities. In this case, and equities should have no correlation or some negative correlation. Of course, each of these scenarios can be reversed, giving us increasing inflation and interest rates that also produce conflicting implications for the correlation. The net effect will be the same. In some cases, have positive correlation to equities, and in other cases, the correlation will be negative. Over long time periods, one expects to see some mix of these circumstances. In any short period, we may see either. Since we don t expect to successfully predict the scenario, our best choice is to rely on historical data and forecast a low correlation between and equities. With this as the framework for the inputs to the asset allocation model, we will, in the following section, put some numbers to the theory. The Asset Allocation Model In order to quantify the value of replacing nominal bonds with, we will use a mean variance computer model, as we do in all of our asset allocation studies. As with any computer model, the mean variance model is only as good as the inputs. In order to run the program, the return, volatility, and correlation for the different asset classes must be entered. This is where the science of computer programs meets the art of projecting returns and risk levels for asset classes. It is very difficult to arrive at a single expected return number for our asset classes. The logical source of information for the input is history. The problem with this is that the story history tells is very dependent upon the specific time period chosen. There are sources of data that will provide the return of stocks and bonds going back to 1926 and beyond in some cases. If you use the entire time period, you get a different answer than if you use the last 50, 30, or 10 years. The important questions that have to be answered are not what an asset class has done in the past but what it will do in the future? In addition, how long into the future should you try to estimate the returns? Should we take into account that we are currently at record high valuations for equities? If we look at the last 10 years for nominal bonds, should we use the rather lofty returns that this asset class generated? Would it be better to consider an environment of high inflation and interest rates or a period like the one we are currently in with relatively low inflation and interest rates? The message that we want to emphasize is that although the model is great for providing detailed answers (many to two decimal point precision) the results are only as good as the inputs used. To make matters worse, we must attempt to model an asset class that has only been around in the United States for three years (). This relative lack of data makes comparisons with the other asset classes even more difficult. As you might suspect, having had numerous philosophical discussions regarding, the one thing we were completely sure about was that the numbers ultimately used were not nearly as important as the comparison of portfolios with and without. With that in mind, it was decided that a number of scenarios would be shown to demonstrate how this substitution would affect the portfolio based on different economic assumptions. The first example uses the collective thinking of those who worked extensively on this project, we call this our current expectations. These inputs are somewhat based on history and somewhat based on expectations for the future, and we believe the numbers to be conservative. The inputs used and the results obtained are shown in the table at the top of the following page. In this example, we will show the inputs, then the expected return and expected risk for the current portfolio (containing nominal bonds), and then the test portfolio with having replaced nominal bonds. In the examples, large cap and small cap stocks include
domestic and international. The small cap stocks also include the small allocation to commodities. The logical result from using instead of nominal bonds, based on these inputs, is that the portfolio would have a slightly higher return and a slightly lower risk. This would be anticipated because the expected return for is higher than for nominal bonds and their risk is lower. The computer confirms and quantifies this expectation. By replacing the nominal bonds with, we have a portfolio that is expected to return 20 basis points more each year and have 39 basis points less volatility. In an attempt to move from the what if to the real world, we used the actual returns from a very bad ten-year period and a very good ten-year period. You have seen this sample period previously; it is the same set the actuary used in the presentation at the January board meeting. The 1973-1982 period was used because the market should not get much worse than it was during that period inflation was on the rise and stocks were in the tank. The nominal results for the asset classes were converted to real returns and entered into the program. The inputs and results are shown in the table to the right. As you can see, will not solve the problems of the world. The expected return for our current portfolio, if we were to experience the same returns as we did from 1973 to 1982, would be a negative1.22 percent, with an expected standard deviation of 17.75 percent. If are used instead of the nominal bonds, the expected return improves to a positive 1.23 percent for an improvement of 2.45 percent with lower volatility. The large allocation to equities will still dominate the portfolio, and the will only soften the blow. Next, we looked at the following ten years, 1983-1992. This period should be considered a very good one for financial assets with falling interest rates and inflation. During this time period, stocks were doing very well. The inputs and results are again shown in the table to the right. The portfolio did not do as well as the current mix. With nominal bonds returning more than 7 percent per year over this period as inflation expectations fell, it is easy to understand why the current portfolio is superior to the. In this example, having in place of the nominal bonds cost the portfolio about 90 basis points of return. The important point from these three examples is that the should buffer the portfolio from the bad times, but it will cost the portfolio in times when nominal bonds are performing well. Current Expectations Return 1983-1992 Period Standard Deviation Intermediate Term Govt. Bonds 3.5% 7.2% Long Term Govt. Bonds 4.0% 10.9% 4.2% 4.5% Large Cap Stocks 8.0% 20.8% Small Cap Stocks 9.0% 29.2% No Replace Portfolio Nominal Return 6.73% 6.93% Standard Deviation 15.58% 15.19% 1973-1982 Period No Replace Portfolio Nominal Return -1.22% 1.23% Standard Deviation 17.75% 16.09% No Replace Portfolio Nominal Return 10.41% 9.44% Standard Deviation 10.61% 9.30%
Summary If we leave you with nothing else, there is one very important concept that is helpful in an analysis of relative to nominal bonds. This concept is called the breakeven inflation rate. This term describes the rate of actual inflation that produces the same return for the nominal bond investor as for the investor. As we discussed earlier, the market is currently telling us that this rate is between 1.5 percent and 1.75 percent. Having looked closely at historical levels of inflation, we know that in no 30-year period since 1929has inflation measured less than 1.75 percent and only twice has it been below 2.0 percent. Both of these periods included the great depression of the late twenties and early thirties. The chart below gives you a visual reference of this point. If you were to buy long nominal bonds, today, you would lock in a yield of 6.15 percent. The same investor can buy a bond yielding 4.15 percent. The current consensus forecast for long-term inflation, generated by a group of leading economists, is 2.5 percent. Based on today s market rates, if this forecast of inflation turns out to be correct, the will outperform the nominal bond by 50 basis points. Without question, this is a very complex subject. Historically, all investors have had are nominal bonds. Now we have that require a reengineering of our thought process in order to fully understand them. We believe this is at least one of the reasons that investors have been slow to endorse. This relative lack of enthusiasm may be why are currently attractive. Investment theory tells us there are no free lunches ; you have got to take more risk to expect more return. While we certainly have no award winning thesis to back up our beliefs, seem to offer pension fund investors, like MOSERS, at least a free snack. 6.15% Nominal 6.65% Inlfation Real 2.50% 4.15% This newsletter will be produced and distributed three to four weeks in advance of each scheduled board meeting with the objective of educating the Trustees regarding investment issues facing the pension fund. If you have questions or would like additional information on any topic contained herein, please contact the investment staff. MOSERS Investment Staff Rick Dahl Chief Investment Officer 632-6160 Jim Mullen Portfolio Manager Fixed Income 632-6164 Pat Neylon Investment Officer Equity 632-6165 Tricia Bisges Investment Officer Fixed Income/Cash 632-6161 Meg Cline Junior Investment Officer Equity 632-6144 Angela Swanigan Operations Officer 632-6166 Karen Holterman Junior Operations Officer 632-6163 Conventional Treasury Cindy James Executive Assistant 632-6147