How well have direct investments in real assets performed over the long term? Martijn Cremers outlines the latest research, and challenges some assumptions about inflation protection and diversification
Direct investments in real assets are distinct from traditional investments in public equities and government bonds – for example, due to a lack of active trading on exchanges. This article provides a summary of our recent research paper, in partnership with Deutsche Asset & Wealth Management Global Financial Institute, on three different real asset classes: natural resources (timberland and farmland), infrastructure (specifically investments in energy infrastructure) and commercial real estate. Only the latter has been an asset class in which pension funds have long held major investments.
Using an international sample of over 800 defined benefit pension funds for 1990–2010, a working paper put together with Aleksandar Andonov and Rob Bauer shows that about 80% of the large pension funds in the sample hold direct real estate investments – about the same percentage in 1990 and 2010. Less than 1% of pension funds held investments in natural resources or infrastructure in 2000, while at the end of 2010 about 30% did.
While growing, portfolio weights in real assets remain generally minor. In 2010, the average allocation to real estate among pension funds active in the asset class was 5.75%, while average portfolio weightings remained below 1% for natural resources and infrastructure.
For each real asset class, our research paper discusses the performance of an index tracking the performance of direct investments using publicly available data. Due to data availability limits, we exclusively focus on US data. For direct investment in commercial real estate, data goes back to 1978, the longest sample. For energy infrastructure, data goes back to 1996 only, the shortest sample. For timberland and farmland, data goes back to 1987 and 1992, respectively. Our data ends in December 2012.
The data on natural resources and commercial real estate comes from the National Council of Real Estate Investment Fiduciaries (NCREIF), a trade association of institutional real estate professionals, and is based on data submitted by its members (mostly pension funds).
NCREIF indices reflect both income and appreciation returns, the latter of which are based on changes in quarterly appraisal values. The data on infrastructure are from the Alerian MLP Infrastructure index, which reflects the performance of 25 publicly traded master limited partnerships (MLPs) engaging in energy transport and storage.
Our paper highlights various ways in which the data used are limited (such that the results should be interpreted with caution). For example, the appraisal-based quarterly data seem smoothed which, if taken at face value, would underestimate their volatility. In addition, the NCREIF investment data do not include the costs of managing the institutional portfolios.
Our research paper investigates the following performance aspects:
• Risk-return trade off;
• Downside risk;
• Correlations with public equities, government bonds and other real asset classes;
• Diversification benefits from adding real assets to a portfolio consisting of equities and bonds;
• Ability to hedge exposure to inflation;
• Exposure to systematic downward shocks in the stock market.
Basic descriptive statistics of the investment performance are presented in figure 1.
The cumulative investment performance of an initial $1 (€0.75) investment from the end of 1995 to the end of 2012 is shown in figure 2.
We compare the risk-return trade off across asset classes using the annual Sharpe ratio – the ratio of the average annual return divided by the volatility of the annual returns – rather than quarterly to minimise smoothing effects.
During 1996–2012, the annual Sharpe ratio ranges from 0.45 for the S&P 500 index to 1.55 for farmland investments. While infrastructure significantly outperformed all other asset classes, its annual Sharpe ratio equals 0.70 and is the lowest among the real asset classes due to its relatively high volatility. Farmland and property both have a Sharpe ratio of close to 1.
Downside risk is assessed using the maximum cumulative loss, or the lowest peak-to-trough return. The maximum losses are largest for public equities and infrastructure, with losses of -45% and -43%, respectively, sustained during the recent financial crisis. Commercial real estate also suffered significant losses, of -24%. Downside losses for investments in natural resources were very limited.
The quarterly returns of all real asset classes have low correlations with equities and long-maturity bonds (below 20%), except the Alerian MLP Infrastructure index that has a 36% correlation with the S&P 500 and a -27% correlation with 20-year Treasury bonds.
Figure 3 illustrates the diversification benefits that could have been obtained by combining a portfolio consisting of equity and bonds (60% in the S&P 500, 40% in bonds) with a real asset portfolio (equally-weighted across the available real asset indices).
Adding real assets to equity/bond portfolios would have resulted in significantly lower volatility but would have left the average returns largely identical. The diversification benefits seem most pronounced for direct investments in farmland and infrastructure.
Real assets have provided almost no inflation hedging benefit over the full 1978–2012 period, measuring inflation by changes in the consumer price index (CPI), except for direct investments in timberland. The performance of the NCREIF Timberland index indicates that the performance of timberland investments is positively related, in an economically meaningful way, to changes in CPI.
None of the other real asset classes show a positive association between investment returns and changes in inflation over 1978–2012. One potential explanation may be that inflation changes were quite limited during most of this period, which may make it harder to estimate any hedging benefits.
The main exception is 1978–1987, for which we have data only for commercial real estate, which was characterised by significant inflation. Over this period, we find strong evidence that commercial real estate provided hedging benefits.
Direct investments in each of the real assets considered are illiquid – for example, with generally long holding periods and higher costs of trading. These costs include gathering and interpreting information, time and effort required in finding a party to trade with, and adverse price impact of trades that are relatively large or done at inopportune times. This illiquidity renders these investments more risky, especially for investors with shorter or uncertain horizons. Such investors would presumably demand a positive liquidity premium – that is, a higher expected return on illiquid investments as a compensation for this greater risk.
It is empirically challenging to estimate any liquidity premium without more detailed transactions data. Instead, the main question we address is whether investments in real assets perform worse when stock market returns are negative.
We find strong evidence for timberland and farmland. The investment performance of both the NCREIF Timberland and Farmland Indices has much stronger exposure to equity market returns when these are negative. The relatively large downside equity betas reflect increased risk for investors, as it means that the considerable diversification benefits discussed above are (albeit temporarily) significantly lower during times when equities go down. However, we find no evidence for increased exposure to equity market risk in down markets for property or energy infrastructure investments.
Martijn Cremers is professor of finance at University of Notre Dame
Real Assets: The real risk from real assets
- Currently reading
Real Assets: Performing in the real world