Timber as an Asset Class

Updated April 2006
PDF April 2006 Journal

In their search for higher returns and risk reduction benefits, more and more investors are moving beyond traditional debt and equity, and including more of what are termed "alternative asset classes" in their portfolios. These include real return bonds, domestic and foreign commercial property, commodities, various hedge fund strategies (e.g., equity market neutral or global macro), and timber. With the latter growing in popularity, it is a good time to look more closely at the supply and demand of returns in this asset class, so that we may reach a conclusion about whether it is over, under, or fairly valued today.

Our basic approach to this issue is to compare the rate of return an asset class is expected to supply (which is equal to its current dividend yield plus the expected real growth rate for those dividends) with the rate of return an investor should rationally demand for holding the asset class (which is equal to the current yield on a long term real return government bonds, plus an appropriate asset class risk premium).

Let's start with the rate of return an investor should logical demand on his or her investment in timber. Over the last fifteen years, the historical spread between the rate of return on the main U.S. timberlands index (published by the National Council of Real Estate Investment Fiduciaries) was 6.5%. However, since this index tracks the performance of directly owned timberland investments, the 6.5% spread must reflect not only compensation for the relative risk of timber as an asset class, but also for the illiquid form in which it is held. Investors in timber based real estate investment trusts that are traded on the stock market (e.g., Timber REITS like Plum Creek and Rayonier) hold a much more liquid investment. Assuming the illiquidity premium equals 2.5%, the adjusted risk premium on liquid timberland investments drops to 4.0%.

Let's now look a little closer at this risk premium, and ask whether it is reasonable in light of the underlying risks that it compensates an investor for holding. The timber cash flow generation process is relatively simple: sun, rain and soil cause trees to grow; costs are incurred to manage the forest, harvest and market the trees; and the price for which they are sold varies. For the asset class as a whole, the physical risks are quite small. For example, Grantham, Mayo and van Otterloo (a major timberland manager) claims the annual risk due to disease, fire and the like is about one half of one percent per year. The potential impact of global climate changes is harder to quantify, but undoubtedly adds more risk to some timber investments (e.g., a cooling climate in northern Europe would slow tree growth) while reducing it elsewhere (e.g., a warming climate in the eastern U.S. would speed tree growth). Operational risks are also small, because the underlying process is biological, with most inputs provided at no cost by mother nature. The major business risks in owning timber are on the demand side of the cash flow generation process. Broadly speaking, they fall into two categories: volume risks and price risks.

Timber demand is affected by two main factors. The first is economic growth; the stronger the economy, the higher the physical demand for wood in all forms (e.g., lumber, panels, cellulose fiber, packaging and paper products). The second is technological innovation, which enables a greater volume of end-use products to be produced from the same amount of wood. A classic example of this is the development of oriented strandboard (OSB) panels, which are made from sawmill scraps that were previously treated as waste. The interaction of these two factors results in the physical demand for timber growing more slowly than the real economy. For example, in "World Agriculture: Towards 2015/2030", the U.N. Food and Agriculture Organization estimates that long-term physical demand for wood will grow by 1.6% per year, versus 3.6% for real world GDP. On the other hand, this historical rate of real GDP growth is not very volatile, which implies that variations in physical timber demand (a measure of risk) shouldn't be very volatile either. But that's not the whole story.

Price risks result from the interaction of the demand for and supply of wood. In recent years, supply has been steadily increasing, due to the planting of more timber plantations and improved forestry management techniques. And, as we have seen, demand for timber is fundamentally linked to the real growth of overall GDP. This implies highly volatile prices. This is just what we found in global data recently made public by the International Monetary Fund. Between 1981 and 2005, the average annual change in real softwood prices (expressed in U.S. dollars) was 0.5%, with a standard deviation (a measure of dispersion around the average) of 9.4%. In contrast, real hardwood prices grew by 1.1% per year on average, but with a standard deviation of 22.3%. When the two are combined into a single global index of timber prices, the 1981 to 2005 period saw an average annual real timber price change of 0.3%, with a standard deviation of 12.1%.

So, let us sum up: at the level of a global asset class, investors face moderate physical risks, but higher business risks, driven principally by price volatility.

We estimate that investors should demand a risk premium of four percent over the current yield on real return government bonds for bearing this risk (assuming an investment in a liquid timber REIT), which is the same as the risk premium on domestic equities and commodities. We should also note that this risk premium implicitly takes into account the quite low historical correlation of real returns on timber with returns on other asset classes, which reflects the partly biological nature of the return generating process. By adding the current yield on real return government bonds (2.4%) to our 4.0% risk premium, we estimate that investors today should demand a real return of at least 6.4% for investing in timber via liquid real estate investment trusts.

The next question is what returns these REITs are likely to provide. We estimate the likely supply of returns by adding to the current dividend yield plus the rate at which dividends are likely to grow in the future. At the time this is written, the current weighted dividend yield on the two timber REITs we use as proxies for the asset class (Plum Creek, PCL, and Rayonier, RYN) is equal to 4.3%. The key uncertainty is the rate at which these will grow in the future. One way to estimate this is to assume that the market for timber investments is operating in equilibrium, and all assets are fully (but not over or under) valued. Since this implies that the supply and demand for returns are equal, you can derive the expected dividend growth rate by subtracting the current dividend yield from the sum of the real return bond yield plus the asset class risk premium. In this case, the calculation is 2.4% + 4.0% - 4.3% = 2.1%. Note that this dividend growth rate reflects a combination of both physical volume growth and real price changes.

Another way to estimate the expected dividend growth rate is to use historical data. Between 1980 and 2005, real world GDP grew by an average of 3.5% per year. Assuming the U.N. Food and Agriculture Organization's estimate is correct, and wood demand will grow by 2.0% less than GDP, the expected physical growth rate is 1.5% (3.5% - 2.0%). If we add the historical average real timber price change of .3% to this, we get an expected dividend growth rate of 1.8%, and an estimate of future returns supplied of 6.1% (4.3% dividend yield plus 1.8% dividend growth). Because this is less than our estimated required return on timber of 6.4%, it implies that timber is slightly overvalued today.

However, if we use the IMF's forecast that real world GDP will grow by about 4.8% per year over the next few years (e.g., reflecting faster growth in India and China), the implied physical growth rate is 2.8% (4.8% - 2.0%). Adding the historical .3% real price growth to this produces an expected dividend growth rate of 3.1%. In this case, you would conclude that timber was undervalued today, since the return the asset class is expected to supply (4.3% + 3.1% = 7.4%) is greater than the return investors demand (2.4% + 4.0% = 6.4%). Of course, an important caveat here is that this forecast is the IMF's "most likely" case, and risks of different outcomes are not evenly distributed around it, with relatively more of them lying on the downside.

On balance, we conclude that timber as an asset class is probably not overvalued today, and may, depending on one's point of view, still be slightly undervalued.

Timberland investment managers may raise some criticisms about this analysis. Specifically, they will point to the potential for earning higher real returns through the application of active management techniques (i.e., they might argue that our estimate of the future supply of returns is too low). Example of these active management techniques include purchasing timberland at attractive prices, wisely choosing when to sell it to maximize the price received (known as "storing it on the stump"), and generating additional income streams from the land on which the trees are growing (e.g., by charging fees to hike, ski, and snowmobile on it). Some of these arguments have more merit than others.

Clearly, if all timberland investment managers were smarter than the companies from whom they purchased timber assets, then the average return on the asset class itself should be higher. However, while this may have been the case in the past (e.g., when integrated forest products companies were selling their timber holdings for less than they were worth), it is probably not true today, and won't be true in the future. Timing timber sales is an easier argument -- while some sellers will wait to sell at higher prices, and in fact receive them, others will sell today and be disappointed in the future (or, perhaps, happy, if prices subsequently go down because of an economic slowdown). In aggregate, these individual wins and losses will balance out, leaving only the price change driven by supply and demand factors for the asset class as a whole. On the other hand, it is clear that generating additional fees from the recreational use of timberland (or, for that matter, its sale for commercial development) is not a zero sum game. On that basis, our estimate of the returns timberland as a global asset class is likely to supply is conservative.

A final argument is that timber is not a single global asset class, but rather a series of national or regional asset classes, where, for example, timber demand and price growth may be higher than we assume. This argument certainly has some merit; for example, many U.S. timber managers point to average real long-term price changes of 2% per year in that market. On the other hand, the fact that many large institutional timber investors have diversified their portfolios across regions (e.g., owning timberland in the U.S. South, Canadian and U.S. Northwest, Chile and New Zealand), as well as the fact that global trade in timber has been growing faster than world GDP, suggests that our view of timber as a global asset class also has merit.

That being said, the fact remains that there is not (so far) a liquid global timber index product available today. As a result, in our model portfolios we use a second-best solution: a mix of two large liquid timber REITS (PCL and RYN) that own a range of timber properties in different locations.

June, 2005

The fact that timber is not included in most futures-based commodity index funds has led many investors to ask if it should be treated as a separate asset class. The answer from a growing number of institutional investors is, "yes." To begin with, timber has a unique three-part return generating process. Current income is provided by cutting and selling the timber. Over time, timber demand tends to grow with real gross domestic product, while timber prices have historically tended to rise at a rate somewhat above inflation.

Capital appreciation on a timber investment comes from the appreciation of the land itself (e.g., as it becomes more valuable to housing developers), and from the natural growth of the trees. Measuring the historical returns and risks on timber is somewhat difficult because of the absence of a standardized tradable product that covers the whole asset class. As a result, different methods have been used to construct "synthetic" indexes. For example, the National Council of Real Estate Investment Fiduciaries is a group of institutional investors who directly own timber and other commercial properties. Their index is constructed on the basis of actual returns their members have earned on timber properties in the United States. In contrast, the Hancock Timber Resources Group uses an econometric approach based on timber prices to estimate returns on timber not only in the United States, but also in two other key production areas, British Columbia and New Zealand. The good news is that both deliver similar estimates of the risk and return for this asset class. Between 1989 and 2004, the average real U.S. dollar return on the (U.S. only) NCREIF Timber Index was 10.67%, with a standard deviation of 8.76%. Between 1963 and 2002, the average real U.S. dollar return on the global Hancock Timber Index was 9.29% with a standard deviation of 12.40%.

However, you cannot invest in a timber index; you can only invest in companies or trusts that own timber, which exposes you to alpha in addition to beta risk. For example, between 1990 and 2004, Plum Creek Timber (PCL), which owns a well-diversified group of properties in the United States, delivered average annual real returns of 25.34%, with a similar standard deviation. However, in New Zealand, the three major timber companies (Carter Holt Harvey, Fletcher Forests, and Evergreen Forests) all had negative average real returns over the past ten years (Fletcher has consequently sold its timber properties, while the other two are reducing their holdings). Other possible timber investments include Rayonier (RYN), Deltic (DEL), TimberWest (TWF.UN), and West Fraser Timber (WFT.TO).

With that important caution in mind, let's take a look at the real returns on timber over the 1989 -2004 period in different currencies, and their correlation with returns for other asset classes over this period.

As you know, we use two different tests to define an asset class. First, it must have a return generating process that is substantially different from those of other asset classes. Second, its correlation of returns with other asset classes must, on average, be less than .65 (which leads to asset allocation solutions that are relatively insensitive to small changes in expected returns). We conclude that timber meets these tests, so we will include it as a possible asset class when we update our model portfolios later this year. In the meantime, we are also including year-to-date nominal returns for timber in our global asset class returns summary. This return is a weighted (70/30) combination of the year-to-date returns on Plum Creek Timber (PCL) and Rayonier (RYN).