Should your company build a new factory? Make a big acquisition? Sell off a division? Like most senior managers, you rely, quite appropriately, on projections of future cashflows to drive such important strategic decisions. These projections, in turn, are often based on the elements of cashflow—such as investment and operating costs—that are relatively easy to forecast (see Exhibit 1). This holds true even in industries such as electronics where costs are very much in flux. But the usefulness of such projections is often limited by their failure to think deeply about the one element of future cashflows that has the greatest overall impact on projected investment returns: price.

Although price forecasts are critical to major strategic decisions, few CEOs are truly comfortable making them. Many rely on hopeful assumptions about future prices, sometimes made in constant and sometimes in current dollars, without recognizing what effect their assumptions have on projected investment returns. Often they decide to keep things even simpler by assuming, implicitly, that future prices will stay where they are today, or that historical price trends will continue.

Such assumptions can produce estimated rates of return that are way off the mark. A few years ago, a new paper machine constructed in the southern United States would have been likely to earn a 13 to 15 percent return if prices remained constant (in constant dollars), but only a 0 to 5 percent return if they followed long-term downward trends. A top management team that relied on the first assumption would probably have decided to go ahead with the machine; a team that favored the latter would not.

Given such uncertainty, can price forecasts be made with assurance? Many CEOs doubt it. When pushed, they will admit something like, "No one can really be certain about what will happen to prices, because they are determined by unpredictable developments in demand, capacity, and cost. Maybe an economist in an ivory tower can get these relationships theoretically correct, but such solutions aren’t realistic or practical enough for us."

My answer to the question, however, is a confident, but qualified, yes. It is possible to forecast price with enough certainty to make good strategic decisions. Are these forecasts 100 percent accurate? No. Do they need to be? Again, no. A reasonable goal should be to help you identify the likely range of future price levels, to understand the economics of your industry better, and to pinpoint the causes of uncertainty. And the most important point of all may be this: if you decide to keep things simple and do not investigate prices in depth, you are still making a price forecast—but there is a big risk it could be wrong.

Why forecasts matter

As you know all too well, companies that spot long-term industry price trends and act on them can make a lot of money. Cruise lines such as Carnival, for example, probably understood that as prices for cruises fell relative to prices for other vacations, demand would boom and, along with it, the opportunity of earning attractive profits from economies of scale. Investing ahead of the trend indeed proved lucrative. In 1987 Carnival’s founder was able to sell 20 percent of the company to the public for about $400 million.

Companies that miss these trends, however, often make decisions that destroy shareholder wealth. Several producers in the US kraft newsprint industry failed to recognize a long-term decline in both demand and price for their products, and continued to add capacity. Not long after, newsprint recycling became mandatory across much of the United States. As a result, those companies’ investments will earn over time only 0 to 5 percent returns, far below their cost of capital.

Cyclical industries present a different kind of opportunity to exploit price movements. The easy decision is to invest when times are good and cash is readily available. But if several competitors make such investments at the same time, they can destroy the market for everyone. The harder—but more profitable—choice is to invest at the bottom of a cycle.

The petrochemical companies that began constructing new plants near the bottom of their industry cycle in the 1980s brought their capacity on line just in time for the cycle’s high point. Consequently, they will earn returns of 13 to 15 percent. By contrast, those that invested when times were good saw their new capacity arrive too late, after prices had already turned down. Moreover, that unwelcome capacity may itself have caused a further decline in prices. These companies will earn only 8 to 10 percent on their investments—significantly below their cost of capital.

Investments in market pulp capacity follow a similar pattern. Capacity additions that were planned when times were good, but that came on line near the bottom of the pulp cycle in the 1980s, will earn 11 to 12 percent returns. Additions planned when times were bad, however, started production near the top of the cycle and will likely earn 17 to 18 percent returns over their service lives.

As these examples show, good price forecasting can make the difference between creating and destroying shareholder value. No one, of course, can reliably predict prices with anything like 99 percent accuracy. No one has a crystal ball, and no one can build a model that accounts for all variables. But in most industries it is possible to develop a solid understanding of price—sometimes enabling you to make confident forecasts of the future, and at other times giving you at least a good understanding of the range of possible price movements.

What influences prices

The first step, naturally enough, is to build a sound understanding of the main factors that influence prices. Imagine, for example, that you are a cruise line executive trying to decide whether to invest $200 million to construct a new ship for the Caribbean market. Any estimate of investment returns will depend on forecasted cruise per diems. But these, in turn, will depend on a complex combination of factors—among them, the mix of different ships, itineraries, consumer preferences, variations in levels of consumer awareness, and the impact of broad economic forces on the cruise market. Making sense out of this kind of complexity may seem intimidating, but it can be done by:

• Conceptualizing the economic drivers of price. The goal is to develop hypotheses about the aspects of supply and demand that interact to define price.
• Making simplifying assumptions that permit these drivers to be quantified. It is simply not possible to do a "perfect" analysis that incorporates every variable; the challenge is to design a model that is "good enough"—one consistent with available industry data, but not so elaborate in its causal linkages that you can never quantify them. When this simplification is done well, the results will closely track the way the real world operates.
• Quantifying these drivers and then using them to verify or modify the initial price estimates—an effort that will often demand considerable data collection and hard work.

These three activities are more or less sequential, although in practice your thinking will move back and forth between them. In the cruise line example mentioned above, your first task would be to consider the possible drivers of cruise per diems. Had you been performing this assessment in the mid-1980s, you would have learned that the number of cruise ships, their size, advertising-driven awareness of them, and their popularity with holidaymakers had all been growing rapidly for some time. Cruise per diems, however, had grown slowly in nominal dollars and had actually fallen compared to the prices of other types of vacation in the Caribbean or Florida.

Given these factors, you would project that a new ship, at current per diems, would earn a highly attractive investment return over the next 30 years. A reasonable first-cut hypothesis would, therefore, be that the cruise market had grown because of the decline in relative prices, and that these prices, as well as the increased spending on marketing, were a consequence of the industry’s need to fill up its large new ships.

In order to prove or disprove this hypothesis, you would then want to make simplifying assumptions so that these facts could be quantified without sacrificing understanding. It might make sense, for example, to exclude the most luxurious ships (for example, upscale sailing ships) or to ignore differences in itinerary, because itineraries can easily be changed. The remaining assumptions must be based on numbers that can actually be collected and verified. If industry averages were not available, you might use your own company’s per diem trends. The point, remember, is not to be 100 percent precise, but only to get reasonably close.

Finally, you would use the data to verify your initial hypotheses—or to identify alternatives. The key here would have been to establish the relationship between volume and price so that it could be plotted on a demand curve. By using such a curve, you could estimate how far future prices would have to fall in order to fill up new ship capacity as it entered the market. And this information, in turn, would provide the basis for forecasting prices. What you would have learned is that the likely price levels would have more than justified buying the ship, and you would have proceeded confidently to buy one new ship, and perhaps several.

In much the same way, suppose that your paper company is evaluating a billion-dollar investment to build a new commodity white paper mill. Much will depend on your forecasts of future paper prices—which will, of course, be determined by the interaction of various factors: the costs associated with production in different countries and with different technologies; the ability of paper machines to make commodity as well as specialty papers; the range of products available in the market; and so on.

To derive and test useful hypotheses about the economic drivers of price, you need to build an industry supply curve like that shown in Exhibit 2, which can then be used—along with current demand levels—to check the accuracy of your predictions against current price levels. In building the curve, you again have to make simplifying assumptions to determine where corners can be cut without harming the insights generated by the forecasting model. For example, the curve might still be reliable even if costs are estimated for only a few of the world’s mills (on the assumption that mills of the same type have similar costs), or if only one grade of paper is considered.

Getting these simplifying assumptions right is a tall order, but very rewarding. Several years ago, a streamlined model of the paper industry forecast prices per ton for 20lb bond rolls at US$610, compared with actual published prices of $612 (Exhibit 2)—an extremely close match in an industry where yearly price fluctuations can be $50 to $100 per ton or more. This accuracy created the confidence to use the model to explore how different events might affect paper prices in the future. The model showed, for example, the point at which a rise in demand would boost prices sufficiently to allow idle high-cost mills to restart and earn a profit (Exhibit 3, situation A). It also showed the point at which increased demand could not be met without the addition of new capacity—and at which prices would have risen high enough to make attractive returns possible on that capacity (situation B).

The model showed too how much some changes—for instance, a jump in energy costs—would lift prices (situation C), at least after an initial period of settling in, and how much other changes—say, a fall-off in demand—would depress them.

A special challenge is presented by industries where demand and/or supply are inelastic—industries where prices are inherently volatile because it is hard for demand or production to respond to price changes, or takes a long time. The world oil market in 1986 is a good example. The Saudis increased production to discipline other world oil producers. They expected prices to fall from $26 per barrel to around $17, but in fact prices plummeted to $9.50. Why?

First, demand was inelastic. Even when prices were much lower than usual, people had no practical way of using their cars significantly more or consuming much greater quantities of energy in industrial settings.

Second, production was inelastic. Most oil-producing countries increased production as prices fell to maintain their total revenues and support their national economies. While some high-cost producers shut down (in West Texas, for example), their impact on total world production was small.

If your industry, like oil, is inelastic, your need to forecast prices will be high because prices can move dramatically in a short timeframe. Equally high is the need for scenarios to reflect the inherent uncertainties in some of the drivers of price levels. It is no surprise that Shell was an early leader in the development of scenario-based planning.

Assessments like these will crack the case in many industries, but not all. Industries such as steel, with numerous competitors and little product differentiation, are commodity-like—that is, their more or less "perfect" competitive market will cause prices to adjust broadly in line with shifts in supply and demand. In others, like branded consumer products and computer software, the markets are highly segmented and thus the reaction of prices to supply and demand changes can be less direct and immediate.

In industries where there are only two, three, or four competitors, companies may set their own strategies and prices with the reactions of their competitors in mind. In such cases, prices may remain high even as demand falls. If you are in this kind of industry, you will also need to assess, perhaps through the use of game theory simulations, both the effects of structural variables like industry concentration and the probable outcomes of dynamic competitive interactions.

In the US airline industry, for example, fare levels tend to fluctuate in response to different industry-wide competitive situations. At the same time, however, there are big differences in fares between routes of a similar distance, depending to a large extent on the pricing strategies and market positions of the airlines competing on each route.

A published report by America West, for example, compared fares for the Chicago-Cincinnati and Las Vegas-San Diego routes, which are of roughly equivalent length. The airlines competing on the former are all traditional carriers with a hub in one city or the other (American and United in Chicago and Delta in Cincinnati). The average fare they charged on this route was $187. On the Las Vegas-San Diego route, however, the average fare was only $34, largely the result of heavy price competition between two very low-cost carriers (America West and Southwest).

How to forecast

Even when you have a solid understanding of your industry’s current price equation, there remains the question of predicting the behavior of key price drivers. Of course, no one can really know the future, and the likely trajectories of price drivers will be harder to foresee in some industries than in others. But it is possible to push back the borderline of uncertainty.

If the question is whether to build a new cruise ship, the place to begin, as we saw, is the analysis of what drives current prices. The next step is to forecast future capacity levels—by, perhaps, finding public reports about ships under construction during the next two or three years—and then to plug this new capacity into the model you have already developed. This, in turn, permits estimates of how far prices (in constant dollars) are likely to fall by the time the new ship arrives on the market.

So far, so good; but these numbers apply only to the forecast of cashflow during that ship’s first year of operation. Also needed is an understanding of how prices might behave later on. One approach to estimating the future price floor is to specify the price levels below which cruise lines could not earn attractive returns on a new ship. This price level could then serve as the background to scenarios embodying different assumptions about how rapidly these low price levels might be reached and whether prices might fall even lower.

Finally, designing a cashflow model for each of the scenarios will yield a confident estimate of the likely returns on the investment in a new ship under each set of circumstances. The goal, then, is not a single point estimate of price, but rather a confident understanding of the risks and opportunities and their associated returns.

A similar analysis is possible for new investments in industrial plants, such as a paper mill. First, quantify competitors’ incentives—that is, the returns they are apt to see from expanding existing mills or building new capacity in different parts of the world.

Next, use your estimates to forecast where in the business cycle competitors might feel justified in adding capacity. It might be a good idea to collect some historical data on when different companies made capacity decisions in the past in order to inform your judgments about how they are likely to behave in the future.

Then you need to assess long-term shifts in demand levels to estimate the point in the business cycle at which prices are most likely to move in response to demand—and, even more important, whether demand will continue its growth trend over time. If it does, new capacity that misses the peak of the cycle may be justified by a rebound in demand at a later time—or at least you will not actually lose money on the investment in the long term.

When managers in the canning industry were considering whether to introduce new canning technologies, the main drivers of price were all related to the economics of replacing existing technologies. By quantifying the impact of the new canning technology on cost, the reception of different can shapes and sizes by retailers, and the reaction of consumers to new cans, the managers assessed how food companies would view the new cans relative to those currently in use. They could then forecast whether future price levels were likely to be high enough to justify the capital investment required to introduce the new cans.

Similarly, some years ago a company had to decide whether it should enter the computer systems operations business—whether to start a new business to compete with companies like EDS in taking over and operating large companies’ data centers. The key driver of future prices was again substitution economics: how much money could be saved by using big, efficiently operated data centers. The company estimated its own operating cost and compared it with the costs of operating a number of different corporate data centers. By focusing on that half of the potential market that had above-average costs, it was able to determine what the total potential saving was, and thus what price range it could expect to realize in the new business.

When CEOs operate in industries they know well, they can, with a disciplined effort, understand the behavior of variables like these. In some situations, this will be enough: their understanding can inform major value-creating decisions. The tougher challenge by far is to understand the effects of factors that lie outside their industries, for example, aggregate macroeconomic performance, foreign exchange rates, or major technological changes—all of which can significantly influence prices in different industries.

Even here, though, you can develop a useful view of how changes in these factors will affect industry pricing. When it is simply not possible to develop definitive forecasts of all key price drivers, you can still improve the quality of your strategic decisions by profiling ranges of uncertainty, estimating the probability of different scenarios playing out, and tying both the estimates and the uncertainty profiles to your more reliable forecasts of the other determinants of price. You will not be predicting the future, but by working to appreciate the full range of possibilities, you will be able to make more richly informed decisions. Regards,