Return Attribution

One way to buy a business is to have a thesis. A thesis on why this particular company is a good deal. Why its advantages make it a good investment. Why changes you make to it will increase its value. But most times, we do not think enough about the durability of our thesis – how long will it be correct? To be right for a little while is not enough. The advantages of a company and improvements made to it have to last far into the future. After all, a business is only worth the present value of future cash flows.

If we do not think carefully enough about the durability of advantages and improvements, we tend to overstate their lifetime – we overestimate returns. Financial modeling makes it especially easy to exaggerate the durability of advantages and consequently misconstrue the long-term value. Analysing modeling techniques and human biases can help identify the root causes of these prediction errors. Further, two examples, an LBO model and a discussion on value investing, clarify how these mistakes go unnoticed and can shine a light on the challenges of correctly attributing returns.

A look at models used for business valuation further stresses the importance of evaluating the durability accurately. Most funds will create discounted cash flow and LBO models for an acquisition as part of their valuation approach. For these models to be of any value, reasonably accurate predictions of financials for the next 5-10 years are needed. 

The structure of such models indicates that advantages need to be durable for many years so that a fund can meaningfully profit from them. After all, without intervention (i.e. cetis paribus), a business with unsustainable advantages would experience sudden drops in revenue and earnings. Furthermore, unless improvements to a company through a fund’s operational experience are lasting, these improvements will only contribute to profitability in the short term. Lastly, prospective buyers will value the business, in part, using the same financial models when the time to exit comes. So they, too, have to believe that advantages will last throughout their holding period. Hence, there is a need for durable moats around investments that are sustainable for at least 5-10 years. Most investors implicitly understand how long their moats will last. Unfortunately, financial modeling makes investors easily susceptible to overstating this durability.

For instance, when modeling an LBO, a one-time improvement may be split up across the holding period. Tweaking the numbers allows the model to arrive at the same outcome as if the improvement were modeled as one-time only. However, while the final value has remained the same, the nature of the advantage or improvement has changed drastically. A one-time improvement is similar to multiplying a metric like free cash flow or enterprise value by a factor representing the improvement. By splitting up the improvement across the holding period, the improvement contributes to the growth of these metrics every year, behaving like a change of basis in a compounding process. Splitting a one-time improvement’s contribution over an investment’s holding period is presented as an indefinitely durable increase in the growth rate. 

There is often a valid reason for splitting one-time improvements up – for instance, they are frequently realised over a few years. Margin improvements or multiple expansion (when reasonably assured) are two such examples. Changes that improve margins or exit multiples occur over many months or years. Furthermore, splitting the realised benefit of some one-time improvements yields a more precise prediction of free cash flow during the time these changes are made. 

Regardless of how such changes to businesses are modeled, investors are implicitly aware of the limited durability of specific advantages and improvements to the underlying companies. After all, margins can only grow so much. However, according to our human nature, we tend to extend trends that cannot continue indefinitely outwards – we overestimate their durability. This tendency leads to an overestimation of revenue and free cash flow growth rates once these improvements no longer contribute to growth. Consequently, our models overstate expected returns. This effect occurs even when the contributions of such improvements are modeled to fade away over the holding period – the issue merely becomes more subtle. The longer outwards we project returns, the larger this problem grows. Furthermore, our tendency to use the first presented value as an anchor makes downward adjustments for marginal, long-term growth rates hard once we have sweetened the deal by including margin and multiple expansion in IRR calculations.

Finally, let us examine how well durability overestimations hide and how challenging they are to untangle from realistic return estimation.

First, consider the case for multiple expansion. While multiple expansion may often be financial engineering, let us regard a valid case: A fund engages in an LBO and models multiple expansion to arrive at an expected IRR. The fund assumes that sustained improvements to the business resulting from their operational experience will lead to higher free cash flow growth rates. Further, assume this growth rate will likely stay elevated for more than five years after the fund exits the investment.

Since high growth rates increase the expected free cash flow an acquiring company will earn in a few years, such growth should be of some value to the buyer of the business once the fund decides to sell. In particular, margin expansion often is one of the drivers of such increases in free cash flow. Splitting the margin increase up and distributing it evenly increases expected free cash flow growth rates by a constant amount each year. However, in reality, assuming successful operational improvements, margin growth and hence free cash flow growth rates will primarily be realised at the beginning of the holding period. The growth rate will likely plateau towards the final years, making the investment appear less attractive to buyers – especially as holding periods increase. This reality is not adequately captured in the model and will likely not allow the fund to exit at the desired multiple. 

By overestimating the durability of an advantage and attributing the improvement to the wrong period (evenly across the holding period instead of at the beginning), final growth rates are estimated to be too high. When the time to exit comes, actual growth rates will already have come down, making the investment less attractive to buyers. A failure to meet projected exit multiples will be the most likely outcome.

For a second example, consider value investing. In value investing, the expected return is a compound return of the price moving closer to intrinsic value and the change of intrinsic value as the company grows revenue and earnings. Once the price has moved closer to intrinsic value, only one source of returns remains. Therefore, an investor holding the asset should expect lower returns after this point in time. Once prices have moved to reflect intrinsic value, expected IRR that considered both drivers of returns no longer accurately reflects expected future IRR – it overestimates it.

Since it is especially hard to estimate when prices will move to reflect intrinsic value, determining when the price movement toward intrinsic value should be accounted for is very hard. Attributing this source of returns to the wrong period or splitting it up leads to errors in return estimation, as previously. At best, a catalyst might give the investor the conviction to expect a price correction soon. At worst, one has little knowledge of when prices will reflect intrinsic value and in which period to account for this driver of returns. Such uncertainty makes distributing this source of price appreciation evenly across the expected holding period even more alluring but no less false.

Business valuation relies on determining the durability of moats around a company. Overestimating the sustainability of a business’s advantages and improvements can lead to incorrect estimates of returns and the business’s current value. This implies a higher long-term growth rate or return where it should not be expected. Modeling margin expansion and price movements towards intrinsic value in value investing illustrate the challenges in arriving at reasonable estimates. Furthermore, our tendency to use initial values as anchors makes correcting errors hard once we overestimate the longevity of a business’ competitive edge and implemented improvements. Increased awareness of these challenges is the first step to conservative and superior estimates.