Sunday, March 29, 2015

What's driving the trend towards intangible value?

The trend for the market values of firms to increasingly diverge from book value, and particularly from tangible book value, is really striking. As I wrote in my previous post on inflation, this is a trend that merits more attention.

So I was happy when I saw Justin Fox's piece for Bloomberg bringing more attention to this pattern. He presents an updated chart using data from Ocean Tomo (which I reproduce here from the original source)



Here's an older chart which also breaks out tangibles (gray) as well as the subset of intangibles that are on the books (brown, mostly goodwill):




It is important to understand exactly what the charts are showing. "Intangible assets" on the first chart is just a residual, the part of aggregate market capitalization that is in excess of aggregate tangible book value. (This is the red plus the brown on the second chart). It seems investors are increasingly willing to pay a lot more for a firm than would seem justified just by looking at the tangible "stuff" on the firm's balance sheet (including property-plant-equipment, inventories, and net financial assets including cash.)

I liked Fox's piece, but I wish his focus had been a little broader. "Most of their value," he writes, "comes from brands, patents, ideas and other intangibles....the modern corporation really is a different, much less bricky-and-mortary creature than its predecessors." Reading this, one thinks of firms like Google and Facebook, that have a lot of valuable software but only a relative handful of highly-skilled employees.

But it's too narrow to think about this just in terms of R&D and intellectual property and "ideas". Other important intangibles are scale, existing relationships with customers and suppliers, and "organizational capital," i.e. the value of having in place a large organization of skilled and trained employees with established procedures.

 In fact, by the measure shown in the figures, Google and Facebook are far from the most "intangible" corporations in the S&P500. Both have positive tangible book value. This would still be true even if you removed their cash.

By contrast, there are over 100 firms in the S&P 500 that have negative tangible book value. In other words, by the measure presented here, more than 100% of their market value is "intangible." The most strongly negative firms tend to be communications companies like Verizon or Time Warner Cable. But there are all sorts of firms on the list. A big part of this is that some firms have a lot of debt, but there is more going on here as well. (Interestingly, financial firms tend to be the most "tangible" by this measure.)

The case of United Rentals

Lets look at one of these negative book value firms: United Rentals (URI). With a market cap of around $9 billion, it is one of the smaller firms in the S&P500, having been added to the index in 2014.

United Rental's business is to rent out heavy equipment and other tools. Customers  include "construction and industrial companies, manufacturers, utilities, municipalities, homeowners, government entities." As of 2014, URI had 881 rental locations in the US and Canada, all filled with big expensive machines available for rent. The firm owns over $6 billion of property-plant-and equipment (about $60 per share). It has over 12,000 employees, more than Facebook. It would seem this firm is very, very brick-and-mortary indeed.

URI stock currently sells for about $90 per share. They are profitable, with a reasonable P/E of 17.5. But they also have over $80 per share of debt, and a tangible book value of equity of approximately negative $25 per share.

All in all, the tangible assets on their balance sheet are worth $2.6 billion less than their liabilities. Why is their stock worth 9 billion dollars? Why is anyone willing to lend them money (which they then use to repurchase their seemingly overpriced shares)? Where does the extra firm value come from?

The 2014 annual report gives an answer in the form of a list of "competitive advantages":

  • Large and diverse rental fleet
  • Significant purchasing power
  • National account program (i.e. relationships with large customers)
  • Operating Efficiencies. (i.e. "Equipment Sharing Among Branches," "Customer Care Center," and "Consolidation of Common Functions.")
  • Strong Brand Recognition. 
  • Geographic and Customer Diversity. 
  • Strong and Motivated Branch Management. 
  • Employee Training Programs. 
  • Risk Management and Safety Programs. 

You could summarize these advantages as "scale" and "organizational capital." (There is some notion of "brand" there as well, although I doubt that a large customer really cares about "brand" in the same way a 13-year old cares about Under Armour, so maybe "reputation" would be a better word.)

Of course scale and organizational capital are not new. But the evidence suggests they may be more important than ever. Why? Perhaps information technology and increased regulation are factors that favor scale. Perhaps the relative decline in the price of manufactured goods relative to wages has inevitably made tangible goods a smaller portion of firm value. Perhaps the workforce (or a portion thereof) is more skilled, but effectively using these skills requires building complex organizations, with a large element of learning by doing.

I don't know the answer, but it's certainly an interesting question.

Sunday, March 15, 2015

Are earnings yields real or nominal? How Inflation can affect the measurement of accounting earnings

This post is inspired by a twitter conversation I participated in last month that started when pseudonymous blogger/tweeter/super-valuation-expert Jesse Livermore posed an interesting question:



At issue is whether stock earnings yields should be considered real or nominal.[1] We all know that with bonds we need to inflation-adjust the nominal yield to find the real yield. But do we need to do any sort of inflation adjustment for earnings?

Jesse Livermore’s answer is no, and this seems to be the default assumption. In an ideal world, accounting earnings would measure sustainable real economic income, and would not have to be adjusted for inflation. The common practice of comparing earnings-based measures of valuation between eras with very different inflation rates relies on the assumption that accounting earnings are real and not nominal. (Note that the inflation adjustments used in calculating the Shiller CAPE do not address the issues treated here.)[2]

But I don't think inflation can ignored so easily; as I will explain, inflation can cause a “distortion” (for want of a better term) in earnings measurement. Furthermore, I will explain how we can attempt to approximately quantify this distortion, for a firm or for a market, using the following earnings adjustment equation:

Inflation adjusted earnings ≈ reported earnings – inflation * book value of tangible equity (t-1)

The basic idea (as discussed in detail below) is quite simple: the nominal value of items on a firm's balance sheet will tend to increase over time just to keep pace with inflation. This purely nominal increase will be reflected as net income, due to the clean-surplus nature of accounting. This causes earnings to be overstated relative to a situation with no inflation.

The adjustment does not require that the book value perfectly measure the true value of the firm. For example, assets are carried at historical cost. Still, if the firm has a relatively stable business model, such that the fraction of true firm value that appears on the balance sheet is constant, then the balance sheet will tend to grow along with inflation, resulting in a portion of reported net income that is purely nominal.

If we take the earnings adjustment equation and divide through by the market value of equity, (and ignore the “t-1” lag), we get the yield adjustment equation:

adjusted earnings yield ≈ reported yield – inflation * tangible-book-to-market ratio

In the US market today, inflation is low, and most firms (especially non-financial firms) have tangible-book-to-market ratios much closer to zero than one, so the adjustment is fairly negligible. But this has not always been true in other times and places. In the 1970s, inflation was above 5% and tangible book to market ratios were well over 50%, implying that inflation could have a meaningful distortionary effect on earnings yields. I suspect this may be an important factor underlying the seemingly low valuations and slow earnings growth in the US market in the 1970s and 1980s.

I will now walk through three highly simplified examples to show how the inflation distortion occurs, then discuss some complicating factors and briefly discuss the implications for valuation. [3]


Example: All Cash


Start with the simplest possible example. A firm starts the period with no liabilities, and only one asset: $100 in cold hard cash. Thus the firm also has $100 of shareholder equity, all of which is “tangible.” Assume the firm conducts some business during the period, resulting in $X in total sales, and $X-10 in total cost of sales plus all other expenses. The income statement is:

Sales                                                                X
Cost of Sales + Other Expenses                 X-10
Net Income                                                     10

Accounting profit for the period is $10. Assume no dividends, stock issues or buybacks, borrowing, or lending. Then at the end of the year the firm again has only one asset, $110 in cash. Likewise shareholder equity has increased to $110, with a $10 increase in the retained earnings account balance.
Now assume we are told that inflation over the period was 10%. Then the $110 cash the firm has at the end of the period is worth exactly the same as the $100 in cash it started with. The $10 gain was purely nominal, real gains were zero. There are no economic profits.
Note that we do not have to worry about the details of the business, how inflation affected each individual transaction that went into X, or even how large X is. Instead we can simply focus on the balance sheet. The principle of clean surplus accounting tells us that net income over a period is equal to the change in the value of shareholder's equity over the period, assuming for simplicity (and without loss of generality) that there are no other transactions with shareholders (i.e. no dividends or stock issues/buybacks.)[4]
In this example, the increase in the value of equity and hence also net income, was a purely nominal increase. If we use the earnings adjustment equation, we see that inflation-adjusted profits were zero:
0$ = $10 – 10% * $100


Example: Inventories

Imagine a retailer that sells widgets. At the beginning of the period they have an inventory of 100 widgets that they purchased wholesale at $1/widget. Assume there are no other assets or liabilities, so the balance sheet just shows $100 in inventories, balanced by $100 in equity. During the period they sell all 100 widgets for $150 total, incurring $40 of expenses along the way (wages, overhead, etc.). At the end of the period, they buy 100 new widgets to restock their inventory. Assume inflation is a uniform 10%, so the new widgets cost $110. The income statement is:

                Sales                                                   150
                Cost of Sales                                      100
                Other Expenses                                    40
                Net Income                                           10

Observe that the firm is now in exactly the same economic position as before, with 100 widgets and no cash. They have shown net income of $10, and the inventory asset and retained earnings equity accounts on the balance sheet have each increased by $10. But there is nothing left over to pay to shareholders, and no real economic earnings.

Note that if inflation is uniform and affects all goods and services at the same rate, the firm could go on like this forever, with each line on the income statement and balance sheet growing at the rate of inflation, showing positive earnings period after period, but never generating any real increase in value or providing any payouts to shareholders.

The inflation adjustment equation looks exactly like the adjustment in the previous example:

                                0$ = $10 – 10% * $100

An important assumption here is that the inventory is carried on the books at replacement cost. Under US GAAP (but not IFRS), many firms instead use LIFO (last-in-first-out) accounting for inventory. This creates a so-called “LIFO reserve,” which makes current profits lower and more accurate in real terms, but makes the balance sheet less accurate. This effectively pushes the inflation distortion out into the future, creating artificially higher accounting earnings in the future periods when the LIFO reserve is eventually liquidated. This is an example of how inflation in one period can distort earnings in subsequent periods (the next example using capital goods will show another way this can happen.) Nevertheless, even with LIFO accounting, if the firm (or market) is in a steady state where the LIFO reserve is not growing as a percentage of firm value, the earnings adjustment equation would still work.

Example: Capital Goods

Now consider a firm that rents out trucks. Assume the only asset is a fleet of trucks, and there are no liabilities. The firm operates in a steady state, where every year the company buys exactly one new truck, and disposes of its oldest truck. Furthermore assume that the trucks are carried on the balance sheet at historical cost, depreciated straight-line over 5 years, and that inflation has caused the price of trucks (and everything else) to increase steadily for several years at a constant rate of 10%.  At the beginning of the period the firm has just purchased a new truck for $50,000. The asset and accumulated depreciation balances for the "fleet-of-trucks" asset would be derived as follows:


BOOK VALUE
Truck Age
Historical Cost
Begin Period
End
Period
Depreciation expense
4
34,151
6,830
-
6,830
3
37,566
15,026
7,513
7,513
2
41,322
24,793
16,529
8,264
1
45,455
36,364
27,273
9,091
0
50,000
50,000
40,000
10,000
Totals
`
133,013
91,315
41,699

Assume that the firm derives $100,000 in revenues from renting out its trucks during the year, and incurs other cash expenses totaling $45,000. Then the income statement is:


Sales                                                      100,000
Depreciation Expense                             41,699
Other Expenses                                       45,000
Net Income                                              13,301

At the end of the year, the firm has $55,000 in cash (sales minus cash expenses): just enough to buy a new truck at the new 10% higher price! Thus the firm is once again left in the same economic position as before it started, with the same number of trucks of the same ages, and no cash. Once again there is no economic profit, despite the reported net income of $13,301. Once again the inflation adjustment is equal to the inflation rate multiplied by the book value of equity at the beginning of the year.

                                0$ = $13,301 – 10% * $133,013

Note this does not depend on the depreciation schedule accurately representing economic depreciation, and the truck asset does not need to accurately measure the value of trucks. For example, the trucks may last for more than 5 years. But none of that matters for the inflation adjustment - all that matters is that the “trucks” asset on the balance sheet has the same relation to true truck value over time, implying that the “trucks” asset needs to grow at the rate of inflation just to keep the real value of trucks constant. 

Discussion


In the interest of simplicity, the previous examples assumed that real profits were zero, and there were no payouts to shareholders. We could easily modify the examples to relax these assumptions without changing the basic lesson.

I have focused on the balance sheet approach, because it seems simpler. But we can also describe these distortions in terms of items on the income statement. In the widgets example, the cost of sales is effectively understated, because it uses the historical cost of widgets rather than the replacement cost when the widget is sold. In the trucks example, the depreciation expense is understated relative to the case of no inflation since the depreciation is calculated based on percent of historical cost, rather than on current cost.

Liabilities

For simplicity, these examples have also ignored liabilities. Liabilities on the balance sheet act much like cash in the first example, except with the sign reversed. In a steady state, if the balance sheet is to grow with inflation, liabilities will also grow, but this growth in nominal liabilities makes reported earnings lower rather than higher. This is why the inflation adjustment is based on equity (assets minus liabilities), rather than assets.

One issue I ignore here is the one-time gains or losses in financial assets and liabilities that will result when inflation shifts unexpectedly. Also ignored is whether/how these gains or losses are marked-to-market on the balance sheet. I suspect that these complications mostly have the effect of shifting earnings between periods, rather than creating permanent distortions in earnings.

Changing vs Stable Inflation

The examples above assumed that inflation was stable over time, which greatly simplifies the analysis. In particular, when the firm has multi-period capital investments, the distortionary effect of inflation on earnings in one period is not confined to that period, but will play out over several subsequent periods.

Consider again the trucks example. Imagine that instead of constant inflation, we had a one-time increase in prices, with no inflation in other years. After this burst of inflation, the "trucks" asset balance would continue to increase each year for a period of five years, until the historical cost of all of the trucks still on the balance sheet caught up to the new post-inflation price of trucks. Thus each year earnings would be overstated until the nominal value of the "trucks" asset reached a new steady state.

This shows that if inflation is volatile rather than constant, the adjustment becomes more complicated. To use the earnings adjustment equation, the current inflation rate would need to be replaced with some weighted average of current and past inflation.

Intangibles

Much of the value of a modern corporation arises from things that do not appear on the balance sheet. These “intangibles” include things such as brands, organizational capital, relationships with customers and suppliers, trade secrets, accumulated R&D, etc.  The distortionary effect of inflation on earnings is driven entirely by growth in assets that are recorded on the balance sheet, because their nominal balance sheet value must grow over time to keep up with inflation (as described in the examples above). Although intangibles usually require investments to maintain their value, these investments are generally expensed immediately, and thus do not show up on the balance sheet at all.   Hence off-balance sheet assets, such as many intangibles, will be irrelevant for the inflation adjustment.

There is a subset of intangibles, however, that do appear on balance sheets (the majority of which is goodwill).  As with other intangibles, any investments needed to maintain theses assets will not show up on the balance sheet.  Thus my assumption is that on-balance-sheet intangibles do not need to grow in order to keep up with inflation. This is why the adjustment equation uses tangible book value rather than total book value.  

Note that even though intangibles may reflect a failure of accounting to fully reflect the value of the firm, this does not invalidate the inflation adjustment equation. It is still the case that the tangible equity portion of the balance sheet will tend to grow to keep up with inflation. 

Relative Prices

A big assumption in this analysis has been that inflation is uniform across all goods. This is, of course, not true in the real world. In particular the cost of manufactured goods has generally fallen relative to other goods and services, and this has surely had some effect on firm balance sheets. This raises the issue of which inflation measure to use, and more generally how to account for relative price changes. I will explain why we should use the overall inflation rate, rather than the price change that applies to the particular assets of the firm.

In a competitive market, anticipated changes in relative prices should not affect the real return on investment. For example, imagine the nominal price of trucks stays constant while overall prices rise. Because of competition, you would expect that this fall in relative price would be passed on to consumers in lower rental fees, such that the real effective return on investment remained constant. The lower relative price of trucks indicates they are actually worth less in real economic terms, and that the firm is likewise less valuable. The lower real value of the "trucks" asset would represent a real economic loss, but this loss is not reflected in accounting income.

In the real world, things might be a somewhat more complicated. First, changes in relative prices that are not anticipated might be expected to change the return on sunk investments, at least in the short term. But this effect can go either direction and should tend to cancel out over time.

Over the longer term, the fall in the relative price of manufactured goods would lead to an increase in the value of intangibles relative to tangibles. In fact the US market has seen a very strong trend in this direction, measured in terms of the share of market value that is captured by tangible book value (see figure 1). (The decline in the relative price of manufactured goods are likely just one factor underlying this trend. Other factors could include changes in technology, industry mix, accounting rules, etc.).



Figure 1




This is a really striking trend that might well have important implications for earnings measurement, and whether/how earnings yields can be compared across time. It probably merits more attention than I can give it here.

On the other hand, the factors causing this shift, including changes in relative prices, have been evolving for decades, and there is no particular reason to believe the trends were any stronger in the high-inflation 1970s than the low-inflation 1990s. The inflation adjustment discussed here should be taken as an adjustment for the overall inflation that is orthogonal to changes in relative prices. 

Other Accounting Distortions

The inflation adjustment here does not require that book value perfectly measure the value of the firm. Likewise, it does not require that accounting earnings perfectly measure economic earnings, if the purpose is to compare valuations over different times or places. For example, perhaps you believe that (for some reason) accounting earnings in the US market consistently overstate true economic earnings. The analysis here merely suggests that they were even more overstated in high inflation periods than in low inflation periods, and that this should be taken into account when doing comparisons.

A similar argument can be made for changes in distortions. There may be reasons to believe that accounting earnings today need to be adjusted for reasons other than inflation in order to make them comparable to earnings in the past. The inflation adjustment discussed here is simply in addition to any other such adjustments.


Implications for valuations

In the US market today, inflation is below 2% and the tangible-book-to-market ratio is below 20% (see figure 1). Therefore the inflation adjustment for the earnings yield will be small, perhaps on the order of 25 basis points. However, the situation was very different in past decades. In the 1970s, general inflation averaged over 5%, and the aggregate tangible-book-to-market ratio was something like 80%. Therefore the yield adjustment equation would suggest we should adjust earnings yields in the 1970s down by as much as 4 percentage points in order to make them comparable to current yields. This would correspond to a shift in the P/E ratio from 10 all the way up to 16.7. I would not be surprised if this is an important factor explaining both the low valuations and slow real earnings growth in the US market in the 1970s and early 1980s.

I don’t know much about emerging markets, but it would not surprise me if inflation might cause earnings yields to be misleading there as well. Proceed with caution.

Comments welcome.




[1] The earnings yield is the reciprocal of the price earnings ratio. We should expect a relationship between the real earnings yield and the long run rate of return, as explained in this wonderful note from Brad DeLong. 

[2]   To calculate CAPE (Cyclically Adjusted PE ratio), Shiller adjusts the price and earnings series to constant dollars, in order to allow combining earnings from different years to create a moving average used to smooth the earnings series. However this does not correct for any bias created in earnings measurement within a single period, which is the issue addressed here.

[3] The basic ideas here are not new. They were well known in earlier high inflation decades, but seem to have been mostly forgotten. However the simple earnings adjustment equations and the application to understanding historical valuations are not something I've seen before.

[4] Clean surplus accounting means that all changes in shareholder equity that do not result from transactions with shareholders (such as dividends, share repurchases or share offerings) are reflected in the income statement.” (See the link below). There are a few items for which clean surplus accounting does not apply, “most notably foreign currency translation adjustments and certain pension liability adjustments,” which are not included in net income, but instead are reported as part of “comprehensive income.” This are probably not too important for this analysis, and I will be ignore them.
 http://financial-education.com/2007/08/11/clean-surplus-accounting/

Tuesday, March 10, 2015

Teach For America - Statistical Insignificance Strikes Again

Once again, as with the Oregon Medicaid Experiment, a prominent study with a "statistically insignificant" result is being misinterpreted by almost everyone.

The headline in the Washington Post’s Wonkblog reads “Teachers in Teach for America aren’t any better than other teachers when it comes to kids’ test scores.” The piece reports on a new randomized evaluation comparing on the latest scaled-up cohort of TFA teachers with non-TFA teachers in the same schools. We are told the new study finds that students of TFA teachers don’t have any positive impact on student test scores compared to other teachers, and this this conflicts with earlier research finding that students of TFA teachers scored higher on math tests.

Of course, that’s not quite the study showed. And, of course, you have to go dive into the study itself to find out what is really going on, since the common practice seems to be that if a result doesn't reach the magical .05 p-value you don't bother to put the point estimate or confidence interval into the press release, or even into the executive summary.

Let’s focus on the differences in math scores. The new research found a point estimate of .05 standard deviations (SD) higher math scores for TFA teachers than for comparison teachers. The the standard error is .05 as well, so the 95% confidence interval is something like [-.05, .15].  By conventional standards we can’t rule out negative impacts as large as .05 SD, or positive impacts as large as .15 SD.

Is .05 SD a large difference? Is .15 SD? Well, it’s hard to say. I guess .05 seems kind of small. The study points out that this corresponds to a one-percentile point difference at the 30th percentile of a Normal distribution, (although I don't know how they got that, it seems like it should be closer to two percentile points. I guess there was some rounding). But anyway it's not nothing, and it's not as if we have a long list of other cheap, feasible methods lying around to allow us to get test score gains. 

The study does helpfully point out that the .12 SD impact they found in reading scores for younger children (which, by the way, was deemed statistically significant, contradicting the headline), corresponds to 1.3 extra months of learning gains. So maybe the estimates impact for math would correspond to an extra 2 or 3 weeks of learning gains, or at the high end as much as a month and a half, although the correspondence might be different for math than for reading. Anyway, not nothing. It might be interesting to know how these gains compare to other differences that have been found in research, such as the gains from teacher experience, or smaller classes. But the report doesn't address that.

As per usual, the study says little or nothing about what differences might be substantively important or achievable, and instead focuses almost completely on statistical significance. It does include this extremely important quote, the kind of thing that really should be part of the executive summary:
Statistical power. Our study had sufficient statistical power to detect moderate to large impacts on student achievement. Minimum detectable effects were 0.13 standard deviations for math and 0.14 standard deviations for reading. In other words, if TFA elementary school teachers truly improved student math achievement by at least 0.13 standard deviations (slightly below the 0.15 standard deviation impact estimate found by the prior elementary school study), there is high likelihood (80 percent) that our study would have found a statistically significant positive impact. 
Apparently the authors think 0.13 SD would constitute a “moderate” impact (how they came to this conclusion they don’t say). So it looks like in fact we can't rule out "moderate" impacts, let alone "small" impacts. In fact, we can’t even rule out the effect in this cohort being the same as the effect in the older cohort from previous research, which had been estimated at .15 SD. The idea that the findings here are at odds with the previous research does not have strong statistical support. As Andrew Gelman likes to remind us, “The difference between ‘significant’ and ‘not significant’ is not itself statistically significant”

But what really got to me was this quote from the report:
Our finding that TFA and comparison teachers were equally effective is robust to multiple sensitivity analyses.
No, you did not find that they were equally effective. In social science, the null hypothesis is never exactly true. It is not plausible that TFA teachers and other teachers are exactly equally effective. If you fail to reject the null, it means your sample size was not large enough. And by your own assessment, you weren't even able to statistically rule out "moderate" differences in effectiveness, let alone prove that the were exactly equally effective. 

What this quote really means is "we kept running regressions, but the sample size never got any larger."

Meanwhile, Jason Richwine at National Review attempts to draw a lesson about teacher training
The fact that TFA requires only a five-week crash course in pedagogy — rather than traditional teacher certification — is another reason to question the value of an education degree.
While I share his skepticism about the value of an education degree, this research can't really say anything about it.The TFA teachers are a highly selected group, with much more elite educational credentials. We can't really conclude anything from this research about the impact of education degrees on typical teachers.

Overall, if we get informally Bayesian, we should probably conclude that the TFA teachers are likely at least a bit more effective than typical experienced teachers in the schools studied. This is consistent with the previous research, as well as the consistent pattern of positive albeit statistically insignificant impacts found in the report.

But to know the overall impact of TFA on test student achievement, you should compare the TFA teachers to the hypothetical teachers that would have been hired in their absence. It seems very likely that this comparison would be even more favorable towards TFA teachers. Comparing TFA teachers to the average, experienced teacher is the kind of mistake a sabermetrician studying sports would never make.