I have been doing a lot of backtesting recently, and I have come to an interesting conclusion.

I have been looking at ordering all the stocks in a "reasonably liquid" universe, sorting them into buckets, and observing the performance of each one.

The common metric used, "price/earnings" ratio (PE ratio), does ok in discriminating between groups of stocks that have high returns and low returns. Buckets with low p/e ratios do better.

For the U.S. market between 01/03/01 and 02/02/08, sorting into 10 declies by P/E, the universe of liquid stocks (market cap > $50m, close price > 1, daily turnover > $200k) the bottom P/E decile returns 19% per annum (rebalanced monthly), well ahead of the 2.8% return of the S&P 500. The next decile returns 17%, the next one returns 14%, and so on.

The evidence clearly shows monotonically increasing returns by choosing stocks with lower P/E ratios, at least at the size of these buckets. Each bucket contains about 500 stocks - so one can be fairly statistically confident of the result.

However, the professional world, increasingly "P/E" is not used, instead more advanced metrics are used, such as EV/EBITDA. Let me define this:

EV=Enterprise Value
EBITDA=Earnings before interest, taxes, depreciation and amortisation

(more information and definitions can be found on wikipedia)

Enterprise value, for the purposes of my test, is defined as (marketcap + totaldebt - totalcash). One of the reasons it is useful to consider the "enterprise value" as opposed to simple "market capitalisation" is that if company A wished to purchase company B outright, it would not only have to buy up all the shares at the current price (market cap) but it would also have to assume the companies debt. If company B had any cash on hand, this would effectively be an "instant rebate", so it gets subtracted from the cost of acquisition.

The motive for using EBITDA are many fold. Some people critisize EBITDA for painting a rosy picture, but the reason adding back the various components can be justified:

-In the above example, EV defines how much would have to be paid to remove all the debt. Since this is without debt, we can safely "add back" interest to the earnings. This provides a "what-if" scenario.

-The reason for excluding taxes are manyfold- but primarily earnings may be temporarily boosted if the company is carrying forward a tax loss from previous years. Since this is only a temporary boost to earnings, in the long run most corperations will pay the same tax rate, so for the purposes of comparison you want to back it out.

-The justification for adding back depreciation is slightly more shaky. However, in the calculation of "enterprise value", no allowance is made for the residual value of the assets of company B. In effect, they have been written off to zero, and further depreciation charges are uncessary, as they are simply a "book keeping" exercise. In reality, this is not necessarily perfect. Some cash may need to be set aside to replace worn assets, but this depends very much on the type of company. Some assets are depreciated but do not literally "wear out" (think radio transmitter). A guage of how much the company needs to invest can be made by looking at the capital expenditure, and free cash flows. The ratio of "free cash flow" to total assets provides a good indication of how asset intensive the company is.

-The justification for adding back amortisation is along similar lines. Amortisation is an annual charge against goodwill. Since the calculation of "Enterprise value" contains no "goodwill" as an effective "rebate", it can be ignored in the on going calculation, it is purely a "bookeeping" exercise with no economic effect.

***

The test using EV/EBITDA, the bottom decile returned 25%, the next one 19%, the next one 14% and so on. The performance is signficantly better than using the P/E ratio.

I would recommend that people focus on EV/EBITDA as a better measure of value than P/E ratios. The EV/EBITDA considers the balance sheet of the company, and excludes a lot of "fuzzy" numbers that are set by arbitrary conventions, rather than true economic reality. One might ask why it should be better measure of value, since even if a company has cash on its balance sheet, investors cannot access and "eat" the cash, and similarly with debt.

However, a company that is debt free makes a signficantly more attractive takeover candidate than one that is loaded with debt.

In addition, a company that is debt free, and/or has a large amount of cash, is more likely to return that cash to shareholders through dividends and/or buybacks. A company saddled with high levels of debt will probably not consider share repurchases until the share price gets significantly low.

Finally, given the issues surrounding capital expenditure, in a two-dimensional ranking system of both EV/EBITDA and Total Assets/Free Cash Flow, the bottom decile (firms with low ratios of EV/EBITDA and Assets/FCF) posted an average annual return of 27%, pretty impressive in a period where the annualised return on the S&P 500 was just 2.8%

very interesting stats olikea thanks for sharing,a queery if I may,why does no-one consider the profitability of the company important when screening for value?Surely profitability and ROE and ROCE and margins are primary considerations?
If you have the time and fascilities to test perhaps incorporate these metrics into your testing the results would be very interesting I'm sure

The justification for adding back depreciation is slightly more shaky. However, in the calculation of "enterprise value", no allowance is made for the residual value of the assets of company B. In effect, they have been written off to zero, and further depreciation charges are uncessary, as they are simply a "book keeping" exercise

This is plain wrong. EV is a measure of what the markets value a company's equity at a point in time, and debt which is a matter of fact. It has nothing to do with book keeping, and certainly not with disregarding assets. Accordingly, it makes no sense in the other side in the EBITDA side of the equation to ignore depreciation. Depreciation is an accounting cost, but an investor would be crazy not to consider the cash cost of replacing worn out assets with fresh capital investment.

For example in BT's reasonably upbeat Q3 results yesterday, the fly in the ointment as far as I could see is the very chunky capital investment programme. BT possibly are a fair investment right now, but the size of the capex programme should cause pause for thought. Look back a few messages ago and I commented yesterday on Yell's Q3 results. That comment was basically a back of the evelope calculation using EV/EBIDTA, but at least with some consideration of the level of capex in that business (I presume intrinsically a concern with low capex demands.)

so a decent measure to use is EV/EBIT (earnings before interest and tax), or alternatively, if ongoing capex is likely to differ materially from the historic depreciation charge, EV/(EBITDA - capex)

Remi

The justification for adding back depreciation is slightly more shaky. However, in the calculation of "enterprise value", no allowance is made for the residual value of the assets of company B. In effect, they have been written off to zero, and further depreciation charges are uncessary, as they are simply a "book keeping" exercise

This is plain wrong. EV is a measure of what the markets value a company's equity at a point in time, and debt which is a matter of fact. It has nothing to do with book keeping, and certainly not with disregarding assets. Accordingly, it makes no sense in the other side in the EBITDA side of the equation to ignore depreciation. Depreciation is an accounting cost, but an investor would be crazy not to consider the cash cost of replacing worn out assets with fresh capital investment.

The purpose of EV/EBITDA is to give an idea of the "payback" period for an acquiring company.

It is quite correct to remove depreciation from the expenses, because assets that have already been purchased have been paid for. If company A takes over company B, then it does not consider the non-liquid assets on the balance sheet. If "Enterprise Value" consisted of "market cap - book value": then you would have to consider depreciation.

Example:

Company A is interested in purchasing company B for £10bn. Company B has £5bn of book value, but no cash and no debt. Total reported corperate earnings are £1bn, which is after £1bn per year of depreciation charges.

Under the "EV" calculation, the purchase price of company B is simply, £10bn. Non-liquid book value is ignored. Assuming no taxes etc. for simplicity, the EBIDTA is £2bn per annum, giving a EV/EBITDA ratio of 5.

Instead, imagine that when purchasing company B, a different metric is used. Company A figures that since it will immediately recieve £5bn of assets on purchase of company B, the "true" cost of purchasing company A is £5bn. However, it would be quite invalid to ignore depreciation, as those assets need to be written off every year. Hence this ratio, which is "(marketcap - book value)/earnings" would also, be 5.

Which ratio is better to use? I would argue that EV/EBITDA is always better. The problem about book value and depreciation is that there is inconsistancy on how they are applied. Some companies in the same industry (airlines for example) will have different deprecation rates for the same assets (aeroplanes). By removing the arbitrary accounting adjustments it makes cross-company comparisons much easier.

HOWEVER.

I did already mention in my original post that depreciation cannot be completely ignored on a "going concern" basis; some of those assets may eventually "wear out" and need to be replaced, which is why I recommend looking at free-cash-flow. It is my assertion that capital expenditure is a much better measure than depreciation of the true economic cost of replacing assets. Some assets that are "depreciated" literally never wear out. A radio broadcasting company can build a transmitter, and afterwards, it need never be replaced. Clearly, an airline on the other hand, will eventually have to replace its aeroplanes. Depreciation charges are arbitrary and do not tell you which is the case.

In any event, there is plenty of evidence that shareholders gain better returns in companies that are not asset intensive. Typically ROE, ROA are reffered to, however, the best ratio that I have tested is "Free-cash-flow/total-assets". Free cash flow is cash flow *after* capital expenditure and is the best measure of what shareholders can "eat". This is why I suggested a two dimensional ranking system that considers both - this is similiar to Greenblatts "little book that beats the market". A problem arises because capital expenditure can be somewhat "lumpy". Companies with good management will try and spread their capex over time, and in any event, you can look at several years capex to build a picture. This is better than simply using "depreciation", which is the accountant's attempt to "smooth" the picture for the benefit of shareholders. In my view, far from "smoothing", it gives a "smudged" picture instead!

so a decent measure to use is EV/EBIT (earnings before interest and tax), or alternatively, if ongoing capex is likely to differ materially from the historic depreciation charge, EV/(EBITDA - capex)

Its a good idea, I just conducted a backtest of EV/(EBITDA-capex), the bottom decile returns 26% per annum, compared to 25% per annum for EV/EBITDA.

The purpose of EV/EBITDA is to give an idea of the "payback" period for an acquiring company.

Exactly, and as the denizens of these boards are not "acquiring companies" why would EBITDA be a useful measure?

Exactly, and as the denizens of these boards are not "acquiring companies" why would EBITDA be a useful measure?

Because share prices are ultimately driven by other market participants, not you. Although companies do not participate in the equity market on a day-to-day basis, when they do, it often has a big effect (think Microsoft-Yahoo).

Also, as I explained, a company with a low EV/EBITDA ratio is in a better position to repurchase its own shares.

Olikea: I just conducted a backtest of EV/(EBITDA-capex), the bottom decile returns 26% per annum...

Hi Olikea,

Thanks for posting your interesting analysis, I am learning!

Quick question: What would be the best (quickest, most accurate) way for me to calulate the above value for (say) the biggest 200 companies in the UK? I.e. what (cheap) data source would you recommend? Ideally I'd want a rolling 3 yr average - latest full year results, previous year, next year's forecast. Is this realistic?

Presumably, picking a sector-diversified portfolio of 20 shares from the bottom two deciles of such a list would produce a 20 share HVP (High Value Portfolio)?!

Loaf (ignorant chartist)

Quick question: What would be the best (quickest, most accurate) way for me to calulate the above value for (say) the biggest 200 companies in the UK? I.e. what (cheap) data source would you recommend? Ideally I'd want a rolling 3 yr average - latest full year results, previous year, next year's forecast. Is this realistic?

Presumably, picking a sector-diversified portfolio of 20 shares from the bottom two deciles of such a list would produce a 20 share HVP (High Value Portfolio)?!

I don't know much about data sources for the UK market. One of the main reason for the focus on the U.S. is that -data is readily available, -it is a large market. My "liquid stocks" universe consists of about 5,000 stocks, (some of which are ADRs of UK plcs). Each decile is 500 stocks, a large number.

There is also an issue of diversification, my tests and experience suggests you need a bare minimum of 50 stocks to be adequately diversified, and if you are risk averse perhaps 200 stocks, though this does apply to the U.S. stocks which has a very large number of small and microcaps, which can be very volatile. If you restrict yourself to the large caps, like the S&P 500, then 20 may be adequate. However, a lot of the "excess" returns are restricted if you choose from a larger cap universe.

A very interesting thread, but without wishing to sound clever I've heard a simpler version of the theory that to my mind reaches the same conclusions through a different route - named...

...PYAD.

i.e. - a share with a good yield, low PTBV, low debt (or more preferably lots of net cash) and a lower than average P/E ratio is statistically more likely to outperform than one that lacks some or all of these features.

I have also seen, in a market in which most shares are getting knocked down, companies with high gearing getting bashed even more than the average, even when revenues and EPS growth look very strong.

B2V

"A very interesting thread, but without wishing to sound clever I've heard a simpler version of the theory that to my mind reaches the same conclusions through a different route - named...

...PYAD."

But olikea is suggesting an alternative measure to the P in PYAD, hence if you still want PYAD shares you could try to change the filter where P=EV/EBITDA rather than P/E, maybe it should be called EYAD ;-)

--nix

nix, I see your point, and I am guilty of a late night oversimplification (on a Friday night, too - I must be getting old...) but the sections that grabbed me in the original post were:

"However, a company that is debt free makes a signficantly more attractive takeover candidate than one that is loaded with debt.

In addition, a company that is debt free, and/or has a large amount of cash, is more likely to return that cash to shareholders through dividends and/or buybacks. A company saddled with high levels of debt will probably not consider share repurchases until the share price gets significantly low."

This to my mind is exactly part of the background thinking behind the PYAD approach, although olikea - who deserves great credit for a good point supported with hard research - reaches it through using EV/EBITDA as what appears to be a more sophisticated measure than bald P/E.

I don't agree exactly with your point about "EYAD" though - what olikea's findings remind us is that P/E alone can be very deceptive as a measure of valuation, and that we must also consider debt. Using Enterprise Value rather than simply Price factors in debts in a similar - though obviously not identical - way to considering P, Y, A, and D separately. Hopefully, use of either approach as a measure of of valuation will be beneficial for value investors.

B2V

I hope you have allowed for survivorship bias in your database - whilst I don't disagree with much, even most, of what you say - some stocks that would've been low pe, or particularly high yield... at the start of the period wouldn't actually have survived to the end. Only a few would enormously skew the results.

But I'm not trying to discourage you!

Harry

I hope you have allowed for survivorship bias in your database

It is a good point, and I think it is worth mentioning for the benefit of everyone the issues of "survivorship bias", and a more sutble issue of "look-ahead bias".

Survivor bias is the issue of defunct companies no longer showing up on historical databases. Therefore, you might erroneously discover that a good strategy is to buy companies on the verge of bankruptcy - only the ones in the database have survived, and presumably have outsized returns. But there is no allowance for those that went bust.

"Look-ahead bias" occurs when information from the future "leaks" into the past, this can occur through fairly subtle means, e.g. if you assumed you traded on the day's high or low, this is a form of look-ahead because the high/low of the day is unknown until the end of the day. Therefore any trading system you must be vigilant that you are not using low/hi of the day in an innapropriate way.

Otherways look-ahead occurs is the issue of companies "restating" their results. When companies earnings are restated, revised up or down, the historical figures are changed to reflect this. However, this information would not have been available at the time. This is the most serious form of look-ahead bias, and with a number of historical databases it can sometimes be difficult to be 100% sure they do not suffer from it. This is one of the reasons the data period tested does not go back further than 2001.

The solution is to make sure your data is "point in time". Thankfully, a number of agencies such as Reuters and Standard and Poor's have been capturing and storing data in real time. They claim to have data at a weekly resolution that contains all information known at the time. The interest in backtesting has only come about more recently, so the further into the past you go, the poorer the quality of the data. However, the more data you have, the more statistically confident you can be of the results. There is something of a trade-off. The data from 2001 is supposedly free of look-ahead and survivor bias, though, aha, they disclaim responsibility for any errors! Nothing in life or the stock market is risk free!

Some assets that are "depreciated" literally never wear out. A radio broadcasting company can build a transmitter, and afterwards, it need never be replaced.

Sorry, but that basically isn't true. Bits of it will have to be replaced, e.g. because they get corroded, or because they are the paintwork protecting the rest from corrosion. Parts of it will have to be replaced because of changes in radio transmission standards - e.g. if the allocation of radio spectrum changes. Parts of it will have to be replaced because of technical advances, if e.g. they can be upgraded to something that is sufficiently more power-efficient.

And even completely "solid-state" equipment without any "moving parts" can simply wear out. We've known it for years in the form of lamp bulbs blowing, but it also applies for instance to computer chips - see http://en.wikipedia.org/wiki/Electromigration for an example of one effect that causes such failures.

Of course, it may be that many such costs are sufficiently low-grade and continuous that they get regarded as running costs rather than capital expenditure. But don't believe in devices that "literally never wear out" without a lot of evidence!

Gengulphus

On the other hand some assets may actually appreciate rather than depreciate.

For example a good quality metal filing cabinet, which will last almost indefinitely ( more than 50 years ), where the cost of replacing it might be considerably in excess of it's depreciated book value.

And of course land and buildings, which tend to appreciate in value over time, even after allowing for maintainance costs.

olikea,

Very interesting thread. Please could you tell me where you get your US data. Do you have to buy it?

Thanks in advance.

David

Please could you tell me where you get your US data. Do you have to buy

www.portfolio123.com

There is a monthly fee but worth every penny, to get a similiar product from S&P costs around $30k annually!

For the U.S. market between 01/03/01 and 02/02/08, sorting into 10 declies by P/E, the universe of liquid stocks (market cap > $50m, close price > 1, daily turnover > $200k) the bottom P/E decile returns 19% per annum (rebalanced monthly), well ahead of the 2.8% return of the S&P 500. The next decile returns 17%, the next one returns 14%, and so on.

I'm not sure this is a sufficiently representative period to validate, or perhaps more precisely, to trump low PE as a value strategy. With respect, it measures the one period in time when debt was probably more accessible, and possible cheaper, than at any other time in modern history.

Regards

Ben

olikea or others,

Can you recommend a stock screener that does a good job with:
EV/EBITDA
Tangible B.V.

For free screeners, I have been happy with ADVFN since MSN Advanced Screener went off line.

Fran

Hi Fran

I am in no way connected with but would be very happy to recommend Sharelockholmes, which I have been extremely impressed with since I started using it some time ago.

http://www.sharelockholmes.com/

I think it includes the items you are after, but suggest you take a look around the site before subscribing.

Subs are less than £5 per month.

Regards,

Dave

On the other hand some assets may actually appreciate rather than depreciate.

For example a good quality metal filing cabinet, which will last almost indefinitely ( more than 50 years ), where the cost of replacing it might be considerably in excess of it's depreciated book value.

And of course land and buildings, which tend to appreciate in value over time, even after allowing for maintainance costs.

Strictly incorrect per IFRS.

The only non-depreciating items that are allowed per IFRS is land.

Buildings usually have a useful economic life of 50 years.

Whilst buildings might appreciate they would need to be revalued and then you hit a higher depn charge.

Thus it does seem reasonable to add back depreciation and amortisation but I'd argue taxation and Interest shouldn't be added back.

So going back to the original post, I'd argue EBDA is preferable. Each to their own of course.