This factor seems to work very well for every universe I test it in:
Eval(AccumDepQ = NA, AccumDepA/CapExTTM, AccumDepQ/CapExTTM)
The idea behind it is that it favors companies who are reducing their capital expenditures while retaining their fixed assets.
Does anyone have any experience with this kind of thing? Comments or thoughts?
Here’s a performance chart going back to 1999 using the Russell 1000 universe.
Yuval,
I created a single factor ranking system using your Eval function and got similar results.
However when I ran it using 100 and 200 buckets the highest rank buckets had much lower performance that lower ranked buckets.
Since I am not sure what the settings are for your test, your tests may not show the same thing with more buckets.
I always check ranking systems using 100 & 200 buckets so that I can see if the highest ranked buckets, the ones from which the Sim or Port will select the recommendations, still out perform the lower buckets.
From a conceptual standpoint is a company that is reducing capex good? There might be situations where that is a negative sign. Maybe putting companies in a growth or value context might provide additional insight.
x
Some food for thought on companies reducing CapEx…
According to Berkshire’s annual reports, one of Buffett’s criteria for identifying “good companies” (besides having the infamous ‘moat’) is that they have comparatively low CapEx. Obviously, businesses that consistently have to spend an outsized proportion of their cash flow on expenditures that are required to maintain the business will be less profitable (and thereby, less attractive as an investment) compared to businesses with relatively low CapEx.
For example, decades ago Buffett was a big fan of investing in newspapers. Established newspapers in a metropolitan area had an almost impervious moat because of years of established reputation and consumer habit. They also had very low CapEx once the initially large expenditure to build plant and equipment (printing presses) was paid for. Significantly increasing the number of newspapers sold each day (usually) did not require additional Capital Expenditures, just marginally more paper and ink (which would fall into the ‘cost of goods sold’ line item).
However, I don’t ever recall seeing Buffett elude to seeking businesses that were [i]reducing[/i] CapEx. I think that Sterling has it right, in that the management of companies that are substantially reducing their historical ratio of CapEx may know something the average investor doesn’t know - and it may be something fundamentally existential about the business. E.g., fast food restaurants that have discovered that sales are contracting and they have fallen out of favor with the public would recognize that continued expansion would reduce profits rather than increase it, so they will stop building new stores (i.e., reduced CapEx).
As Denny pointed out, when looking at narrower (100 and 200) buckets and the highest buckets have lower returns than buckets further down, it is the classic sign of a factor/formula telling you something is amiss. My guess is that the rebalance period and turnover of the RS/Sim will also play a role, since companies with significantly declining CapEx that are held for long periods may be prone to BK.
Chris
[quote]
This factor seems to work very well for every universe I test it in:
Eval(AccumDepQ = NA, AccumDepA/CapExTTM, AccumDepQ/CapExTTM)
The idea behind it is that it favors companies who are reducing their capital expenditures while retaining their fixed assets.
Does anyone have any experience with this kind of thing? Comments or thoughts?
Here’s a performance chart going back to 1999 using the Russell 1000 universe.
[/quote]I don’t think it’s doing what you think it is. Just because CapEx is low compared to depreciation, doesn’t mean that the company is decreasing CapEx compared to the past.
To actually get companies that are reducing Capex use something like: CapExTTM / LoopAvg(“Capex(CTR, TTM)”, 5).
Secondly, why are you using AccumDepQ instead of annual depreciation?
Depreciation vs. CapEx background:
Depreciation is an accounting fiction. CapEx is fact.
For example, when a company buys an asset for $30 million, they take a depreciation expense every year for thirty years or something like that. Sometimes the depreciation is real, such as railroads. Sometimes the depreciation is just accounting, such as TV broadcasting towers. So, in many cases depreciation is misleading.
On the other hand, CapEx is not accurate either. Some CapEx is for growth, but some is used to maintain the existing facilities. How do you differentiate between the two? Furthermore, CapEx is lumpy. Years can go by without putting too much maintenance into an asset, but then one year it needs a whole bunch of maintenance. So, CapEx measures the actual cash going out, but it is not normalized.
For capex intensive industries, it might make sense to buy a company soon after it completed its CapEx and RandD investments for a new product. It has seen some of its new product expenses, but not its resulting profits and cash flows.
I could see this with biotech companies that are now in their approval or go to market stage.
I can also see this for declining industries that have fallen on hard times and just throttled back on capex. Think oil and gas.
This strategy might work but I think you need to consider the universe first.
Thanks for setting me straight, Chaim. What I really should be looking at is gross plant to capex, not accumulated depreciation. That’s more what I had in mind, and it works better too.
What I was thinking of is companies like Cabot Corp, which has sunk so many millions into fixed assets that they don’t need to do much capital expenditures any more to generate return on those fixed assets.
A factor like that will be sensitive to the industry/sector you’re looking at. For companies in businesses that are not by their natures capital intense, it won’t mean much one way or the other.
Separately, the accounting standards poo-bahs make life difficult for us by not allowing tech companies to capitalize R&D. This means that for a large swatch of the market, major build-the-future expenditures never find their way into capex, gross plant, accumulated depreciation, etc. or into any equivalent account. Capex can still mean something for tech companies, assuming they actually do build physical assets instead of outsourcing. But even with a company that doesn’t the outsource, capex misses a lot of valuable information and just counts it as a dat to day overhead expense. (This, by the way, is what confused so many of the loudest and boldest AMZN bears a few years ago, when the company ramped up its spending and drove eps deep into the red).
This is why, to me, a ranking system’s usefulness can only be measured in terms of the buy-rules/screen with which it’s used. Factors like this are best when you pre-qualify to work only with low-tech manufacturing companies. I suggest testing any ranking systems like this against a universe constructed along those lines. And whatever results you see, consider them in the context of the world; outsourcing, insourcing, interest rates, etc. And if you ever wondered what might be a good occasion to pull some economic series’ into you modeling, this seems worthy of experimentation. Should companies be punished/rewarded for reducing/increasing capex in the face of improving/deteriorating conditions?
Well, if this doesn’t apply to tech companies, how do you explain this? I’m using one rule for ranking, eval(grossplantq = na, grossplanta/(capexttm+0.01), grossplantq/(capexttm+0.01)). (I always prefer quarterly numbers for balance sheet items.) The first chart is applying the rule to the Russell 3000, the second is applying it to only tech companies in the Russell 3000, the third is applying it only to industrials. All three are using max for the time period and a 4-week rebalance. It seems to work across the board. The tenth decile is not always the highest, but as a general rule, one could say that companies whose gross plant is much larger than their capital expenditures perform a lot better than those whose capital expenditures are high and gross plant is low, whatever industry they’re in.
For what it’s worth, Thomas Bulkowski did a study of ten-baggers from 1992 to 2007 and found that almost 60% had decreased their capital spending in the year before their growth began.
