Authors: Philip Vanstraceele and Luc Allaeys
A version of this paper can be found here
Introduction
This is our fourth post in our series on “net nets” having previously analyzed “Benjamin Graham’s Net Current Asset Values: A Performance Update” by Henry R. Oppenheimer, “Graham’s Net-Nets: Outdated or Outstanding?” by James Montier and “Testing Benjamin Graham's Net Current Asset Value Strategy in London” by Ying Xiao and Glen C. Arnold.
The focus of this post is to analyze the research paper, “Studying Different Systematic Value Investing Strategies on the Eurozone Market” conducted by investment practitioners Philip Vanstraceele and Luc Allaeys.
The paper examined the performance of a variety of value investing screens during the 10 year period from 1999 to 2009 for stocks listed in the “Eurozone”. Of interest to us, the paper examined the performance of stocks identified by a “Net Current Asset Value” (NCAV) screen.
Our objective was to analyze the study itself; determine its reliability, draw our own conclusions and glean, if any, actionable advice for the practitioner of the Net Current Asset Value (i.e. “net net”) method of investing. While the authors examined four value investing screens, we focus our analysis on the Net Current Asset Value method.
Key Methodology
Valuation metric:
“NCA= Current Assets (cash, inventories and accounts receivable) – Total Liabilities
This strategy calls for buying stocks trading at 2/3 or less of their NCAV.
In our methodology the NCAV-ratio should be greater than 1,33, and we define the ratio as Net Current Assets Value / Market Value.
This ratio is used to find companies that are trading below their net current assets value.”
For clarity, it should be noted that a ratio of 1.33 of Net Current Asset Value/Market Value equals 75% of NCAV (not 2/3 i.e. 67% of NCAV more typically associated with the strategy as mentioned by the authors). In addition, there is no mention of the treatment of preferred shares, therefore we assume they were not taken into consideration.
Weighting: “equally weighted basis”
Holding/rebalancing period: “buy and hold strategy for 1 year… and rebalanced the portfolios once a year at exactly the same time” i.e. on 13 June.
While rebalancing on the same date each year is consistent it nonetheless exposes the results to “timing luck”. For example, if portfolios were rebalanced in March it is highly likely that the returns would appear materially lower given that most equity markets reached their trough in March 2009. In contrast, if portfolios were rebalanced in October, it is highly likely that the returns would appear materially higher given that most equity markets saw appreciable recovery between March 2009 and October 2009.
The point being, consistency, while desirable, cannot alone be equated with reliability.
Market capitalization and the number of holdings:
The authors examined 20 stock and 50 stock portfolios and did so at various minimum market capitalization cut offs.
The portfolios examined are identified below:
Reliability
While there are numerous biases/errors that can be made when conducting studies, below we have analyzed those we deem most likely to impact a study of this nature:
1. Survivorship bias
“Most studies don’t include companies that went bankrupt or that were taken over by others. In our test these company’s [sic] were not excluded.”
Based on the above the data source appears to be free from survivorship bias.
2. Look Ahead bias
“When you are using data for your stock ranking that was not available at the moment of portfolio formation, your results will suffer from look-ahead bias. This biases results upwards.
We worked with accounting data from the prior fiscal year and waited 6 months to form our portfolios and actual trading.”
Given the “six-month lag” the study appears to be free from look ahead bias.
3. Time period bias
The study spans 10 years and we classify this as an “inadequate/unreliable” period.[i]
However, given the period incorporated the Great Financial Crisis that saw many markets drawdown 50% we do accept the premise put forth by the authors:
“All things considered, the last decade has been quite turbulent for the stock markets all over the globe. We had one bull market and two major crashes (a “luxury” for back-testers).”
Nonetheless, an investor would be well advised to refrain from making capital allocation decisions based solely on a 10 year back test.
4. Data source and treatment
Data source: “Thomson DataStream”
Excluded industries: “We excluded Financial- and Insurance companies from our screeners!!”
Universe:
“In the Euro Monetary zone there are +/- 4.080 companies with a primary quote on the stock market. If we exclude the financials and insurers are excluded +/- 3.400 companies remain.”
We assume that the abovementioned represents the stock universe available for examination “from the end of 1998” and that new listings were incorporated after the commencement of the test period[ii].
Dividends:
“We back tested the different strategies and calculated the price-index “PI” (excluding dividends)”.
Excluding dividends when measuring returns would have certainly understated the return achieved. Furthermore, for firms that were able to sustain the quantum of their dividends during the Global Financial Crisis, they would have, on the balance of probability offered a particularly high dividend yield. Thus, not accounting the dividend component would have, in our view materially impacted the reliability of the back test. Confusingly, in a subsequent set of tests the authors state: “Here we included the dividends”. So, when examining the results, we need to be mindful as to which set of tests we are analyzing given some accounted for dividends and some did not.
Eurozone:
The “Eurozone” consists of European Union (EU) member states which have adopted the euro (€) as their common currency and sole legal tender. The authors did not expressly state which countries/markets formed part of the Eurozone during the period studied. Complications arise given some member states adopted the Euro during the study period. Were they included following their adoption of the Euro or were only markets that had adopted the Euro at the commencement of the study included? While not mentioned in the study our correspondence with the author confirmed that countries that adopted during the period were included. That means Greece, in consequence of their adoption of the Euro on 1 January 2001 would have formed part of the study. While seemingly trivial had they been excluded it may have had a material impact on the results given the performance of their stock market during the period examined. Furthermore, it is during major drawdowns that net nets are more likely to be available and therefore their potential exclusion could have had a material impact on results.
For completeness we have included the list of countries that had adopted the Euro at the commencement of, and during the study period (13 June 1999 to 13 June 2009) in the endnotes[iii].
5. Citation Error
It is stated that:
“One research study, covering the years 1970 through 1983 showed that portfolios picked at the beginning of each year, and held for one year, returned 29,4%, on average, over the 13 year period, compared to 11,5% for the S&P 500 Index.”
The authors are almost certainly referring to the study “Benjamin Graham’s Net Current Asset Values: A Performance Update” by Henry R. Oppenheimer. While minor, the comparative benchmark to which they are likely referring to was actually the “NYSE-AMEX Index” as opposed to the stipulated “S&P 500 Index”.
6. Journal rating/credibility[iv]
This study was not, to our knowledge, published in a top tier academic journal and therefore cannot be granted the “additional credibility” that may come with such publication.
Initial Reliability Assessment: Despite a very concerted effort to ensure the study did not suffer from major biases the relatively short test period (10 years) leads us to conclude that in isolation the results cannot be fully relied on, however, they could provide the basis for further research.
Portfolio Return Analysis
Let the games begin!
So as identified above the authors test net nets across different market capitalizations and number of holdings. For example, they examined net net portfolios with minimum market capitalization of EUR 25m and 20 companies (“MV25 C20”) and net net portfolios with a minimum market capitalization of EUR 5bn and 20 companies (“MV5.000 C20”).
Wait, what?!
Is that to say in each and every year the authors were able to identify at least 20 net nets (per their specified criteria – “NCAV-ratio should be greater than 1,33”) across the entire spectrum of market capitalizations tested? Well, we could believe that is possible for the portfolios with a minimum market capitalization of EUR 25m, but to find at least 20 net nets with a market capitalization of EUR 5bn for ten consecutive years?! That seems utterly implausible! And so, it is…
In fact, what was actually tested were those firms that traded at the highest relative “ratio of Net Current Assets Value / Market Value” (MV) not what was specified in the methodology of the study (refer section “Key Methodology”)! Significantly, there was no way to have definitively reconciled this incongruity other than by asking the authors.
Now, having identified that this study does not examine what it stated (i.e. firms trading at a NCAV ratio of greater than 1.33, or said in more familiar terms, 75% of NCAV), but rather firms trading at the highest relative ratio of NCAV / MV (i.e. the “cheapest) we proceed with a review of the returns achieved.
Significantly, and thankfully, the authors examined the returns by calculating the Compound Annual Growth Rate (CAGR) (as opposed to the arithmetic average return). Out of interest, the arithmetic average return for the DJ Euro Stoxx was -0.24% which, prima facie doesn’t sound calamitous given the literal calamity that transpired in financial markets during the period examined. However, the truth of the matter is (gross of fees) an index tracking investor compounded at -3.13% (including dividends)[v] and therefore lost a cumulative 27.22% for the 10 years ended 13 June 2009!
We have restated the returns as reported in the study with regard to portfolios constructed using the NCAV/MV metric below:
For simplicity and brevity, we will review the returns of the relatively cheapest stocks as measured by NCAV/MV for three groups:
1. MV25 C20 (minimum market capitalization of EUR 25m and 20 companies):
The NCAV MV25 C20 generated a CAGR of 13.44% vs the DJ Euro Stoxx -3.13%, an outperformance of 16.57%. Certainly, generating a double digit return while the market return was negative would have been heartening!
2. MV250 C20 (minimum market capitalization of EUR 250m and 20 companies):
The NCAV MV250 C20 generated a CAGR of 11.06% vs the DJ Euro Stoxx -3.13% an outperformance of 14.19%. Significantly at a minimum market capitalization of EUR 250m some institutional investors would be able to access the outperformance.
3. MV5.000 C20 (minimum market capitalization of EUR 5bn and 20 companies):
The NCAV MV5000 C20 generated a CAGR of -1.55% vs the DJ Euro Stoxx -3.13% an outperformance of 1.58%. While the outperformance of 1.58% for the largest capitalization stocks examined appears trivial, it is important to note that the cumulative consequence was a loss of 11.45% (100 - 85.55) as opposed to 27.22% (100 - 72.78); a little testament to the power of compounding!
Observing the returns over time one can’t help but notice that returns between the largest group examined (“MV 5000 C20) along with the market (“DJ Euro Stoxx) appear to have varied materially from those attained by the portfolios of smaller firms.
Below we chart the annual return of four (of the 13) portfolios tested along with the annual market returns[vi].
While oftentimes directionally consistent, the variation in the magnitude of returns over the time period studied is noteworthy.
Calculating the correlation coefficient between the returns of the portfolios charted above demonstrates that the returns of smaller stocks took an entirely different path to those of larger stocks and the overall market.
Lastly, we chart below the compound annual growth rate of the 20 stock portfolios (i.e. “C20”) along with that of the market:
From the chart above it would appear that as the minimum market capitalization was increased returns fell in a consistent, though not perfect sequence. A regression analysis was not conducted for various “factors”. However, a reasonable hypothesis might attribute the pattern as consistent with the “size factor”, along with the probability that as the investment universe shrunk (by increasing the market capitalization requirement) it contained firms with a lower relative NCAV/MV ratio (i.e. more expensive).
Now, let’s not forget, these tests excluded dividends when it came to the quantification of returns to the NCAV portfolios, however the DJ Euro Stoxx Index included dividends[vii]. Including dividends, it would be reasonable to assume that it would have resulted in a material increase in returns to the NCAV portfolios not accounted for in the results displayed.
So, there it is, the cheapest firms as identified as those with the highest relative NCAV/MV ratio produced market beating return from 1999 to 2009 in the Eurozone across a full spectrum of market capitalizations - case closed.
Well, not so fast…
Decile Return Analysis
The authors also split their stock universe up into deciles (with various minimum market capitalization requirements) and presented the returns (this time including dividends).
The authors make a particularly thought provoking comment in their conclusion:
“Using the NCAV and splitting it up in deciles sometimes gives you strange results but you have to consider that using NCAV doesn’t always reflect “how cheap” a company is.”
We analyse this comment in further detail.
The minimum market capitalizations for the decile tests are EUR 10m, EUR 100m and EUR 500m respectively.
To get a sense of the size of the universe, and therefore the number of firms in each decile, we provide an extract from the study identifying the size (and market capitalization) of the investment universe examined:
While the market cap classifications above don’t neatly match those used in the decile analyses, we can deduce that the minimum numbers of firms contained in each decile for the 10m, 100m and 500m minimum is not less than 190, 71 and 38 firms respectively[viii].
Given that we have a MV100 C20 portfolio and a MV500 C20 portfolio to examine we will focus on the deciles returns with the same respective minimums i.e. 100m and 500m - this ensures the market capitalizations of stocks match when conducting our analysis.
Below we restate the returns of decile 1 (D1)[ix]and decile 2 (D2) (minimum market capitalization 100m) alongside those of the MV100 C20 portfolio for comparative purposes:
Recall MV100 C20 consists of the 20 firms that possessed the highest ratio of NCAV/MV i.e. the cheapest stocks. Therefore those 20 firms would be subsumed within D1. While the cheapest 20 firms generated a CAGR of 10.48% the cheapest decile, which contained those firms, generated a CAGR of -2.18%. Yet, D2 generated a CAGR of 11.93%.
The above implies a lack of robustness to the hypothesis that cheapness, as measured by the NCAV/MV is the driver of returns, for if it was, we would expect a consistent diminution of returns as the NCAV/MV decreased (i.e. more expensive), as opposed to the inconsistent sequence identified above.
Furthermore, it should be noted that the 10.48% CAGR achieved by the MV100 C20 portfolio excluded dividends, whereas the -2.18 CAGR of D1 included dividends, further confounding the results.
A chart of annual returns for D1, D2 and MV100 C20 is provided below:
Continuing in the same vein below we restate the returns of D1 and D2 (minimum market capitalization 500m) alongside those of the MV500 C20 portfolio for comparative purposes:
The results with a minimum market capitalization of 500m are very similar to those found for a minimum market capitalization of 100m. Specifically, MV500 C20 consists of the 20 firms that possessed the highest ratio of NCAV/MV i.e. the cheapest stocks. Therefore those 20 firms would be subsumed within D1. While the cheapest 20 firms generated a CAGR of 7.23% the cheapest decile which contained those stocks generated a CAGR of -6.89% (a 51% cumulative loss!). Yet, D2 generated a CAGR of 7.23%.
Again, the above implies a lack of robustness to the hypothesis that cheapness, as measured by the NCAV/MV is the driver of returns, for if it was, we would expect a consistent diminution of returns as the NCAV/MV decreased (i.e. more expensive), as opposed to the inconsistent sequence identified above.
A chart of annual returns for D1, D2 and MV500 C20 is provided below:
Conclusion
The authors stated that they were examining the returns to stocks that traded at a minimum NCAV/MV ratio of 1.33 (i.e. 75% of NCAV), however we subsequently confirmed that what was actually tested were stocks that traded at the cheapest relative valuation as measured by NCAV/MV with no constraints to the absolute value of the ratio itself.
Initially, the CAGR achieved by portfolios across a wide spectrum of market capitalizations seemed to definitively indicate the firms with the highest NCAV/MV (i.e. the cheapest) reliably and consistently outperformed the market as measured by the DJ Euro Stoxx. However, the robustness of the results was brought into question when we analysed the returns to the decile portfolios which revealed an inconsistency which could not be explained.
The lack of robustness in returns notwithstanding, and despite a very concerted effort to ensure the study did not suffer from major biases the relatively short test period (10 years) would have always led us to conclude that, in isolation, the results could not be fully relied on. Nonetheless, while uncertainty as to the efficacy of the NCAV/MV as a standalone metric remains, the results also imply that it is also worthy of further examination.
Sometimes, to see further one must dig deeper.
Notes:
[i] For reference:·
< 10 years; inadequate/unreliable
11 to 20 years; somewhat reliable·
> 20 years; more reliable·
> 40 years; most reliable
[ii] Confirmed by author.
[iii] Eurozone countries (https://en.wikipedia.org/wiki/Euro): Monaco, San Marino and Vatican City do not have a stock exchange.
[v] The DJ Euro Stoxx index included dividends (confirmed by author).
[vi] We opted not to chart all 13 portfolios to ensure the chart was readable.
[vii] While the NCAV portfolios excluded dividends the DJ Euro Stoxx index included dividends (confirmed by author).
[viii] 10m: 77+308+334+1189 = 1908, 1,908/10 = 190.8. 100m: 77+308+334 = 719; 719/10 = 71.9. 500M: 77+308 = 385/10 = 38.5.
[ix] D1 contains the firms highest NCAV/MV ratio i.e. the cheapest.