Saturday, September 10, 2011

Valuation models and value drivers of stock returns: The Malaysian context

This post is a summary of the research I’ve undertaken for my thesis with regards to price multiples valuation models such as price-to-earnings (PER), price-to-book (PBV), price-to-sales, price-to-cash flow, EV/EBITDA, EV/Sales etc coupled with value drivers of share prices in the Malaysian context. The aim of this study is to give investors a better understanding of the appropriate valuation models and value drivers of stock returns in making investment decisions coupled with providing a faster way of analyzing the whole stock universe (though still nothing beats an in-depth analysis of individual stocks). This could be quite lengthy, so pardon me. If you don’t feel bored, read on :p Perhaps you could get a tip or two if you’re interested to do similar studies on other markets probably, and maybe share with me as well :)

Samples used are 373 firms listed in KLSE that cover most of the constituents of FBM EMAS spanning from year 2000 to 2010.

Identifying comparable firms:
Firstly before analyzing the appropriate price multiples to use, identification of comparable firms is needed. There are basically three industry classification systems available from Bloomberg terminal (The lifeline of most investment or finance professionals) for Malaysian firm such as Global Industry Classification Standard (GICS), Industrial Classification Benchmark (ICB) and Bloomberg Industry Classification System (BICS), though there are plenty of other classification systems such as SIC, Dow Jones, Fama and French Classification etc which are not available in Bloomberg though. To determine the most appropriate classification system, the system which has the highest explanatory power of industry’s averages of variables over the individual firms’ variables coupled with the lowest intra-industry variances is considered the most appropriate. The variables tested include PER, PBV, PS, ROE, operating margin, sales growth and stock returns, representing the valuations, profitability and growth of the firms. It was found that GICS clearly had the best results by having the highest explanatory powers and lowest intra-industry variances. The results were consistent with prior researches on EU and US markets as well where GICS clearly outperformed other industry classification systems. Therefore, next time when you want to extract comparable firms in Bloomberg terminal, GICS would likely give you the closest comparable firms for your analysis.

Appropriate valuation models:
Valuation models include market value multiples and enterprise value multiples based on 1-year forward and 1-year trailing net income, profit before tax, operating profit, EBITDA, sales, book value and cash flow. 1-year forward values are based on ex-post data, assuming perfect foresight by research analysts. Two empirical tests were done to determine the appropriate valuation models. Firstly, OLS cross-section regression and valuation errors were done to determine how best the fair values computed by different valuation models explain and fit onto the stock prices, a method commonly used by prior researches in the past. However, this could be of little contribution to investors as they would not be able to take advantage of the arbitrage opportunities if the fair values fit perfectly the stock prices. A more practical empirical model would be the convergence test where the convergence rates of the market values towards the fair values are measured. The winner among all the valuation models for the OLS regression as well as convergence test was price-to-earnings before tax (P/EBT), followed by PER and PBV. The worst valuation models appeared to be price-to-sales (P/S) and price-to-cashflow (P/CF). For convergence test, price multiples based on forward values outperformed trailing values, implying that investors would have greater arbitrage opportunities using forward values. Other observations included: (1) Market value multiples outperformed enterprise value multiples; (2) As we move the value drivers from bottomline to the topline of the income statement (i.e. net income to sales), results were poorer; (3) Convergence rates of market values towards the fair values improved as convergence duration increased, an indication of the inefficient market that Malaysia had and contrary to US findings where convergence results deteriorated as time went by.

Some of the industries in Malaysia and their appropriate value drivers for valuation models are shown below:

In summary, forward earnings and book values are appropriate models to use in equity valuation. EV/EBITDA and EV/Sales which were highly revered by some researchers in the past appeared to be poor valuation models to use. P/CF and P/S also might not contribute much to equity valuations in Malaysia. Hmmm…..The results are quite in line with the valuation models that are commonly used among research analysts. Perhaps the popularity of PER and PBV among research analysts could have caused the results to favor these two, as this might be a self-fulfilling prophesy as investors use these models to bring the market values towards the fair values computed by these models.

Firm-specific value drivers of stock returns:
This could be useful for deciding which firms to invest should the firms have similar upside based on their fair values. Several value drivers are tested, such as beta, book-to-market (inverse of PBV), earnings yield (inverse of PER), dividend yield, net gearing and market capitalization. Multivariate analysis using panel data regression tests is used to determine the explanatory powers and significance of these value drivers. All in all, book-to-market and market capitalization had the most significant impact on stock returns, followed by net gearing and beta. Earnings yield and dividend yield appeared insignificant in most of the industries. Book-to-market, market capitalization and dividend yield are negatively correlated to stock returns whereas net gearing, earnings yield and beta are positively correlated to stock returns. Surprisingly, beta appeared to be not so significant in affecting stock returns, rendering the application of CAPM in Malaysia rather pointless :P On the other hand, higher net gearing in fact favor stock prices, of course provided that the borrowings do not bring the firms close to default risks. This could be due to greater efficiency in the capital structure where higher borrowings could bring in tax savings and at the same time allow greater expansion of business operation. Most of the industries have more or less similar results as the overall market, except for agricultural products (Mainly oil palm companies) which had net gearing as the most significant driver, and construction firms of which dividend yield was significant in the negative direction to stock returns. Banks and industrial conglomerates (like Sime Darby) are not affected by net gearing at all.
In summary, lower market capitalization, lower book-to-market ratio and higher net gearing favor higher stock returns. 

GICS provides the best industry classification of firms

Most appropriate valuation models: Market value multiples based on forward values of earnings and book values performed the best. Enterprise value-based models coupled with sales, cash flow and EBITDA multiples performed poorly.

Lower book-to-market, lower market capitalization and higher net gearing significantly affect stock returns in the positive direction.


  1. Have you finish your thesis?

  2. Hi David,

    Very interesting, you must have spend a huge amount of time on this!

    Forward earnings, who determines that, there is some subjectivity there?

    I use mostly:
    - current PER, where I "normalize" the "E", take out one off items etc.
    - long term average ROE
    - and then all the rest: debt, cash flow, dividends, branding (or other moats), etc

    I would have thought that especially the average ROE is pretty important.

    I like your recommendations regarding unit trusts, good funds with long track records bought with minimal commission, for many people a much better solution then trying to outdo the market themselves.

    I have to study your stock picks, they sound pretty good.

    Keep up the good work.

  3. LCChong: Ya. Just finished. Will be having a few months holidays after this as well. Am trying to catch up with the market for now and finding gems in other parts of the world :P

    MA Wind: Thanks a lot :) What I've done is a general view on valuation models and not exactly a recommendation for in-depth analysis of each company. Personally, I would include all that you have mentioned in my analysis of individual stocks. And forward earnings I've been using the ex-post data which might differ from analysts' recommendations. There's a lot of subjectivity in forecasted earnings and it'll be difficult to do a test on it for hundreds of companies while many companies I covered in my analysis had no analysts recommendations for each year throughout the last decade.

    ROE is important to me as well. Surprisingly, for value drivers of stock returns, none of the prior researches had done a test on it, thus I've not included it in my thesis, lack of references. Perhaps I should include it in my future researches. Thanks.

  4. Very interesting research for stocks in Bursa which is hard to find. I wish I could read your full research article on this in order to see which are the important drivers for Bursa stock returns. Hope you can share your opinion on the followings:
    1)Why is that only 10 years records were used when the KLSE has records since more than 30 years ago? Is the analysis of only 10 years data robust enough especially there were 2 serious market down turn from 2000-2003 and 2007-2009? And the greatest bull market from 1990 to 1997 was missed?
    2)Does your stock return a total return figure which includes dividend (which should be) or just on stock price only?
    3)In your statement “Book-to-market, market capitalization and dividend yield are negatively correlated to stock returns whereas net gearing, earnings yield and beta are positively correlated to stock returns”
    a.I am surprised that Book-to-market and dividend yield are negatively correlated to stock returns as they are contradicting to conventional believe. I have been seeing stock prices rise all the time when dividends, especially a special dividend is declared. This also contradicting what Fama and French and Robert Shiller and others did for the US market and other researchers for other parts of the world.
    b.Your answer to question 2 above may shed some light as stock price will be lower ex of dividend.
    c.I am surprised also that higher gearing significantly affect stock returns in the positive direction. I wonder survival bias has been taken into account.

  5. Hello KC. Thanks a lot for your comments :) For Q1, 10 years due to limitation of time available for my thesis. Given more time, I would extend it to 30 years. I do admit that 10 years might be too short to draw solid conclusions. If possible, it'll be even better to split the time frame to bull markets and market downturns as I think the characteristics of value drivers would change as well.

    Q2: The stock prices are adjusted for dividends, stock splits etc.

    Q3: Book-to-market results indeed were contrary to the findings by Fama French and many others. However, I did find that some studies on Australian market had the same sign for book-to-market as Malaysia and their market cap is positively related to stock returns..hmmm..Some studies found that the negative correlation of book-to-market with stock returns happen in less developed markets such as Thailand and Malaysia, contrary to developed ones. One reason could be due to strong earnings usually associated with low book-to-market but high default risks associated with high B/M (Fama and French), thus investors tend to be more cautious and avoid low B/M stocks. But then again higher risk demands higher returns by investors, right? I've no answer to this. Hope you could shed some light.

    Net gearing was actually consistent with prior studies with earlier works like Bhandari 1988 - Higher risk and perhaps more efficient capital structure. Dividend yield's results were inconsistent in the past. Some positive and some negative. The results I've got for dividend yield were very insignificant overall. Thus it could be said that dividend yield overall had very little influence on stock returns as compared to the other significant value drivers. The sign of dividend yield could be misleading as well owing to its insignificance. Survivorship bias is not taken into account since the method I used is panel data analysis thus companies that do not have the data for all the years are ignored. Further research on cross sectional analysis which could take this into account could make help confirm the results.


  6. Hi David,
    Let’s see if we can utilize your research outcome to profit from Bursa. I hope you do not mind some of my critical comments. I certainly would be happy if you would critically comment on my points of view. That is how we can learn from each other.
    First your statements of “Most appropriate valuation models: Market value multiples based on forward values of earnings and book values performed the best”, and “1-year forward values are based on ex-post data, assuming perfect foresight by research analysts”. The question is where to get this forward earnings when I am deciding whether want to buy that stock or not now? Research analysts have perfect foresight? I thought I have read that research has shown that analyst earnings forecasts were out widely and at more than 80% of the time?
    Secondly your statement “Lower book-to-market, lower market capitalization and higher net gearing significantly affect stock returns in the positive direction.” Looking at the evidence (1927-2001 in US), lower PBV stocks (not the inverse of your book-to-market) have generally outperformed high PBV stocks on average, though there have been extended period when low PBV stocks underperformed as well. This is also confirmed with all other international markets such as Europe, Japan and even emerging market like Korea, have their low PBV stocks outperforming the rest by 1.06% to 3.26% using the data from 1981 to 1992. Low PBV stocks perform best when the overall market is in the doldrums, reflecting their status as defensive stocks. I believe the Malaysian stocks would behave similarly if you use a long enough data and also data at a different period, because the explanation of it would be more plausible.
    Lower market capitalization stocks perform better? Try using the 10-year monthly KLSE Second Board index (A good proxy for small cap) from 31/1/2007 to 31/12/2006 when the index dropped from 612 points to less than 100 point; and then compared with the returns of those higher capitalized stocks during the same period. High net gearing stocks better return? I wonder how many of those highly geared and small capitalized stocks vanished into thin air during the 1997 Asian financial crisis and the sacking of Anwar Ibrahim saga.
    In short, I think a single metric cannot be used to explain any anomaly in the market. For example, a small capitalized stocks may earn excess returns but they must be accompanied with say ROE>10%, Debt-to-equity ratio less than 1, expected growth rate of >10% etc.

  7. "Assuming perfect foresight" - It's an assumption. Maybe I should just use the term 'ex-post' 1-year forward. Lower market cap outperforming higher market cap I think it's quite consistent even with developed markets. Agree with the rest of your statements :))