The indexes used for comparing the performance of value versus growth stocks in the Thai market were the MSCI Thailand Value Index and the MSCI Thailand Growth Index. All the data were extracted from the database Bloomberg. The end of month value for both indexes were collected for the period from December 1999 to December 2016. The risk free rates for Thailand for all these years were extracted from Bloomberg and equate to the 10 year local treasury bond yield (longest time series available in the data base). For the previously mentioned period the value index generated returns of approximately 156% while the growth index generated a return of 120%. The MSCI indexes are frequently used as benchmarks by actual institutional investors in this market. It seemed reasonable then to use these indexes, rather than creating an artificial basket of stocks representing value and growth investments. In the indexes used, there is no double counting, in other words, there are no companies included simultaneously in the value and the growth indexes.

The performance of both indexes can be seen in Figure 1 and the risk adjusted comparison in Table 1. There were only three years in which the indexes moved in opposite directions. These years were 2001, 2006, and 2011. In all these three years the growth indexes had negative returns while the value indexes had positive returns. Of the 17 years analyzed the value index outperformed the growth index in 10 years. On a risk adjusted basis, the results are similar with the point estimate for the Sharpe ratio being bigger for 9 out of the 17 years analyzed. The point estimate for the correlation between the two indexes for the entire period was rather high, 0.931, but this correlation did change over time (Table 2). The smallest correlation for these two indexes was in 2005 (0.683) while the highest correlation was in 2007 (0.986).


Figure 1. Performance of MSCI Thailand Value Index and MSCI Growth Index (December 1999–December 2016).

Table 1. Value and growth index risk adjusted performance.

Year Return Volatility Sharpe
Value Growth Value Growth Value Growth
2000 −59.06 −42.94 25.37 18.44 −1.17 −1.06
2001 24.74 −11.70 −2.89 1.37 1.83 −0.80
2002 32.25 5.03 1.62 0.25 1.82 0.15
2003 144.14 87.88 126.67 77.23 1.85 1.93
2004 −4.15 −7.95 0.33 0.63 −0.53 −0.81
2005 4.72 16.41 0.77 2.69 −0.04 0.7
2006 1.14 −12.32 −0.14 1.52 −0.20 −1.00
2007 21.99 40.63 8.93 16.51 0.32 0.7
2008 −50.62 −46.70 23.64 21.81 −0.46 −0.61
2009 63.76 62.44 39.81 38.99 0.77 1.1
2010 39.41 33.6 13.24 11.29 0.55 0.59
2011 1.87 −3.73 −0.07 0.14 −0.03 −0.25
2012 21 33.16 6.97 11 0.47 0.77
2013 −8.75 −12.64 1.11 1.6 −0.36 −0.39
2014 5.75 21.47 1.23 4.61 0.06 0.41
2015 −25.01 −11.51 2.88 1.32 −0.37 −0.58
2016 24.64 20.27 4.99 4.11 0.49 0.42

Table 2. Correlation between value and growth index.

Period Correlation Period Correlation
2000–2016 0.931 2008 0.979
2000 0.926 2009 0.985
2001 0.766 2010 0.984
2002 0.752 2011 0.968
2003 0.981 2012 0.781
2004 0.693 2013 0.95
2005 0.683 2014 0.934
2006 0.699 2015 0.928
2007 0.986 2016 0.951

As a first step, the normality of the data was tested using an Anderson–Darling test. For the entire time series (from December 1999 to December 2017) the null assumption that the data follow a normal distribution is rejected as a 5% significance level. Nevertheless, it should be noted that when the test was performed for each individual year in the majority of the cases the Anderson–Darling test was unable to reject the hypothesis that the data follow a normal distribution (Table 3). As there are conflicting data regarding the issue of normality of distribution on these stocks returns and in accordance to the majority of the existing literature regarding this issue it was not assumed that the index returns follow a normal distribution. Hence, a non-parametric test was used. The non-parametric test used to compare both indexes was the Wilcoxon test. The null hypothesis of equal medians, comparing the MSCI Thailand Value and the MSCI Thailand Growth Index, was rejected in all cases (including when analyzing the entire time series together) with the exception of the 2016 period (Table 4).

Table 3. Anderson–Darling test results (p values).

Value Index Growth Index
Period p Value Period p Value Period p Value Period p Value
2000–2016 0.0005 2008 0.1356 2000–2017 0.0005 2008 0.0416
2000 0.0457 2009 0.0584 2000 0.7174 2009 0.1854
2001 0.6723 2010 0.5481 2001 0.6877 2010 0.142
2002 0.7148 2011 0.9548 2002 0.8705 2011 0.9158
2003 0.141 2012 0.2885 2003 0.2341 2012 0.3784
2004 0.4648 2013 0.6015 2004 0.4265 2013 0.5588
2005 0.5676 2014 0.6882 2005 0.1751 2014 0.6125
2006 0.1638 2015 0.0489 2006 0.99 2015 0.8385
2007 0.2003 2016 0.3447 2007 0.5052 2016 0.5252

Table 4. Wilcoxon test results (p values).

Period p Value Period p Value
2000–2016 0.00009 2008 0.0226
2000 0.0014 2009 0.0061
2001 0.00008 2010 0.0002
2002 0.0002 2011 0.00004
2003 0.0304 2012 0.00004
2004 0.00004 2013 0.00004
2005 0.00004 2014 0.00004
2006 0.00004 2015 0.0024
2007 0.00004 2016 0.5444

Another option instead of using indexes is to create portfolios of stocks directly according to some of the characteristics of value and growth investing. The approach followed for the creation of these indexes is similar to the one used in (Lakonishok et al. 1994). These authors used four metrics to classify companies into two categories; value and growth. One of the metrics that they used, and of the most frequently mentioned in the literature is the P/E ratio. First, a list of all the companies listed in the Bangkok Stock Exchange with positive earning as of December 1999 was obtained. Companies with extensive suspension periods were excluded from the index. It should be noted that the liquidity in some of those names was not too high with some of them not having daily trading. Only companies with relatively liquid stocks were included in the analysis. Those companies were grouped into four different groups according to their respective P/E values. For instance, the highest 25% of companies, according to their P/E were included in group one, the following 25% in group two and thereof. Some authors chose to use more subgroups , for instance in 10% intervals, but given the relatively small amount of stocks that satisfied our criteria in the Thai market on that date it seemed preferable to use a classification into four groups. The top and bottom groups, according to their P/E values, were selected to represent growth and value stocks. Each of these groups contained 16 stocks. An equal weight index was then created with all these 16 components. The returns on both indexes can be seen in Figure 2 and the correlation data in Table 5. Low P/E stocks are typically associated with value investments while high P/E stocks are typically associated with growth stocks.


Figure 2. P/E based indexes.

Table 5. Correlation between P/E based indexes.

Period Correlation Period Correlation
2000–2016 0.78174 2008 0.975461
2000 0.812026 2009 0.523414
2001 0.816988 2010 0.602125
2002 0.820738 2011 0.933659
2003 0.754619 2012 0.527316
2004 0.877633 2013 0.757883
2005 0.606471 2014 0.832368
2006 0.895574 2015 0.942001
2007 0.417671 2016 0.672125

Portfolios were also created using the cash flow per share metric. Similarly to the previous case a list of the companies listed in the Bangkok Stock Exchange at the end of December 1999 was used as a starting point. Then the cash flow per share was extracted from the data base Bloomberg for each of those stocks and arranged into four buckets. Only companies with positive cash flows were included. Due to these limitations, only 50 companies were left on the overall list. The top and bottom buckets contained 12 companies each. The returns of the indexes created using this criteria can be seen in Figure 3 and the correlation data in Table 6.


Figure 3. Cash flow per share indexes.

Table 6. Correlation between CF per share based index.

Period Correlation Period Correlation
2000–2016 0.4852 2008 0.848
2000 0.672 2009 0.6363
2001 0.2964 2010 −0.0332
2002 0.7588 2011 0.9453
2003 0.3188 2012 0.4962
2004 0.7007 2013 0.7799
2005 0.6963 2014 0.7726
2006 0.6692 2015 0.8096
2007 0.33 2016 0.4283

A third approach used to construct portfolios was to use the price to book value metric. Like in the previous cases, the starting point was the list of companies listed in the Bangkok Stock Exchange as of the end of December 1999. Then the price to book value metric was obtained from Bloomberg for each of the companies and arrange accordingly. The top and bottom buckets each contained 14 companies. Low price-to-book value is typically associated with value investment strategies while high price-to-book value is normally associated with growth stocks. The returns of the indexes created using this criteria can be seen in Figure 4 and the correlation data in Table 7.


Figure 4. Price-to-book value indexes.

Table 7. Correlation between price-to-book based index.

Period Correlation Period Correlation
2004–2016 0.441862 2010 0.907774
2004 0.592733 2011 0.323907
2005 0.215935 2012 0.375791
2006 0.717272 2013 0.869932
2007 0.421321 2014 0.672415
2008 0.744101 2015 0.621149
2009 0.373382 2016 0.593515

The final metric used for the classification of companies was the average five years sales growth for the companies. Due to data availability, the time series using this metric is shorter compared to the other metrics. This was necessary in order to maintain a reasonable number of stocks in each index. The starting data point, used for the classification of stocks was the end of December 2003, rather than the end of December of 1999 like in the previous cases. Also in this case 14 companies were included in the top and bottom buckets. The returns of the indexes created using this criteria can be seen in Figure 5 and the correlation data in Table 8.


Figure 5. Growth rate based indexes.

Table 8. Correlation between growth rate based indexes.

Period Correlation Period Correlation
2004–2016 0.5288 2010 0.4837
2004 0.7643 2011 0.8978
2005 0.6105 2012 0.3383
2006 0.7094 2013 0.8642
2007 −0.0192 2014 0.7182
2008 0.9152 2015 0.9297
2009 0.5545 2016 0.442

Similarly to the previous cases, in which the MSCI indexes were used, the first step was to do a test regarding the normality of the portfolio returns built using the previously mentioned four different metrics. For consistency considerations the Anderson–Darling test was selected as the appropriate test. Similarly to the previous cases, when the entire data series is analyzed the hypothesis that the returns follow a normal distribution can be rejected at a 5% significance level for most of the indexes. However, for the majority of the individual years such assumption cannot be rejected (Table 9, Table 10, Table 11 and Table 12).

Table 9. Anderson–Darling test results (p values) – P/E indexes.

High P/E (Growth) Index Low P/E (Value) Index
Period p Value Period p Value Period p Value Period p Value
2000–2016 0.0856 2008 0.6197 2000–2017 0.0266 2008 0.7659
2000 0.7589 2009 0.5882 2000 0.1636 2009 0.1152
2001 0.9174 2010 0.2188 2001 0.5051 2010 0.9869
2002 0.4234 2011 0.0497 2002 0.9001 2011 0.6515
2003 0.0898 2012 0.1058 2003 0.2042 2012 0.2395
2004 0.6387 2013 0.6381 2004 0.861 2013 0.141
2005 0.9702 2014 0.8968 2005 0.9532 2014 0.2548
2006 0.0588 2015 0.4995 2006 0.7878 2015 0.2732
2007 0.6345 2016 0.3836 2007 0.2395 2016 0.848

Table 10. Anderson–Darling test results (p values) – CF indexes.

High CF (Growth) Index Low CF (Value) Index
Period p Value Period p Value Period p Value Period p Value
2000–2016 0.0005 2008 0.5851 2000–2017 0.0005 2008 0.14
2000 0.7957 2009 0.1115 2000 0.094 2009 0.4685
2001 0.3336 2010 0.1156 2001 0.0012 2010 0.6809
2002 0.7295 2011 0.2771 2002 0.8003 2011 0.6201
2003 0.3014 2012 0.0049 2003 0.99 2012 0.8733
2004 0.0899 2013 0.9707 2004 0.1902 2013 0.9433
2005 0.0594 2014 0.8208 2005 0.4703 2014 0.2981
2006 0.2813 2015 0.3044 2006 0.8948 2015 0.6345
2007 0.9722 2016 0.0791 2007 0.5834 2016 0.3376

Table 11. Anderson–Darling test results (p values) – P/B indexes.

High P/B (Growth) Index Low P/B (Value) Index
Period p Value Period p Value Period p Value Period p Value
2000–2016 0.0887 2008 0.1867 2000–2017 0.0005 2008 0.0634
2000 0.5612 2009 0.9359 2000 0.1339 2009 0.8552
2001 0.9231 2010 0.9349 2001 0.0005 2010 0.4962
2002 0.6289 2011 0.8636 2002 0.8058 2011 0.1498
2003 0.1842 2012 0.8211 2003 0.1979 2012 0.3646
2004 0.1086 2013 0.6615 2004 0.7007 2013 0.4188
2005 0.9725 2014 0.7093 2005 0.2488 2014 0.587
2006 0.2162 2015 0.3643 2006 0.3334 2015 0.6184
2007 0.4302 2016 0.9677 2007 0.4859 2016 0.9822

Table 12. Anderson–Darling test results (p values) – growth indexes.

High 5-Year Growth Index Low 5-Year Growth Index
Period p Value Period p Value Period p Value Period p Value
2003–2016 0.0005 2010 0.9879 2000–2017 0.0005 2010 0.7783
2003 0.8536 2011 0.3257 2000 0.3796 2011 0.3355
2004 0.99 2012 0.5195 2001 0.9216 2012 0.7945
2005 0.2609 2013 0.8973 2002 0.2245 2013 0.1647
2006 0.9372 2014 0.1289 2003 0.2026 2014 0.7564
2007 0.6334 2015 0.6888 2005 0.207 2015 0.6352
2008 0.2836 2016 0.6251 2006 0.3588 2016 0.6175
2009 0.318     2007 0.4816    

Wilcoxon tests were then performed, as in the previous cases, to compare the indexes built using the P/E, P/B, Cash flow per share and five-year growth rate. Interestingly, the results of these tests (Table 13, Table 14, Table 15 and Table 16) fail to reject the hypothesis that the returns come from distributions with the same median. This might relate to the fact that perhaps using only one of these criteria to allocate companies into the value or growth categories is not enough and further analysis is needed. The issue of poor liquidity of some of the stocks included in the analysis is also acknowledged. While significant efforts were made to include only liquid securities, given the limited number of stocks that satisfied the previously mentioned criteria, some not highly liquid stocks were included in the portfolios. This, while a factor to take into account, is unlikely to be the sole factor between the discrepancy of the MSCI indexes and the portfolios built from individual stocks satisfying some broad market criteria such as P/E ratios.

Table 13. Wilcoxon Test results (p values) – P/E indexes.

Period p Value Period p Value
2000–2016 0.5275 2008 0.5834
2000 0.4025 2009 0.3123
2001 0.6236 2010 0.3123
2002 0.931 2011 0.8399
2003 0.665 2012 0.141
2004 0.8852 2013 0.665
2005 0.9999 2014 0.7508
2006 0.977 2015 0.665
2007 0.665 2016 0.3123

Table 14. Wilcoxon Test results (p values) – CF indexes.

Period p Value Period p Value
2000–2016 0.9348 2008 0.5834
2000 0.2602 2009 0.1939
2001 0.8852 2010 0.4025
2002 0.997 2011 0.8852
2003 0.2366 2012 0.4357
2004 0.977 2013 0.665
2005 0.977 2014 0.8852
2006 0.2855 2015 0.4357
2007 0.4357 2016 0.665

Table 15. Wilcoxon Test results (p values) – P/B indexes.

Period p Value Period p Value
2000–2016 0.578 2008 0.665
2000 0.2366 2009 0.8399
2001 0.6236 2010 0.665
2002 0.4705 2011 0.8852
2003 0.931 2012 0.0999
2004 0.8852 2013 0.665
2005 0.4357 2014 0.977
2006 0.5834 2015 0.3708
2007 0.3408 2016 0.5067

Table 16. Wilcoxon Test results (p values) – five-year growth indexes.

Period p Value Period p Value
2000–2016 0.8341 2010 0.931
2003 0.665 2011 0.999
2004 0.3708 2012 0.4025
2005 0.665 2013 0.997
2006 0.7075 2014 0.931
2007 0.3123 2015 0.3408
2008 0.5834 2016 0.795
2009 0.8399