1. Introduction
The term “frontier markets” has emerged over the last few years and describes certain countries that are investable but less established than emerging and developed markets.
Institutional investors, who utilize benchmarks to define investment universes, have generally followed classifications from index providers to define the set of countries that fall into this category.
Two of the largest index providers, Russell and MSCI, follow an approach focused on economic development and operational market criteria. The
Russell Global Index Series Construction and Methodology (2016) defines Russell’s economic criteria for the frontier market classification as being based on relative
income (as measured by the World Bank), development status (as published by the International Monetary Fund), and country risk (as measured by the Economist Intelligence Unit). With respect to market operational criteria, Russell focuses on FX (foreign exchange rate), repatriation and stock transfer restrictions, as well as liquidity, foreign ownership limits, and allowable account structures. The
MSCI Global Investable Market Indices Methodology (2015) follows a similar approach, but also establishes a minimum number of eligible stocks for each category.
As
Speidell (2011) notes, the methodologies differ in that Russell includes a frontier market country if a listed company does business there, regardless of where it trades, whereas MSCI does not; this results in a longer list of countries (notably Senegal and Papua New Guinea at the time Speidell published his book) appearing in the Russell indices.
Griffin et al. (2015) introduce a useful way to classify frontier markets at the early stage (those that are excluded from frontier indices for failing to meet minimum requirements), the conventional stage (countries that are current constituents of core frontier indices and are not under review to be upgraded to emerging markets), and the transitional stage (those that are under review to be upgraded —or have already been upgraded—to emerging status by some index providers). The authors analyze valuation and growth characteristics for each group and find differences in liquidity, commission levels, and economic freedom, and correlations with developed markets, across these three areas. The authors, however, do not delve into potential investment strategies within emerging markets, other than to advocate diversification in order to take advantage of the lower cross-sectional correlation existent in frontier markets.
In this chapter we present potential investment strategies by performing a detailed financial–statistical analysis to construct quality indicators and to assess the performance of different frontier markets equity portfolios that have been selected according to three quality dimensions.
Asness et al. (2013) define quality as a characteristic that “all else equal, an investor should be willing to pay a higher price for: stocks that are profitable, growing, and well managed.” They go further by defining a quality-minus-junk (QMJ) factor that focuses on the right-hand variables of the Gordon growth model (P/B = (profitability × payout ratio)/(required return—growth)).
The topic of quality investing was introduced in Benjamin
Graham’s (1973) Intelligent Investor. However, due to the difficulties in defining quality in an objective manner, this factor has received significantly less attention than other well-known factors, such as value, momentum, and size.
Joseph
Piotroski (2000) utilizes some simple accounting-based fundamental analysis and applies it to differentiating between winners and losers in firms that score well in book-to-value metrics.
Novy-Marx (2014) presents a good literature review of the quality factor, including some asset management industry publications, and performs quantitative tests on historical performance for the Russell 1000 and Russell 2000 universe using different definitions of quality. He defines quality in seven different ways and tests the historical performance after adjusting for market, size, and value. He obtains stronger results in the small capitalization universe (the Russell 2000) and suggests that the impact of quality considerations could be larger in more inefficient markets.
In the same venue of research,
Asness et al. (2015) arrive at a similar (almost identical) conclusion, arguing that the small-cap premium becomes stronger once lower quality stocks have been screened out.
In this chapter we introduce an investment strategy that selects corporations from the frontier markets universe based on several quality variables used in academia and in the industry. Given the restrictions of the data, we were able to select and classify corporations based on three value-oriented financial ratios (book-to-price (BP), earnings-to-price (EP), and earnings yield (EY)) and three quality oriented ratios (gross profits-to-assets (GPA), return on invested capital (ROIC), and return on equity (ROE)). These ratios were selected consistent with the availability of data in our sample.
Our results show that portfolios based on GPA produce the best risk-adjusted results, followed by the results from portfolios built on the ROIC. Also, we compare the countries selected in the high-quality portfolio with those described in the
Griffin et al. (2015) classification and determine that with the exception of Egypt and Mauritania (that are not considered in their paper), the vast majority of the 20 countries present in the portfolio of high-quality companies are classified as either conventional or transitional.
The chapter is structured as follows: In
Section 2 we describe the data used, define the quality variables, present the methodology followed to construct the portfolios, and describe several problems with the data and the way we create our final dataset. In
Section 3 we present the results, and in
Section 4 we conclude and present future venues of research.
2. Data and Methodology
In this section we briefly describe the data and the qualitative variables used in this chapter. Later, we describe the methodology and detail the steps followed in cleaning the data and comparing the results.
2.1. Data
The dataset in our analysis comes from Worldscope. We obtained quarterly data for 35 frontier markets (as defined by Russell Indexes) and a total of 3554 corporations (all corporations available in the database that are domiciled in those countries, regardless of their inclusion in any index). The sample period is the first quarter of 1998 through the fourth quarter of 2014. The numbers are expressed in US dollars.
In
Table 12.1 we present the countries in the sample and the number of corporations in each of them. It is important to note that many of these corporations have a small capitalization and are highly illiquid; however, since our purpose is to show whether there is a premium for quality investing in frontier markets, we treat all corporations as equal.
Table 12.1
Countries Present in the Dataset Used in the Analysis and Their Corresponding World Region and Number of Corporations for Dataset Countries
Country ID |
Country name |
Region |
Number of corporations |
ARG |
Argentina |
Latin America |
172 |
BGD |
Bangladesh |
Asia |
393 |
BGR |
Bulgaria |
Eastern Europe |
415 |
BHR |
Bahrain |
Middle East |
70 |
BWA |
Botswana |
Africa |
44 |
CYP |
Cyprus |
Eastern Europe |
301 |
EGY |
Egypt |
Africa |
330 |
EST |
Estonia |
Eastern Europe |
56 |
GAB |
Gabon |
Africa |
1 |
GHA |
Ghana |
Africa |
44 |
HRV |
Croatia |
Eastern Europe |
318 |
JAM |
Jamaica |
Latin America |
72 |
JOR |
Jordan |
Middle east |
279 |
KAZ |
Kazakhstan |
Asia |
105 |
KEN |
Kenya |
Africa |
74 |
KWT |
Kuwait |
Middle East |
269 |
LKA |
Sri Lanka |
Asia |
367 |
LTU |
Lithuania |
Eastern Europe |
112 |
MLT |
Malta |
Eastern Europe |
164 |
MUS |
Mauritania |
Africa |
396 |
NAM |
Namibia |
Africa |
36 |
NGA |
Nigeria |
Africa |
275 |
OMN |
Oman |
Middle East |
176 |
PAK |
Pakistan |
Asia |
906 |
PNG |
Papua New Guinea |
Asia |
21 |
QAT |
Qatar |
Middle East |
68 |
ROU |
Romania |
Eastern Europe |
1023 |
SRB |
Serbia |
Eastern Europe |
142 |
SVK |
Slovakia |
Eastern Europe |
176 |
SVN |
Slovenia |
Eastern Europe |
234 |
TTO |
Trinidad and Tobago |
Latin America |
30 |
TUN |
Tunisia |
Africa |
95 |
TZA |
United Republic of Tanzania |
Africa |
11 |
UKR |
Ukraine |
Eastern Europe |
381 |
VNM |
Vietnam |
Asia |
767 |
We also downloaded 37 financial indicators (
Table 12.2).
Table 12.2
Financial Indicators Present in the Dataset
Financial indicator
Cash earnings return on equity
Common equity
Common equity (USD)
Common shares outstanding
Common stock
Cost of goods sold (excluding depreciation)
Current assets (total)
Depreciation
Dividends
Dividends provided for or paid (common)
Earnings before interest and taxes (EBIT)
Gross income
Long-term debt
Market capitalization (USD)
Net cash from continuing operations
Net income available to common
Net income before extra items/preferred dividends
Net income before preferred dividends
Net income/starting line
Net margin
Net sales or revenues
Operating income
Preferred stock
Pretax income
Return on equity (per share)
Selling
Total asset turnover
Total assets
Total assets (USD)
Total assets (as reported)
Total debt
Total liabilities
Total liabilities and shareholders’ equity
Total shareholders’ equity
Trailing 12 months net profit (USD)
Trailing 12 months net sales/revenues (USD)
Working capital
In addition, we obtained monthly stock market returns for most of the corporations domiciled in the countries in the sample. These numbers are indices computed in local currencies. In the analysis that follows, we assume an exchange rate–neutral investor (ie, all the returns are expressed in local currency); while this is a simplifying assumption, we believe this is a relevant analysis because the majority of the markets are still dominated by local investors. In addition, as we will see in
Section 3 later, our sorting methodology (based on defined financial ratios) yields diversified portfolios across countries.
We used the indices to construct the return series used to evaluate the quality strategies. Moreover, we assume that investors have no particular interest in any one specific country and that costs related to capital mobility are not significant. These assumptions go back to our main objective of exploring the existence of a quality premium while understanding that investability might still be limited.
2.2. Quality Variables
The basic premise of the chapter is to test whether GPA, as suggested by Novy-Marx, constitutes a quality dimension that can help investors select profitable stocks in frontier markets. In order to do this, and to increase the robustness of our study, we also construct five additional financial indicators commonly used in the industry. BP, EP, and EY are value metrics that we use for comparison purposes. ROIC and ROE are quality metrics to build robustness around the idea of the feasibility of quality investing in frontier markets. A definition for each indicator is presented next.
• Gross profits-to-assets (GPA):
GPA=Revenues−COGSTotal assets
(12.1)
• Book-to-price (BP):
BP=Book equityMarket equity
(12.2)
where market equity is market capitalization and book equity is common equity.
• Earnings-to-price (EP):
EP=Net incomeMarket equity
(12.3)
where net income is net income/starting line.
• Earnings yield (EY):
EY=EBITEnterprise value
(12.4)
where enterprise value (EV):
EV=Market equity+long term debt+current liabilities+preferred stock−cash and short term investments
(12.5)
where net cash from continuing operations is used as a proxy for cash and short-term investments.
• Return on invested capital (ROIC):
ROIC=EBITTangible capital
(12.6)
where tangible capital (TC) is defined as:
TC = Property, plant, and equipment + working capital
(12.7)
where TC is estimated as:
TC=Total assets−current assets+working capital
(12.8)
• Return on equity (ROE):
ROE=Net incomeCommon equity
(12.9)
2.3. Methodology
After cleaning the data (as described next) we built the six financial ratios described previously (GPA, BP, EP, EY, ROIC, and ROE) for each company in our sample. We determined the company’s average and median value across all quarters and ranked it accordingly, without consideration to market cap. We decided to sort the data into four portfolios based on this ranking: high, medium high, medium low, and low. (For example, the high portfolio based on GPA contains 25% of the corporations with the highest GPAs). Finally, we evaluated the portfolio returns based on each of the rankings of each of the six financial indicators.
2.4. Cleaning the Data
The data used has several challenges common to frontier (and even emerging) markets. Among the problems found are:
• Incomplete data at the corporate level. Not all the corporations have sufficient financial indicators to construct the financial ratios needed to build the portfolios. In these cases, we removed the corporations.
• Incomplete time series by corporation.
• Not all of the corporations have data for every year. Through an in-depth analysis, we observed that most corporations started to report their data more frequently after 2008 (at the beginning of our sample period), and that most of the corporations had available data until the third quarter of 2014 (the end of our sample period). This is 27 quarters of data.
• Not all the corporations have data in every quarter in our sample period. We only included corporations with at least 14 quarters available and with more than 90% of those 14 quarters belonging to the later part of the sample (to avoid corporations that were left the sample due to bankruptcy or other unknown reasons).
After limiting our data to those corporations with complete information, our sample size was reduced from 3554 to 1446 corporations. Our sample was further reduced when we matched each corporation with its monthly return series. In that pairing, several corporations that were present in the quarterly data were either not available or with not enough return information in the monthly data. We removed those corporations from both datasets (quarterly and monthly) because we needed the stock returns to test the profitability of the strategies built based on the aforementioned six financial ratios.
Our final sample was thus reduced from 3554 to 1042 corporations and from 35 to 22 countries. In
Table 12.3 we present the countries and the number of corporations considered in each for them in the final analysis.
Table 12.3
Countries and the Number of Corporations per Country, Used in the Analysis
Country ID |
Country name |
Number of corporations |
ARG |
Argentina |
55 |
BHR |
Bahrain |
15 |
CYP |
Cyprus |
7 |
EGY |
Egypt |
114 |
EST |
Estonia |
10 |
GHA |
Ghana |
7 |
HRV |
Croatia |
54 |
JAM |
Jamaica |
12 |
JOR |
Jordan |
70 |
KAZ |
Kazakhstan |
2 |
KWT |
Kuwait |
158 |
LKA |
Sri Lanka |
91 |
LTU |
Lithuania |
21 |
MLT |
Malta |
1 |
MUS |
Mauritania |
2 |
OMN |
Oman |
35 |
PAK |
Pakistan |
162 |
QAT |
Qatar |
16 |
ROU |
Romania |
2 |
SVK |
Slovakia |
2 |
SVN |
Slovenia |
10 |
VNM |
Vietnam |
195 |
Before moving to the results section,
Table 12.4 presents some economic data for the countries included in the sample; we will use this information later in our results, as we observed some interesting relationships between these economic variables and the representation of each country in the different strategy portfolios.
Table 12.4
Economic and Population Indicators of the Countries Used in the Portfolio Analysis
Country |
GDP PPP (in billions USD) |
Real GDP growth (%) |
GDP per capita (USD PPP) |
Population |
Unemployment rate (%) |
Argentina |
927 |
−1.7 |
22,100 |
43,024,374 |
8 |
Bahrain |
62 |
3.9 |
51,400 |
1,314,089 |
4 |
Cyprus |
25 |
−3.2 |
28,000 |
1,172,458 |
16 |
Egypt |
945 |
2.2 |
28,000 |
86,895,099 |
13 |
Estonia |
35 |
1.2 |
26,600 |
1,257,921 |
9 |
Ghana |
109 |
4.5 |
4,200 |
25,758,108 |
11 |
Croatia |
87 |
−0.8 |
20,400 |
4,470,534 |
21 |
Jamaica |
24 |
1.1 |
8,700 |
2,930,050 |
14 |
Jordan |
80 |
3.0 |
11,900 |
7,930,491 |
12 |
Kazakhstan |
421 |
4.6 |
24,100 |
17,948,816 |
5 |
Kuwait |
284 |
1.4 |
71,000 |
2,742,711 |
3 |
Sri Lanka |
217 |
7.0 |
10,400 |
21,866,445 |
4 |
Lithuania |
79 |
3.0 |
26,700 |
3,505,738 |
11 |
Malta |
13 |
2.2 |
31,700 |
412,655 |
6 |
Mauritania |
13 |
6.8 |
3,400 |
3,516,806 |
31 |
Oman |
164 |
3.4 |
44,100 |
3,219,775 |
15 |
Pakistan |
884 |
4.1 |
4,700 |
196,174,380 |
7 |
Qatar |
323 |
6.5 |
92,400 |
2,123,160 |
0 |
Romania |
387 |
2.4 |
19,400 |
21,729,871 |
7 |
Slovakia |
150 |
2.4 |
27,700 |
5,443,583 |
13 |
Slovenia |
61 |
1.4 |
29,400 |
1,988,292 |
14 |
Vietnam |
510 |
5.5 |
5,600 |
93,421,835 |
3 |
US |
17,460 |
2.4 |
54,800 |
318,892,103 |
6 |
European Union |
17,610 |
1.4 |
38,300 |
511,434,812 |
10 |
China |
17,630 |
7.4 |
12,900 |
1,355,692,576 |
4 |
Source: CIA world fact book.
All values are 2014 estimates. We have included the United States, the European Union and China for comparison.
3. Results
In this section we present the portfolio results based on rankings determined by each of the six financial ratios explained in
Section 3. In the first part of this section we present the results for which the rankings are driven by the weighted average values per corporation of each of the financial ratios. In the second part
we provide the results based on rankings determined by median values. The use of median values is a robustness check against possible outliers that may be present in the data and that we were not able to identify.
Recall that we create four portfolios per financial indicator—high, medium high, medium low, and low—based on number of corporations and following both a cap-weighted and an equal-weighted approach. Based on this, in the last subsection we concentrate our efforts on understanding whether there are significant differences in portfolio results based only on the GPA indicator, as suggested by Novy-Marx.
3.1. Portfolio Results Based on Weighted Average Values of Financial Ratios
In this section we compare the portfolio results for the top 25% ranking (by number of stocks) of each financial ratio presented in
Section 3. The rankings are driven by the average values per corporation for each of the financial ratios. Within each portfolio, each company is weighted by its market cap relative to the market cap of all corporations in that portfolio. In
Table 12.5 we present the findings.
Table 12.5
Monthly Portfolio Results From Top 25% Ranking of Every Financial Ratio
Statistic |
BP |
EP |
EY |
ROIC |
GPA |
ROE |
Mean return annualized |
2.9% |
6.6% |
8.4% |
8.6% |
9.7% |
4.4% |
Median return annualized |
6.0% |
7.3% |
9.4% |
9.4% |
9.6% |
6.0% |
Annualized standard deviation |
5.0% |
4.9% |
5.3% |
5.4% |
5.1% |
6.6% |
Maximum return annualized |
59.6% |
71.2% |
128.9% |
128.9% |
128.9% |
128.9% |
Minimum return annualized |
−37.3% |
−43.6% |
−43.6% |
−55.2% |
−45.2% |
−55.2% |
Sharp ratio on monthly observations |
0.17 |
0.38 |
0.44 |
0.44 |
0.53 |
0.19 |
Number securities |
260 |
260 |
260 |
260 |
260 |
260 |
Number returns >0 |
158 |
178 |
183 |
184 |
184 |
162 |
Number countries in portfolio |
12 |
12 |
17 |
22 |
22 |
21 |
The rankings are driven by the average values per corporation for each of the financial ratios. The portfolio weights are given by their market capitalization relative to the total market cap of the corporations in the portfolio. The returns are expressed annualized percentages. In this table we present results based on book-to-price (BP), earnings-to-price (EP), earnings yield (EY), return on invested capital (ROIC), gross-profit-to-assets (GPA), and return-to-equity (ROE). The mean annualized returns are significantly different at the 5% significance levels for the three quality variables used. The data spans from the first quarter of 2008 to the third quarter of 2014. The returns used have monthly frequency.
From this table we can see that in terms of average portfolio returns, ROIC and GPA portfolios are the best ones (even when compared to value factors such as BP, EP, and EY). However, considering the risk-adjusted returns (average divided by the standard deviation), the GPA portfolio is the clear winner (0.53).
In
Table 12.5 and through this chapter we present absolute returns. This is because our dataset is strictly confined to Worldscope, and, as noted earlier, after filtering companies for data availability some key benchmark counties ended up being excluded from the analysis. Additionally, our objective is to determine whether—given available data—companies with more quality characteristics outperform those with fewer quality characteristics; a focus on absolute returns suffices for this.
The use of portfolios weighted by market capitalization brings some realism to the investment strategies but also introduces potentially higher idiosyncratic risk (larger-cap companies will have more representation). To control for this, we also ran the analysis based on equal weighted portfolios and the results were very similar as shown in the next subsection.
The number of positive absolute returns provides an interesting piece of information. In the case of the corporations that belong to the BP top quartile portfolio, 158 of 260 (57%) had a positive average return for the period of analysis. In contrast, 184 of 260 (70%) had a positive average return in the case of the GPA portfolios.
One important thing to note is that the top quartile portfolios based on BP and EP are heavily concentrated in very small number of countries; this is possibly driven by systematic undervaluation driven by political situations at different points in time. This is clearly not the case for the portfolios created using
ROIC, GPA, and ROE. In
Table 12.6 we present the average country weights over the period of study for each of the portfolios presented in
Table 12.5.
Table 12.6
Country Weights for Portfolios From Top 25% Ranking of Every Financial Ratio
Country ID |
BP |
EP |
EY |
ROIC |
GPA |
ROE |
ARG* |
1.2% |
1.5% |
8.5% |
9.2% |
10.8% |
8.5% |
BHR* |
|
|
0.4% |
2.3% |
2.3% |
2.3% |
CYP* |
0.4% |
0.4% |
0.8% |
1.5% |
1.2% |
1.5% |
EGYNP
|
0.8% |
2.7% |
18.5% |
16.5% |
15.0% |
16.2% |
EST* |
|
|
0.4% |
1.9% |
2.3% |
1.2% |
GHA** |
|
|
|
0.8% |
0.8% |
0.4% |
HRV* |
0.4% |
0.8% |
5.4% |
4.2% |
4.6% |
3.1% |
JAM** |
0.8% |
2.3% |
2.7% |
0.8% |
1.2% |
|
JOR* |
|
|
1.2% |
2.7% |
3.1% |
2.7% |
KAZ* |
0.8% |
0.8% |
0.4% |
0.4% |
0.4% |
0.4% |
KWT*** |
|
|
1.9% |
14.2% |
13.1% |
21.2% |
LKA* |
3.1% |
5.4% |
10.8% |
6.2% |
6.9% |
4.6% |
LTU* |
|
|
0.8% |
1.9% |
3.1% |
1.2% |
MLT** |
|
|
0.4% |
0.4% |
0.4% |
0.4% |
MUSNP
|
|
|
0.8% |
0.8% |
0.8% |
0.8% |
OMN* |
|
|
0.4% |
3.5% |
3.8% |
3.5% |
PAK* |
16.5% |
22.3% |
22.3% |
16.2% |
13.8% |
16.9% |
QAT*** |
1.9% |
0.8% |
4.6% |
5.0% |
5.0% |
5.4% |
ROU* |
0.4% |
0.4% |
0.8% |
0.8% |
0.8% |
0.8% |
SVK* |
|
|
0.4% |
0.4% |
0.8% |
0.4% |
SVN* |
|
|
0.8% |
1.5% |
1.9% |
1.5% |
VNM* |
73.5% |
62.3% |
18.1% |
8.8% |
8.1% |
7.3% |
No. countries in portfolio |
12 |
12 |
17 |
22 |
22 |
21 |
Frontier market country classification according to Griffin et al. (2015). The symbol * indicates a country is at the conventional stage, ** is early stage, *** is transitional stage, and NP indicates the country not present in their classification.
In
Table 12.6 we can observe that portfolios created using BP and EP are heavily concentrated in corporations located in Vietnam and Pakistan, creating potential concentration risk (at the country and regional levels). The weights of these two countries when focusing on these two financial ratios are 89.9 and 84.6%, respectively.
The portfolios created with EY show an investment concentration in four countries (Egypt, Sri Lanka, Pakistan, and Vietnam), with a combined weight of 69.7%; however, in an equal-weighted analysis, 70% of the portfolio is concentrated in three countries. The last three variables (ROIC, GPA, and ROE) provide investors with more diversified portfolios in both equal- and cap-weighted scenarios.
In
Table 12.6 we also classify frontier market countries following
Griffin et al. (2015). According to our results, 15 countries are conventional, 3 are early stage, and 2 are transitional frontier markets. Also it is important to note that
68% (15 out of 22) countries that are present in the high-quality portfolios are conventional, 14% (3 out of 22) are early stage, 9% (2 out of 22) transitional, and 9% (2 out of 22) are not considered in the
Griffin et al. (2015) classification. Also it is important to note that
Griffin et al. (2015) consider 23 frontier markets to be conventional; this means that 65% (15 out of 23) of the countries that are present in our analysis are classified as conventional. In other words, we arrive at the conclusion that conventional and transitional countries are mostly the ones for which there is enough data to run a quantitative analysis of quality characteristics, and also the ones that are represented in a portfolio of high-quality companies.
3.2. Portfolio Results Based on Median Values of Financial Ratios
This section is different from the previous one in terms of the statistic used to rank BP, EP, EY, ROIC, GPA, and ROE, as well as in terms of the weighting methodology used to construct the portfolios. A potential problem with simple averages is that they can be seriously affected with extreme values (outliers). Also, the use of capitalization weighting portfolios can introduce idiosyncratic risk. To deal we both problems we used the median value—a central-tendency statistic that is immune to outliers—and we construct equal-weighed portfolios.
The results presented in this section are thus based on a ranking using the median values of the financial ratios considered in this chapter. Again, we present the portfolio results of the top 25% ranking of each financial ratio presented in
Section 3.
Table 12.7 presents the results for these portfolios.
Table 12.7
Monthly Portfolio Results From Top 25% Median Rankings of Every Financial Ratio
Statistic |
BP |
EP |
EY |
ROIC |
GPA |
ROE |
Annualized average returns |
1.2% |
6.2% |
10.0% |
12.7% |
11.4% |
10.0% |
Annualized median returns |
2.4% |
6.2% |
12.7% |
12.7% |
12.7% |
12.7% |
Standard deviation |
5.5% |
5.2% |
5.2% |
5.5% |
5.2% |
6.2% |
Maximum return annualized |
67.7% |
67.7% |
127.8% |
127.8% |
127.8% |
127.8% |
Minimum return annualized |
−37.2% |
−37.2% |
−45.3% |
−45.3% |
−48.0% |
−55.4% |
Risk-adjusted returns based on monthly observations |
0.09 |
0.32 |
0.56 |
0.62 |
0.61 |
0.46 |
Number of securities |
260 |
260 |
260 |
260 |
260 |
260 |
Number returns >0 |
149 |
171 |
194 |
195 |
202 |
198 |
Number countries in portfolio |
6 |
6 |
12 |
18 |
21 |
20 |
The rankings are driven by the median values per corporation for each of the financial ratios. In this table we present results based on book-to-price (BP), earnings-to-price (EP), earnings yield (EY), return on invested capital (ROIC), gross-profit-to-assets (GPA), and return-to-equity (ROE). The data spans from the first quarter of 2008 to the third quarter of 2014. The returns used have monthly frequency. The mean annualized returns are significantly different at the 5% significance levels for the three quality variables used
The results from this table are very similar to the ones presented in
Table 12.5, in which the ranking was made based on the average values of the financial ratios. This can be considered as a robustness check of our results.
Once again, the quality metrics appear on top. From this table we can see that in terms of average portfolio returns, ROIC is the best metric. In terms of median portfolio returns, once more ROIC portfolios are the best ones. In this case, considering the risk-adjusted returns (average divided by the standard deviation), the ROIC and GPA portfolio are similar (0.621 and 0.614, respectively).
Comparing the number of positive returns with the ones presented in
Table 12.5, we can see that there is a slight improvement for EP, EY, ROIC, and ROE. Positive returns improve by 2, 20, 7, and 1 for EP, EY, ROIC, and ROE, respectively. GPA positive average portfolio returns decreased by 1, to 202.
Again, the value portfolios based on BP, EP, and EY are heavily concentrated in a very small number of countries (6, 6, and 12 for BP, EP, and EY, respectively). This is clearly not the case for the quality portfolios created using ROIC, GPA, and ROE. The country weights are very similar to those presented in
Table 12.5.
3.3. Portfolio Results Based on the Gross-Profit-to-Assets
In this section we consider only the portfolios created using the GPA ratio.
The objective is to see if there are significant differences in profitability and risk among the portfolios created by separating the corporations into four groups based on the rankings of GPA and weighting each portfolio based on market cap. The portfolio names are high (top 25%), medium high, medium low, and low (last 25%). The rankings were made based on average and median values of each corporation’s GPA. The portfolio returns information is presented in
Table 12.8
Table 12.8
Results for Portfolios Created by Separating the Corporations Into Four Groups Based on the Rankings of GPA
Statistic |
High
|
High
|
Medium high
|
Medium high
|
Medium low
|
Medium low
|
Low
|
Low
|
Average returns |
9.7% |
10.1% |
2.8% |
2.7% |
−2.9% |
−1.9% |
−7.9% |
−9.1% |
Median returns |
9.6% |
9.8% |
3.5% |
3.4% |
−2.1% |
−1.6% |
−8.6% |
−9.4% |
Standard deviation |
5.1% |
4.9% |
5.4% |
5.6% |
5.5% |
5.3% |
6.0% |
6.0% |
Maximum returns |
128% |
128% |
101.0% |
101.0% |
59.6% |
59.6% |
67.2% |
67.2% |
Minimum returns |
−45.2% |
−36.8% |
−54.8% |
−54.8% |
−48.1% |
−48.1% |
−65.0% |
−65.0% |
Risk-adjusted returns based on monthly observations |
0.53 |
0.57 |
0.15 |
0.14 |
−0.15 |
−0.11 |
−0.40 |
−0.46 |
Number of corporations |
260 |
260 |
260 |
260 |
260 |
260 |
262 |
262 |
Number returns >0 |
184 |
187 |
156 |
155 |
120 |
125 |
83 |
76 |
Number countries in portfolio |
22 |
22 |
17 |
17 |
15 |
14 |
17 |
18 |
The portfolio names are high (top 25%), medium high, medium low and low (last 25%). The data spans from the first quarter of 2008 to the third quarter of 2014. The returns used have monthly frequency.
Rankings based on average GPA values.
Rankings based on median GPA values.
The first thing to note is that there are significant differences in average and median returns between the high and medium-high portfolios compared to the other two portfolios (medium low and low). Also, it is clear that the risk-adjusted returns of the high portfolios are almost double those of the medium-high portfolios.
In terms of country diversification, the high portfolio is the most diversified one. In
Table 12.9 we present the details of the country weights for each of these portfolios. This table shows that the corporations that made up the four portfolios are well diversified across the board. This could imply that the difference in portfolio performance is thus mainly based on the characteristics of every component, selected according to the quality financial ratio GPA.
Table 12.9
Country Weights for the Portfolios Created by Separating the Corporations Into Four Groups Based on the Rankings of GPA
Country ID |
High (%) |
Medium high (%) |
Medium low (%) |
Low (%) |
ARG |
10.8 |
5.4 |
3.5 |
1.5 |
BHR |
2.3 |
1.5 |
1.5 |
0.4 |
CYP |
1.2 |
0.8 |
|
0.8 |
EGY |
15.0 |
10.8 |
10.4 |
7.6 |
EST |
2.3 |
1.2 |
|
0.4 |
GHA |
0.8 |
0.8 |
|
1.1 |
HRV |
4.6 |
4.6 |
6.2 |
5.3 |
JAM |
1.2 |
1.5 |
1.5 |
0.4 |
JOR |
3.1 |
3.5 |
8.5 |
11.8 |
KAZ |
0.4 |
|
|
0.4 |
KWT |
13.1 |
21.5 |
15.4 |
10.7 |
LKA |
6.9 |
7.7 |
10.4 |
9.9 |
LTU |
3.1 |
1.9 |
1.5 |
1.5 |
MLT |
0.4 |
|
|
|
MUS |
0.8 |
|
|
|
OMN |
3.8 |
5.8 |
1.9 |
1.9 |
PAK |
13.8 |
20.8 |
12.7 |
14.9 |
QAT |
5.0 |
0.8 |
0.4 |
|
ROU |
0.8 |
|
|
|
SVK |
0.8 |
|
|
|
SVN |
1.9 |
0.4 |
0.4 |
1.1 |
VNM |
8.1 |
11.2 |
25.8 |
29.8 |
Portfolio names are high (top 25%), medium high, medium low, and low (last 25%). The data spans from the first quarter of 2008 to the third quarter of 2014.
The country weights are similar in the case of the portfolios based on median rankings.
For these particular cases, we present a very simple correlation analysis in order to see if there is a linear relationship between the country weights and the economic and population indicators presented in
Table 12.4.
In
Table 12.10 we can observe several interesting results. First, most of the portfolio weights are highly correlated with the countries’ GDP. Recall that the GDP is used as a measure of economic development. Thus, this positive and significant linear relationship makes perfect sense and gives support to the idea that quality is important in frontier markets.
Table 12.10
Correlations Between Economic and Population Indicators and Portfolio Weights
|
High |
Medium high |
Medium low |
Low |
GDP PPP (in billions USD) |
0.60 |
0.66 |
0.47 |
0.30 |
Real GDP growth |
0.19 |
0.29 |
0.01 |
0.10 |
GDP per capita (USD PPP) |
−0.37 |
−0.56 |
0.00 |
0.23 |
Population |
0.79 |
0.81 |
0.63 |
0.24 |
Unemployment rate |
−0.29 |
−0.21 |
−0.19 |
0.24 |
Second, the portfolio weights are also positively and significantly correlated to the population size of the countries. Population is regularly considered as a market-size and market-potential indicator. Again, this result is consistent with intuition.
Finally, it is interesting to note that the countries weights are negatively related to the unemployment rates. There are a couple of plausible reasons for this. First, a high unemployment rate can be considered a negative signal for investors who see countries with high unemployment rates as less developed economically and socially, and second, that countries with more investment (higher portfolio weights) are the ones that have lower unemployment rates. Thus, a double causality could be in play here. We have left this interesting topic for future research.
All of the other correlations present mixed results.
A final word before finishing this section: results based on the other financial ratios presented in
Section 3 show the same pattern. First, differences in results based on rankings derived from the averages versus the medians are not significant. Second, in most of the cases the differences in average portfolio returns between high and medium-high portfolios are minimal.
However, the risk-adjusted returns are also larger for the high portfolios most of the time.
4. Conclusions and Further Research
This chapter presents a way to build portfolios with frontier market stocks based on six financial ratios: BP, EP, and EY as ways to measure value, and GPA, ROIC, and ROE as ways to measure quality.
Our results show that portfolios based on GPA produce the best risk-adjusted results, followed by the results from portfolios built on the ROIC. These results conform to the ones obtained by
Novy-Marx (2014), who found similar results in the United States using Russell 1000 and Russell 2000 stocks.
Our results are robust to possible outliers in the value and quality indicators, since the ordering based on the average and median values of the indicators provide very similar results.
Our results also show that risk-adjusted returns decrease with the quality variable, that is, the high-quality portfolio has better risk-adjusted returns than those of lower quality (in terms of GPA).
In terms of country diversification and for all the cases analyzed, quality characteristics tended to produce more diversified portfolios, whereas value indicators were concentrated at times.
Other interesting findings appear when we compare the portfolio composition to certain economic and population metrics. First, most of the portfolio weights are highly correlated with the countries’ GDP. As GDP is used as a measure of economic development, this positive and significant linear relationship is consistent with intuition. Second, the portfolio weights are also positively and significantly correlated to the countries’ population size. Population is regularly considered as a market-size and market-potential indicator. Again, this result is consistent with intuition. Finally, the countries’ weights are negatively related to unemployment rates.
We also compare the countries selected in the high-quality portfolio with those described in the
Griffin et al. (2015) classification and determine that, with the exception of Egypt and Mauritania (that are not considered in their paper), the vast majority of the 20 countries used in our research are classified as either conventional or transitional frontier markets. It is important to note that
Griffin et al. (2015) consider 23 frontier markets to be conventional. This means that 65% (15 out of 23) of the countries that are present in our analysis fall within the conventional category.
Given the results presented in this chapter, a natural extension could be to perform a more in-depth analysis of the high-quality portfolios to see whether there are significant differences that can be exploited by investors; to do this, we would have to incorporate liquidity metrics, place more emphasis on cap-weighted portfolios, and establish a rebalancing frequency (rather than keeping the set of quality companies constant through the sample period). Also, our expectation is that the availability of data will increase over time in frontier markets; our analysis in this chapter was constrained to one data source, but typical professional money managers have the ability to combine several sources to arrive at a more complete set of information. Future research should try to include more information sources. A better dataset would also provide opportunities to build more financial ratios and quality metrics that could be compared to gross profitability.
Furthermore, the definition of quality used by Novy-Marx (ie, GPA) seems to be applicable to frontier markets; this suggests that other inefficient spaces—such as emerging markets—will likely benefit from this definition too. Exploring this effect in emerging markets is another natural avenue for future research.
In this chapter we assumed a currency-neutral investor, and we believe this is an acceptable assumption, given the considerable importance of local participants in each market. However, for international investors the impact of currency effects can be significant, and an analysis of this impact (and the extent to which it is present for different types of corporations) constitutes another interesting area for future research.