Capítulo 2: Análisis de la situación en la UCI y propuesta de solución
2.3 Propuesta de método de evaluación de factibilidad para proyectos informáticos
2.3.7 Procedimiento para el análisis de factibilidad económico
Burmeister and Wall (1986), Berry et al. (1988) and Chan et al. (1990) successfully model the return generating process of securities compromising the US market. However, it is by no means certain that the APT framework can successfully be applied to developing markets.
Burmeister (2003) acknowledges that while linear factor models of the return generating process can successfully be constructed for a number of developed markets (e.g., the UK, Germany and Japan), there are possible difficulties for developing markets.
Kandir (2008) investigates the role of macroeconomic factors in explaining returns on Turkish stock portfolios over the July 1997 to June 2005 period. The APT framework is employed to model the return generating process and Kandir (2008) first traces the development of asset pricing models and their contribution to providing a conceptual framework. In doing so, the influence of asset pricing models and notably the APT on contemporary literature dealing with time series models of stock returns is demonstrated.
Chen et al. (1986) are credited for motivating the extensive study of linkages between returns and macroeconomic factors representative of systematic risk within a multifactor framework.
The influence of the APT framework is further evident in the identification of risk factors;
Kandir (2008) refers to Chen et al. (1986) and Clare and Thomas (1994) amongst others49 to identify macroeconomic factors for inclusion in the return generating process specification of Turkish stock returns. A multifactor model incorporating changes in industrial production, consumer prices, money supply, the exchange rate, short-term interest rates and returns on the MSCI World Index is proposed to explain Turkish stock returns. Aside from Kandir’s (2008) direct references to the APT framework, it is clearly evident that this model reflects the influence of the framework in structure and factor composition.
Returns to be explained are the returns on size, book-to-market, earnings-to-price and leverage ratio sorted portfolios. The model is estimated by regressing returns on the abovementioned factors using the LS methodology. Kandir (2008) reports that the factors that are statistically significant in the return generating process are the exchange rate, the short-term interest rate and returns on the MSCI World Index. Together, these factors explain approximately 30 percent of variation in returns. It is noteworthy that the identity of significant factors does not change according to portfolio formation criteria. This suggests
49 One of these studies is that of Chen (1991) who draws upon APT framework and Chen et al. (1986) in his study of the predictability of stock returns.
that the limitation of the APT model noted by Clare and Thomas (1994) whereby the identity of priced factors differs according to portfolio formation criteria, does not affect inferences relating to the return generating process.50 Kandir (2008) refers to Chen et al. (1986) to establish the consistency of the results. The finding that industrial production and the inflation rate are statistically insignificant is inconsistent with the findings of Chen et al.
(1986) who find that these factors explained expected returns. However, these findings suggest that these factors do not explain the time series behaviour of returns or are irrelevant to the Turkish market. Further interpretations of significant relationships are provided by Kandir (2008); exchange rates are significant because of increasing trade and tourist activities, interest rates impact Turkish stock returns because of their role as a proxy for alternative investment opportunities and the MSCI Word Index plays an important role as a result of Turkey’s increasing integration within world markets (see Clare & Priestley, 1998).
In contrast, it is argued that oil prices do not have a significant impact upon returns as there may be more important factors of production for Turkish companies and industrial production does not affect returns because of the under-development of the Turkish stock market. The insignificant relationship between stocks and inflation suggests that Turkish stocks are not a hedge against inflation. The money supply does not appear to impact real activity.
Kandir (2008) applies the APT framework within the context of a developing market. In modelling the return generating process, the APT framework is extensively treated as a conceptual basis for the structure and composition of the return generating process specification. While a number of macroeconomic factors are identified in the return generating process, the impact of factors such as industrial production, oil prices and inflation – factors which play an important role in developed markets – is statistically insignificant.
This provides support for Burmeister (2003) postulation that there are difficulties in constructing linear factor models of the return generating process in developing markets. The foremost difficulty is the identity of the risk factors themselves; factors that describe the return generating process are likely to differ across markets and need to be identified in these markets.
While Kandir (2008) applies the APT framework to a single developing market, Bilson et al.
(2001) investigate the return generating process of twenty emerging stock markets over the
50 Not only are the same factors statistically significant for returns on portfolios formed according to differing criteria, explanatory power and estimated factor sensitivities are consistent.
January 1985 to December 1997 period.51 As such, Bilson et al.’s (2001) study is far more extensive in scope. It is argued that two variants of multifactor models exist. The first assumes perfect integration with returns being driven by global risk factors whereas the second, primarily informed by the work of Chen et al. (1986), assumes complete segmentation whereby domestic factors drive returns. Bilson et al. (2001) list goods prices, oil prices, the money supply, real activity, exchange rates, interest rates and trade factors as some of the risk factors that drive returns. Although the authors recognize that the selection of factors is subject to criticism on the grounds that selection is subjective and arbitrary, guidance is provided by prior research and the role of APT literature in factor identification is evident in the abovementioned list. This list includes factors considered by Chen et al.
(1986), Hamao (1988), Beenstock and Chan (1988), Clare and Thomas (1994) and Priestley (1996). A further problem is the identity of the international risk factor which Bilson et al.
(2001) argue can be represented by returns on a value-weighted world index. The use of an international risk factor again finds support in APT literature (see Clare & Priestley, 1998).
Bilson et al. (2001) however extend the theory by suggesting that regional influences may also play a role if countries are regionally integrated.
An initial a five-factor model motivated by APT literature incorporating four domestic factors and returns on a value-weighted world market index, the MSCI World Index, is chosen by Bilson et al. (2001) to model returns. Changes in the money supply, the prices of goods (inflation), real activity and exchanges rates represent domestic factors. Bilson et al. (2001) acknowledge that while the APT framework employs unexpected components, factors in the study are employed in their raw form.52 The base specification is further extended to include a political risk measure, a trade sector factor, interest rates, a regional factor, the price-to-earnings ratio and the (aggregate) dividend yield. Although, the last two factors are not fully consistent with the approach of Chen et al. (1986) who hypothesize returns to be a function of macroeconomic factors and non-equity factors, their inclusion is justified by the ability of factors constructed from the same market to better explain returns relative to macroeconomic factors (see Chen, 1991; Van Rensburg, 2000).
51 Argentina, Brazil, Chile, Colombia, Greece, India, Indonesia, Jordan, Korea, Malaysia, Mexico, Nigeria, Pakistan, Philippines, Portugal, Taiwan, Thailand, Turkey, Venezuela and Zimbabwe.
52 Bilson et al. (2001) use this term to describe factor series that consist of expected and unexpected components. In other words, “pure” innovations are not used. The use of expected and unexpected components of factors is common place in recent studies of the return generating process (see Sadorsky, 2001; Sadorsky &
Henriques, 2001). This represents a relaxation of the assumptions of the APT framework.
For the restricted five-factor model, results indicate that returns on ten markets are positively and significantly influenced by returns on the MSCI World Index whereas the exchange rate is statistically significant for eleven markets with a predominantly negative impact. The money supply is statistically significant for five markets, with the relationship between returns and changes in the money supply being predominantly positive. Inflation and real activity significantly impact returns on two individual markets, namely Mexico and Portugal.
Reported F-statistics indicate that the factors in the model are jointly significant for ten out of the twenty markets and the average R is 0.136 ranging between -0.00 and 0.38 for 2 Columbia and Indonesia respectively. A number of important inferences arise from these results. Firstly, the same risk factors do not explain returns across markets suggesting that there may be risk factors that are relevant to specific markets. In other words, the structure of the return generating process differs across markets. Secondly, as the factors are jointly significant in ten out of the twenty markets, a multifactor model may not be applicable to all markets. Finally, the spread of explanatory power also suggests that not all emerging stock markets can be adequately described by a multifactor return generating process derived within the APT framework. These results suggest that the identification of the return generating process is not a straightforward task. Such a proposition is consistent with the findings of Poon and Taylor (1991) who suggest that factors that explain returns are likely to vary across markets. Furthermore, these findings highlight a potential limitation of the multifactor APT framework; it may impossible to fully describe the return generating process relying only upon macroeconomic factors (Van Rensburg, 1996; Bilson et al., 2001).
The unrestricted model is estimated next. Results indicate that the MSCI World Index loses much of its significance; returns for three markets are significantly related to returns on the index. However, returns are significantly related to returns on regional indices in twelve out of twenty instances suggesting that regional risk factors are more important than international factors. The exchange rate is statistically significant for fourteen markets – a finding re-affirming this factor’s importance. Money supply and real activity are significant for two markets, and the inflation, country risk and the trade factors are significant for three markets.
Interest rates have a statistically significant impact upon returns in four markets. Notably, the price-to-earnings ratio and dividend yield are statistically significant for returns in sixteen and ten markets respectively suggesting that factors constructed out of the same market are better at explaining returns relative to macroeconomic factors used in place of unidentified APT
risk factors. This suggests that the APT framework should be extended to consider equity related aggregates. On average, the unrestricted model explains 60 percent of the variation in returns. Bilson et al. (2001) conclude by stating that within a multifactor setting, a larger set of factors improves the fit of the model.
Like Kandir’s (2008) approach, Bilson et al.’s (2001) approach strongly reflects the influence of the APT framework. However, the sporadic significance of certain factors across markets and inconsistent explanatory power emphasizes the argument that the same risk factors are not applicable across markets and that not all markets can be described by an APT framework derived model. These findings – extensive in scope - suggest that there are possible limitations and difficulties in applying the APT framework to developing markets.
This is especially pertinent given that the South African stock market can be classified as a developing market. Moreover and in light of the limitations of Van Rensburg’s (1996, 2000) analysis, it remains to be seen whether the APT framework can be applied and generalized to an extended number of South African return series. If the APT framework can be applied in an extensive investigation of the behaviour of South African stock returns, only then can it be said that the framework is applicable within the South African context.