Portal de Periódicos da CAPES

The Portuguese equity risk premium: what we know and what we don’t know

2005; Chapman and Hall London; Volume: 15; Issue: 7 Linguagem: Inglês

10.1080/09603100500038799

1466-4305

Rui Alpalhão, Paulo Alves,

Financial Risk and Volatility Modeling

Abstract Estimates of appropriate equity risk premiums are abundant in finance textbooks. Unfortunately, these estimates are ill suited to small and data scarce markets such as the Portuguese. The literature is reviewed to select techniques to overcome this difficulty, and estimates of equity risk premiums suited to the Portuguese market produced. Historical equity premiums are computed and the study finds what is believed to be a better understanding of the subject with the help of the Godfrey–Espinosa approach and of implied risk premiums. The Godfrey–Espinosa model is applied to a number of other European markets, and it is concluded that the Portuguese market implies a higher exposure to risk, namely when compared to other Euronext member markets. It is concluded that the valuation of Portuguese equities should carry a higher risk premium than the ones generally suggested in finance textbooks, and that the merger of the Lisbon Stock Exchange with Euronext should lead to a reduction in the appropriate risk premiums for Portuguese blue chips. Notes Siegel (Citation1992) puts forward significant variations in the level of the equity risk premium during 1987 as one (out of two) possible explanation for the behavior of stock prices around the 19 October 1987 crash, when the S&P 500 index fell by 20.5%, the greatest single-day decline in history. To discuss whether it does or not is beyond the scope of this paper; nevertheless, the view that stock prices do not follow a random walk is supported by several authors, namely Lo and MacKinlay (Citation1988). Other authors, such as Fama and French (Citation1988) and Finnerty and Leistikow (Citation1993), find that, although stock prices follow random walks, these are not stationary. For simplicity, an arithmetic average is taken. For a discussion of the pros and cons of arithmetic and geometric averages, Cooper (1996). In the following sections, geometric averages were chosen in part due to the reasons stated in the section on asset pricing models. Damodaran (Citation2003) computes standard errors of risk premium estimates based on a 20% annual standard deviation in stock prices of 2.8% with a 50-year estimation period, against 6.3% taking a 10-year period. Portugal is, in this study, included among the developed markets. Since part of the factors behind the added volatility are, presumably, also behind risk free rates differentials that also affects cost of capital computations through domestic risk free rates, and following Erb et al. (Citation1995), that report that 40% of share volatility is explainable through credit risk differences, Godfrey and Espinosa compute risk premium using 60% of adjusted betas for emerging markets. This will not be done so in the Portuguese case. For a discussion of the implications of the alternative view of international capital markets integration, Eijgenhuijsen and Buckley (1999). For another approach to the segmentation angle, see the Brealey et al. (1999) ‘locally financed firms’. These are, of course, pre-EMU results. Nowadays, one can hypothesize that they would persist only on a Euroland basis. The interested reader is referred to Keck et al. (1998). This view is found to be too strict, as British data looks perfectly adequate for the computation of historical risk premiums. Daily price (expressed in percentage) of a synthetic issue of a bond whose present value is calculated with a YTM resulting from the three year average rate weighted by the duration of the member bonds (all listed fixed rate government bonds). Annualized monthly averages of daily close data. In Godfrey and Espinosa (1996), Portugal's adjusted beta is reported to be roughly two. This difference is attributed to the different calculation periods. In this case only in the later period (after 1950) of their extremely long time series (1872–2000).