Table 6 provides descriptive statistics for the sample. Per share values are in nominal
currency of the countries, so comparisons are difficult. However, we can compare ROE
(the ratio of EPS to average book value) across countries and the differences are
substantial. Median ROE ranges from 6 percent (over 8 years) in Korea to 14 percent (over 5 and 2 years) in Indonesia and the Philippines. For comparison, King and Langli
(1998) find ROE over the 1980s and 1990s to be about 6 percent in Germany, 10 percent
in Norway, and 13 percent in the UK. In the US, this measure has averaged around 13 percent over the last 20 years. Prior research has speculated that such differences across
countries may reflect differences in conservatism of accounting methods. As noted above,
for example, Taiwan does not record either goodwill asset revaluations (most conservative)
while both are recognized in the Philippines (least conservative).
Table 7 reports the pair-wise correlation between stock price and accounting variables
for all countries. For all countries except Korea the rank (Spearman) correlations are
greater than the product±moment (Pearson) correlations. However, the patterns and
significance of the parametric (Spearman) and non-parametric (Pearson) correlations are
similar. Stock prices are strongly correlated with BVPS and with EPS for all countries. The
pair-wise correlations between price and EPS are approximately the same as between price
and BVPS. The correlation (Pearson) between BVPS and EPS is high in Thailand (.81),
the Philippines (.78), and Korea (.74), and relatively low for Taiwan (.24). If EPS is used
as a proxy for residual earnings, the high correlation between BVPS and EPS (they are
related by size) may make it difficult to partition value relevance between book value and
earnings. However, the pair-wise correlations between BVPS and REPS are far smaller,
significant in some cases, insignificant in others, and sometimes negative. This is
anticipated since there is no reason to expect residual (unexpected) per share earnings
to be related to book value. One important reason for using REPS in the empirical analysis
is to avoid the high correlation between BVPS and EPS.