La Porta’s 1997 Analysis Revisited with Social Indicators
In this article, we have followed a similar pattern for the presentation of data to that used in La Porta et al. (
1997a). In Table II of that paper (p. 1138), they listed 49 countries, grouped by legal origin, and reported empirical measures of financial and legal variables with means calculated for each legal origin. Our equivalent to Table II in La Porta et al. (
1997a) is Table
2; it differs from the La Porta et al. Table II in three ways. First, we have added a number of social indicator variables. The social indicators comprise: the under 5 child mortality rate (U5MR 01-04); two separate measures of income inequality (R10/P10—based on the richest and poorest deciles, and the Gini index—based on income distribution across the whole population); the log of the prison population (Log Pr Pop) and the proportion of women in the lower, or a single, house of legislators (% Women MPs). These variables are defined in more detail in Table
1a.
Table 1
Description of the social indicators (a) and summarised description of the variables reported in Table 1 of La Porta et al. (
1997a) (b)
(a) |
U5MR 01-04 | Mean under 5 child mortality rate for the years 2001–2004. Source UNICEF’s State of the World’s Children Reports 2003–2006. |
R10/P10 | The ratio of the income or expenditure share of the richest 10% of a population to that of the poorest 10%. Source UN Human Development Report (HDR) 2007–2008. |
Gini index | Gini coefficient of income inequality. Source UN HDR 2007–2008. This measure, unlike the R10/P10 ratio, is based on income levels for an entire population. |
Log Pr Pop | Log of prison population as at January 2007. Source HDR 2007–2008. |
% Women MPs | Percentage of Women in lower or single house of legislators as at 31 May 2007. Source HDR 2007–2008. |
(b) |
Ext cap/GNP | The ratio of the stock market capitalisation held by minorities to gross national product in 1994. |
Domestic firms/pop | Ratio of the number of domestic firms listed in a given country to its population (in millions) in 1994. Source: emerging market fact book and world development report 1996. |
IPOs/pop | Ratio of the number of initial public offerings of equity in a given country to its population (in millions) for the period July 1995 to June 1996. |
AntiDir Rights | An index aggregating shareholder rights. |
Second, we have restricted the number of countries investigated. The 49 countries considered by La Porta et al. span a very wide range of social and economic development. Had we used the same list of countries, any associations that may exist between social indicators and legal origins may well have been masked by the very large differences attributable to other factors. Such masking would be especially relevant to health indicators which show gross discrepancies between rich and poor countries lying on opposite sides of the “epidemiological transition”.
9
,
10 Our choice of countries is, therefore, a subset of the 49 investigated by La Porta et al. and is based on the method adopted by Collison et al. (
2007) which investigated child mortality in wealthy nations. In order to compare rich countries on the same side of the epidemiological divide, Collison et al. (
2007) restricted their examination to the 24 wealthiest OECD countries. Of those 24, only 22 are included in the current analysis since two, Iceland and Luxembourg, were not included in the La Porta et al. study. The restriction in the number of countries has not altered the relative outcomes originally recorded by La Porta et al. Their aggregated ranking of countries, when grouped by legal origin, persists when their data is restated for the smaller group of developed OECD countries. Both sets of aggregated figures, for the larger and smaller groups, are shown in bold in Table
2.
Table 2
External capital markets and social indicators
Australia | 0.49 | 63.55 | – | 4 | 6 | 12.5 | 35.2 | 2.10 | 24.7 |
Canada | 0.39 | 40.86 | 4.93 | 4 | 6.5 | 9.4 | 32.6 | 2.03 | 20.8 |
Ireland | 0.27 | 20 | 0.75 | 3 | 6 | 9.4 | 34.3 | 1.86 | 13.3 |
New Zealand | 0.28 | 69 | 0.66 | 4 | 6 | 12.5 | 36.2 | 2.27 | 32.2 |
UK | 0.49 | 35.68 | 2.01 | 4 | 6.5 | 13.8 | 36.0 | 2.09 | 19.7 |
USA | 0.39 | 30.11 | 3.11 | 5 | 8 | 15.9 | 40.8 | 2.87 | 16.3 |
English origin avg |
0.50
|
43.2
|
2.29
|
4
|
6.5
|
12.3
|
35.9
|
2.20
|
21.2
|
La Porta et al. avga
|
0.60
|
35.45
|
2.23
|
3.39
| | | | | |
Belgium | 0.17 | 15.5 | 0.3 | 0 | 5.5 | 8.2 | 33.0 | 1.96 | 34.7 |
France | 0.23 | 8.05 | 0.17 | 2 | 5.5 | 9.1 | 32.7 | 1.93 | 12.2 |
Greece | 0.07 | 21.6 | 0.3 | 1 | 5 | 10.2 | 34.3 | 1.95 | 13 |
Italy | 0.08 | 3.91 | 0.31 | 0 | 5.25 | 11.6 | 36.0 | 2.02 | 17.3 |
Netherlands | 0.52 | 21.13 | 0.66 | 2 | 5.5 | 9.2 | 30.9 | 2.11 | 36.7 |
Portugal | 0.08 | 19.5 | 0.5 | 2 | 5.5 | 15.0 | 38.5 | 2.08 | 21.3 |
Spain | 0.17 | 9.71 | 0.07 | 2 | 5.25 | 10.3 | 34.7 | 2.16 | 36 |
French origin avg |
0.19
|
14.2
|
0.33
|
1.29
|
5.36
|
10.5
|
34.3
|
2.03
|
24.5
|
La Porta et al. avga
|
0.21
|
10.00
|
0.19
|
1.76
| | | | | |
Austria | 0.06 | 13.87 | 0.25 | 2 | 5 | 6.9 | 29.1 | 2.02 | 32.2 |
Germany | 0.13 | 5.14 | 0.08 | 1 | 5 | 6.9 | 28.3 | 1.98 | 31.6 |
Japan | 0.62 | 17.78 | 0.26 | 3 | 4.5 | 4.5 | 24.9 | 1.79 | 9.4 |
Korea, Rep of | 0.44 | 15.88 | 0.02 | 2 | 5.25 | 7.8 | 31.6 | 1.99 | 13.4 |
Switzerland | 0.62 | 33.85 | – | 1 | 5.5 | 9.0 | 33.7 | 1.92 | 25 |
German origin avg |
0.37
|
17.3
|
0.15
|
1.8
|
5.05
|
7.0
|
29.5
|
1.94
|
22.3
|
La Porta et al. avga
|
0.46
|
16.79
|
0.12
|
2.00
| | | | | |
Denmark | 0.21 | 50.4 | 1.8 | 3 | 4.25 | 8.1 | 24.7 | 1.89 | 36.9 |
Finland | 0.25 | 13 | 0.6 | 2 | 4.75 | 5.6 | 26.9 | 1.88 | 42 |
Norway | 0.22 | 33 | 4.5 | 3 | 4 | 6.1 | 25.8 | 1.82 | 37.9 |
Sweden | 0.51 | 12.66 | 1.66 | 2 | 3.25 | 6.2 | 25.0 | 1.91 | 47.3 |
Scandinavian avg |
0.30
|
27.26
|
2.14
|
2.5
|
4.06
|
6.5
|
25.6
|
1.87
|
41.0
|
La Porta et al. avga
|
0.30
|
27.26
|
2.14
|
2.5
| | | | | |
Third, we have reported in Table
2 only a subset of the indicators from La Porta et al. (
1997a). The indicators reported are those from the first four columns of Table II in La Porta et al., and they measure various proxies for shareholder protection and the vitality of equity markets. This has been done to aid clarity of the exposition. The definitions of these variables are reproduced, in summary form, in Table
1b. The rest of the financial/legal variables considered by La Porta et al. (
1997a) are reproduced in an
Annex to this article (in “Supplement to Table 1b” and “Supplement to Table 2”) and they are also included in statistical investigations which appear later in the article.
The basis of the choice of the social indicators listed in Table
1a, merits some explanation. Results reported in Collison et al. (
2007,
2010) highlighting the poor performance of the Anglo-American countries suggested the possibility of an underlying systemic relationship involving poor societal well-being which could be linked to income inequality. The epidemiological literature provides additional evidence to support such a proposition (see, for example, Wilkinson and Pickett
2009). Two measures of income inequality were chosen and are explained in more detail in Table
1. They are the widely used Gini coefficient which takes into account income levels across an entire population, and a second, more extreme, measure which is a ratio based only on the income received by the top and bottom population deciles. The child mortality and prison population variables were chosen as examples of indicators which previous research had shown to be related to income inequality (see, for example, Wilkinson and Pickett
2009). The percentage of women MPs was selected since it seemed to be a potential discriminator between common and civil law traditions; the former being identified with the preservation of established interests. It seemed plausible that such values could be manifested in various ways:
If the mechanics of a particular electoral system exclude to a large degree members of a particular ascriptive group (women or otherwise), then more often than not that is damning evidence that the system is excluding the
interests of that particular group from the structures of decision-making power….Indeed, the degree to which a system successfully includes women can indicate a propensity for the system to include other disenfranchised minorities. (Reynolds
1999, p. 549)
The statistical significance of the relationships between social indicators and legal origin and between social indicators and some of the specific measures related to investor protection used by La Porta et al. are examined in some detail below. But a number of striking patterns emerge from a visual inspection of the mean statistics for the social indicators chosen. Consistent with the results from Collison et al. (
2007,
2010), the common law (i.e. English origin) countries have the worst child mortality figures and the highest levels of income inequality. The common law countries also have the largest mean prison population in the OECD and this result too is consistent with findings reported by Wilkinson and Pickett (
2009) that “more unequal countries have higher rates of imprisonment than more equal countries” (p. 148).
The association of income inequality with a range of social ills which is now widely established in the epidemiological literature is reflected in the consistent ranking of the legal origin groups across the income inequality, child mortality and prison population indicators. In each case, the Scandinavian countries perform best, followed by the German group. The French group is consistently ranked third while the common law countries are consistently ranked at the bottom.
11
The final social indicator, percentage of women MPs, is of a different type to the others but, as discussed above may be considered as a proxy for the progression of the democratic impulse and so could be construed as having features in common with the other measures. This indicator again shows, we would argue, the worst performance being found amongst the common law countries (in aggregate); although admittedly, there appears to be little difference between the three non-Scandinavian groups. Close inspection reveals that the results are influenced by a few outliers. In the German group, the Asian nations have low figures which may be accounted for by differing cultural traditions. Were these to be removed, a rather different gradation of means would be apparent. However, the figures for France, Greece and Ireland are all also relatively low. A cultural/historical examination of possible explanations for these figures goes beyond the scope of this article. But we would venture that, prima facie, this evidence, taken together with that pertaining to income inequality, is consistent with the position advanced above by Reynolds regarding the distribution of power and influence in society.
The next part of this section presents a more rigorous statistical examination of the data in Table
2 (and the rest of the La Porta et al. data which appears in the
Annex to this article) but it already appears that, compared to the civil law countries, the common-law tradition is associated with greater inequality and, possibly, with a relatively conservative approach to social development.
Statistical Analysis
The empirical analysis in this section of the article has a number of parts. Initially, the social indicator variables (the under 5 child mortality rate, the two measures of income inequality—R10/P10 and the Gini index, the log of the prison population and % women MPs) are examined for each of the legal-tradition categories which La Porta et al. derive. Specifically, the mean (median) of each of the social indicators is calculated for all four legal tradition groupings of countries and a test of the null hypothesis that the means (medians) were equal is conducted. The second empirical component of the current article examines the relationships that exist between the various proxies for investor protection which La Porta et al. employ when grouping countries and the social indicators examined in the current investigation. In particular, correlation analysis is used to study the sign and size of any relationships that may be present. The third empirical part of the analysis distils the information in the ten investor protection proxy variables, used by La Porta et al., into a number of principal components and regresses these components on the social indicators for the developed countries considered in the current study. In this way, a comprehensive investigation is undertaken to determine whether certain investor protection proxy variables and some legal tradition groupings of countries are associated with better indicators of social health and development such as under 5 child mortality, measures of income inequality, the size of the prison population or the representation of women among elected members of a country’s parliament.
The initial investigation focused on whether the five social indicator variables being considered varied across the four groupings of countries from La Porta et al. based on legal traditions. The results from this analysis are shown in Table
3. The top half of this table reports the findings from an analysis of means while the bottom half documents the results for an investigation of the median values for each of the social indicator variables; the median analysis is reported because some of the descriptive statistics in Table
2 suggested that the variables might not be normally distributed. In the top half of the table, the mean value of each social indicator together with its standard deviation is provided for all four legal traditions. An
F statistic and its
p value are then reported for a test of the null hypothesis that these means were equal. In the bottom half of the article, median values and their associated
Z statistics are provided for each of the four groupings of countries and an
H statistic together with its
p value are shown for the null hypothesis that these median values were equal.
Table 3
An analysis of the social indicator variables according to a country’s legal origin
English | 6.500 | 0.775 | 12.250 | 2.534 | 35.850 | 2.758 | 2.203 | 0.352 | 21.167 | 6.664 |
French | 5.357 | 0.197 | 10.514 | 2.256 | 34.300 | 2.466 | 2.030 | 0.089 | 24.457 | 11.036 |
German | 5.050 | 0.371 | 7.020 | 1.651 | 29.520 | 3.347 | 1.939 | 0.090 | 22.320 | 10.457 |
Scandinavian | 4.063 | 0.625 | 6.500 | 1.098 | 25.600 | 0.983 | 1.874 | 0.040 | 41.025 | 4.730 |
F statistic | 18.18 | 9.28 | 15.73 | 2.72 | 4.64 |
p value | 0.00 | 0.01 | 0.00 | 0.08 | 0.01 |
English | 6.250 | 3.540 | 12.500 | 2.730 | 35.600 | 2.430 | 2.097 | 1.990 | 20.250 | −1.220 |
French | 5.500 | 0.180 | 10.200 | 1.450 | 34.300 | 1.450 | 2.017 | 1.020 | 21.300 | −0.530 |
German | 5.000 | −1.210 | 6.900 | −2.230 | 29.100 | −1.610 | 1.978 | −0.980 | 25.000 | −0.940 |
Scandinavian | 4.125 | −2.980 | 6.150 | −2.470 | 25.400 | −2.810 | 1.881 | −2.470 | 39.950 | 3.060 |
H statistic | 17.86 | 15.70 | 14.19 | 9.32 | 9.64 |
p value | 0.00 | 0.00 | 0.00 | 0.25 | 0.02 |
An analysis of Table
3 reveals that sizeable differences exist in the measures of social health across the four groupings of countries based on La Porta et al.’s classification scheme. In particular, the mean level of under 5 child mortality in countries where the legal tradition has an English common law origin (mean = 6.500) is 60% higher than in Scandinavian countries (mean = 4.063). A similar picture emerges from the other four social indicator variables considered. Specifically, countries where the legal system is based on an English common law tend to have the greatest income inequality (according to both the R10/P10 and Gini index variables), the highest average prison populations and smallest percentage of women MPs. The Scandinavian countries perform best. In between, the countries where the legal tradition is based on French Law have the second highest (i.e. second worst) indicators across the first four variables, while those where the legal origin is German in character are ranked third.
12 In some of the subsequent analysis, the four legal origins are labelled Legal Origin (LO) 1–4 according to their order in Table
3 and, for the purposes of the correlation analysis in Table
4, they are ranked in this order as the Legal Origin variable.
Table 4
Correlation analysis
Legal origin | 1.000 | | | | | | | | | | |
ExCap/GNP | −0.1.86 | 1.000 | | | | | | | | | |
AntiDir | −0.423* | 0.522* | 1.000 | | | | | | | | |
Firms/Pop | −0.364* | 0.425* | 0.671* | 1.000 | | | | | | | |
IPOs/Pop | −0.222 | 0.413* | 0.656* | 0.688* | 1.000 | | | | | | |
CredR | 0.238 | −0.030 | 0.016 | 0.004 | −0.126 | 1.000 | | | | | |
Debt/GNP | −0.168 | 0.481* | 0.311 | −0.032 | −0.186 | 0.226 | 1.000 | | | | |
GDP growth | −0.148 | −0.129 | 0.162 | −0.164 | −0.214 | −0.139 | −0.011 | 1.000 | | | |
Log GNP | −0.125 | 0.336 | 0.040 | −0.307 | 0.181 | 0.050 | 0.557* | 0.186 | 1.000 | | |
1s1vote | 0.162 | 0.063 | −0.119 | −0.031 | −0.413* | 0.066 | −0.023 | 0.313 | 0.073 | 1.000 | |
Rule of law | 0.096 | 0.260 | 0.208 | 0.273 | 0.441* | −0.096 | 0.216 | −0.382* | −0.056 | −0.503* | 1.000 |
U5MR 01-04 | −0.912* | 0.349 | 0.418* | 0..416* | 0.219 | −0.266 | 0.248 | 0.056 | 0.179 | −0.285 | 0.065 |
R10/P10 | −0.838* | 0.036 | 0.293 | 0.350 | 0.206 | −0.204 | −0.003 | 0.009 | 0.035 | −0.240 | −0.207 |
Gini index | −0.801* | −0.007 | 0.146 | 0.199 | −0.029 | −0.243 | 0.037 | 0.044 | 0.006 | −0.188 | −0.200 |
Log Pr Pop | −0.649* | 0.049 | 0.248 | 0.199 | 0.019 | 0.130 | 0.336 | −0.027 | 0.214 | −0.261 | 0.088 |
% Women MPs | 0.484* | −0.112 | −0.093 | 0.012 | 0.285 | 0.300 | −0.172 | −0.417* | −0.368* | −0.533* | 0.553* |
A more detailed inspection of Table
3 reveals that there is some variability within the country groupings for the social indicator variables being studied. In particular, some of the standard deviation figures were large. This seems to be especially the case for English common law countries where the standard deviation values were highest for three of the five social indicator variables being examined. For example, the standard deviation value of the log of prison population variable for English common law countries of 0.352 is nearly four times as large as the next highest standard deviation number. By contrast, Scandinavian countries tend to be much more homogenous in terms of the social indicator variables since the standard deviation values are smallest for four of the five measures being examined.
Despite this variability within groupings, the picture that emerges from Table
3 is that a very consistent pattern exists in terms of the rankings of the country groupings according to their social indicator variables. The
F statistics confirm that the mean values for each social indicator are not equal across the four country groupings. All of the
F statistics were large and statistically significant at the 10% levels; indeed, four of the
p values are less than the critical value of 0.05. This finding is confirmed by an analysis of the median values and their corresponding
H statistics. With the exception of “% Women MPs”, the rankings of country groupings based on median values are identical to those based on their mean counterparts. Further, the null hypothesis that the medians are equal across the four country groupings can be rejected for four of the five social indicator variables; the exception to this general finding related to the log of prison population where the
H statistic is only 9.32 and its
p value is 0.25.
The Spearman correlations
13 (a) among the investor protection and legal origin variables and (b) between the investor protection as well as legal origin variables and the social indicator measures are displayed in Table
4. Based on the results from Table
2, and the ranking of the Legal Origin variable based on Table
3, one would of course expect the correlation findings to confirm that a relationship exists between the legal origin of a country and its social indicators. However, this table goes further by examining whether a relationship exists between (i) the investor protection measures on which the legal origin grouping is based and (ii) the social indicator variables. Further, the table highlights whether there are correlations among the different investor protection measures which La Porta et al. employ or whether each one is capturing a different aspect of the legal origin grouping used by La Porta et al.
A visual inspection of Table
4 reveals that there is a strong negative association between: under 5 child mortality; income inequality; as well as the size of prison population; and the legal origin measure used: of course these associations are to be expected given the construction of the legal origin variable. In addition, the association between the % of women MPs and legal origin is positive, consistent with the figures showing that Scandinavian countries have a much larger representation of female elected representatives in their Parliament relative to their common law counterparts. When the investor protection variables were examined, however, relatively few of the correlations were statistically significant; in fact only six correlation values have
p values of less than 0.05: Anti Dir and U5MR 01-04, FirmsPop and U5MR 01-04, GDP Growth and % Women MPs, Log GNP and % Women MPs, 1s1vote and % Women MPs, Rule of Law and % Women MPs. The remaining 44 correlations in the bottom panel of Table
4 are not statistically different from zero at the 5% level.
In the top half of Table
4, there is some evidence that the investor protection variables are correlated with one another. Of the 45 correlation values calculated, 12 were statistically significant: ExCapGNP and AntiDir, ExCapGNP and FirmsPop, ExCapGNP and IPOsPop, ExCapGNP and Debt/GNP, AntiDir and FirmsPop, AntiDir and IPOsPop, FirmsPop and IPOsPop, 1s1vote and IPOsPop, Debt/GNP and Log GNP, GDP Growth and 1s1vote, GDP Growth and Rule of Law, 1s1vote and Rule of Law. Such a result is hardly surprising since many of the variables were constructed from a common component (e.g. GNP) while all were presumably selected by La Porta et al. because they helped to paint a picture about one issue (namely the protection of investor rights) within a country. All of these significant correlations had the expected signs. For example, it is not surprising that the correlation between ExCapGNP and Anti Dir is positive at 0.522 since one would expect the index value aggregating shareholders rights in a country to be high where the ratio of the capitalisation held by minority shareholders to GNP is high.
Since there is some evidence of a relationship among the investor protection variables from La Porta et al. studies, it was decided to use a statistical approach to take account of this correlation before examining the association between social indicators and the investor protection variables using regression analysis.
14
To examine the possible relationship between indicators of social performance and the various investor protection variables in the La Porta et al. studies, Principal Components Analysis (PCA) was employed to identify relevant factors from the pool of data under consideration. PCA is a method which significantly reduces the number of variables from p to a much smaller set of k derived orthogonal variables that retain most of the information in the original p variables. The k derived variables which maximise the variance accounted for in the original variables are called principal components (PCs). After applying this analysis to the data series of each of the developed countries being studied, the dominant PCs are then extracted and used as inputs into a regression analysis to seek to explain the social indicators included in the study. The use of PCA is appealing for a number of reasons. First, it allows a large number of theoretically important factors that may affect the social indicators to be considered and second, it can be used effectively in conjunction with multiple regression analysis by addressing the problems of multicollinearity; specifically, because the k derived variables are orthogonal to each other, multicollinearity should not be present.
Table
5 summarises the results from applying PCA to the investor protection variables considered in the La Porta et al. papers. In particular, the bottom part of Table
5 details the eigenvalues and proportions of variance explained by the PCs, while the top part of Table
5 summarises the factor loadings for the dominant PCs. The data in Table
5 clearly show that across all 22 countries examined, the bulk of the variability in the original ten investor protection variables can be explained by 4 PCs. For example, the variance, or eigenvalue, of the first PC is 3.027. It explains 30.3% of the total variance of the ten investor protection variables. The second PC has an eigenvalue of 2.291 and accounts for 22.9% of total variance of the 10 variables. The third and fourth PCs also have eigenvalues greater than 1 and account for 17.0 and 11.3%, respectively, of the variability in the investor protection measures across the different countries. The proportion of variance explained by the remaining six PCs is relatively low and their eigenvalues are all small.
Table 5
A principal component analysis of the La Porta et al. investor protection variables
ExCap/GNP | 0.326 | −0.438 | 0.002 | −0.012 | −0.034 | 0.272 | 0.715 | 0.228 | −0.112 | −0.225 |
AntiDir | 0.454 | −0.184 | 0.283 | −0.187 | −0.223 | −0.230 | 0.044 | −0.339 | −0.122 | 0.641 |
Firms/Pop | 0.369 | 0.051 | 0.489 | 0.217 | −0.284 | 0.045 | −0.320 | −0.166 | −0.164 | −0.579 |
IPOs/Pop | 0.401 | 0.065 | 0.246 | −0.402 | 0.525 | 0.147 | −0.194 | 0.247 | 0.466 | −0.023 |
CredR | 0.104 | −0.229 | 0.077 | 0.792 | 0.440 | 0.148 | −0.123 | −0.005 | 0.007 | 0.271 |
Debt/GNP | 0.233 | −0.391 | −0.401 | 0.152 | −0.282 | −0.451 | −0.189 | 0.127 | 0.504 | −0.140 |
GDP Growth | −0.257 | −0.409 | 0.274 | −0.158 | 0.389 | −0.593 | −0.040 | 0.119 | −0.342 | −0.179 |
Log GNP | 0.118 | −0.377 | −0.469 | −0.265 | 0.194 | 0.339 | −0.361 | −0.412 | −0.294 | −0.121 |
1s1vote | −0.288 | −0.409 | 0.249 | −0.104 | −0.362 | 0.372 | −0.362 | 0.474 | −0.016 | 0.227 |
Rule of Law | 0.403 | 0.292 | −0.311 | 0.032 | −0.002 | −0.136 | −0.180 | 0.566 | −0.521 | 0.107 |
Eigenvalue | 3.027 | 2.291 | 1.696 | 1.132 | 0.546 | 0.447 | 0.359 | 0.283 | 0.130 | 0.090 |
Proportion | 0.303 | 0.229 | 0.170 | 0.113 | 0.055 | 0.045 | 0.036 | 0.028 | 0.013 | 0.009 |
Cumulative | 0.303 | 0.532 | 0.701 | 0.815 | 0.869 | 0.914 | 0.950 | 0.978 | 0.991 | 1.000 |
The Kaiser criterion was used to select the PCs which should be used as inputs for the regression analysis. The criterion recommends that only those PCs with eigenvalues greater than or equal to 1, should be retained (Kaiser
1960). Jolliffe (
1972) has suggested a cut-off point of 0.7. However, in this instance, Jolliffe’s criterion results in the same number of components being retained as Kaiser’s criterion of the eigenvalue being greater than or equal to 1 (Dunteman
1994). Therefore, the adoption of these criteria led to the retention of four PCs, which we describe in more detail below. Together, these four PCs account for 81.5% of the variance in the investor protection variables. Therefore, the dimensionality of the dataset can be reduced from 10 to 4.
The values in the top half of Table
5 indicate the factor loadings of the PCs that are identified from the data. In particular, the top half of the table therefore highlights the variables that have large coefficients of either sign in each PC vector.
15 The first PC, which is shown in column 2, has high positive correlations with AntiDir, Rule of Law and IPOsPop and negative correlations with GDP Growth as well as 1s1vote. This PC primarily reflects strong shareholder rights and a vibrant new issue market; we have labelled it “Outsider Capitalism” in the current analysis. The second PC shows large negative co-efficients for ExCapGNP, GDP Growth and 1s1vote and can be interpreted as small stock market/low growth variable. We label this PC “Insider Capitalism” in the remainder of the article. The largest co-efficients for the third PC are positive for FirmsPop and negative for GDP Growth as well as Debt/GNP. This can be interpreted as a large stock market/low growth/low debt variable; as a result, we label this PC as the “Small Economy” variable. The fourth PC is mainly associated with strong “Creditor Rights”.
In the final part of the empirical analysis, the dominant PCs together with legal origin dummy variables are used as inputs to a regression analysis in order to explain the social indicator variables of the 22 developed countries included in this study.
16 Five regression models are considered. First, the under 5 child mortality figures of the sample countries are regressed on each of the four PCs as well as three dummy variables representing legal origin (LO2 (French), LO3 (German) and LO4 (Scandinavian)); a variable was not added for LO1 (English) as the regression equation would have been over-specified. Instead, the impact of Legal Origin 1 is accounted for in the constant term: all of the other co-efficients need to be interpreted relative to the level of under 5 child mortality rate (U5MR 01-04) in English common law countries.
17 Four similar regression equations were estimated for the other social indicator variables. These regression models took the form:
$$ {\text{SIsi }} = \beta_{0} + \beta_{ 1} {\text{PC}}_{{ 1 {\text{i}}}} + \beta_{ 2} {\text{PC}}_{{ 2 {\text{i}}}} + \beta_{ 3} {\text{PC}}_{{ 3 {\text{i}}}} + \beta_{ 4} {\text{PC}}_{{ 4 {\text{i}}}} + \beta_{ 5} {\text{LO2}} + \beta_{ 6} {\text{LO3 }} + \beta_{ 7} {\text{LO4 }} + \varepsilon_{\text{i}} $$
(1)
where SIsi is the social indicator s for country i (s = U5MR 01-04, R10/P10, Gini index, Log Pr Pop and % Women MPs), PC
i is principal component for country i, LO is the Legal Origin dummy variable for French (LO2), German (LO3) and Scandinavian (LO4) legal traditions. Finally, ε
i is a random error term.
Table
6 reports the results from estimating Eq.
1.
18 In particular, the table details the co-efficient of each PC and Legal Origin dummy variable, with their corresponding
p values. The adjusted
R²s for the five regressions are also shown. An inspection of Table
6 suggests that a significant relationship exists between some of the social indicator measures and the PC as well as a number of the legal origin dummy variables. The strongest and most significant associations are between under 5 child mortality as well as income inequality and legal origin variables. For example, the co-efficients for the legal origin variables are negative for the U5MR 01-04 equation suggesting that under 5 child mortality is lower in countries which do not have an English common law tradition; for those countries with a German or Scandinavian legal tradition, the co-efficients on the U5MR 01-04 variable are statistically significant at the 5% level. A similar picture emerges for the Gini index equation where Legal Origin 3 and Legal Origin 4 dummy variables have co-efficients of −9.068 and −10.810 with
p values of 0.020 and 0.000, respectively. For the R10/P10 (% Women MPs) variables, only the co-efficient for the Legal Origin 4 countries is significantly negative (positive) at the 5% level.
Table 6
Regression results
Constant | 6.468 | 0.000 | 11.608 | 0.000 | 36.869 | 0.000 | 2.134 | 0.000 | 19.319 | 0.002 |
PC1 | 0.069 | 0.606 | 0.235 | 0.683 | −0.494 | 0.470 | 0.0489 | 0.375 | 1.964 | 0.268 |
PC2 | −0.027 | 0.818 | −0.256 | 0.611 | −0.115 | 0.846 | −0.009 | 0.845 | 2.508 | 0.114 |
PC3 | −0.049 | 0.710 | 0.209 | 0.715 | −0.024 | 0.971 | −0.010 | 0.858 | −1.713 | 0.498 |
PC4 | −0.154 | 0.252 | 0.277 | 0.623 | −0.008 | 0.991 | −0.020 | 0.704 | 4.244 | 0.025 |
LO2 | −1.065 | 0.092 | −0.533 | 0.835 | −3.083 | 0.317 | −0.051 | 0.835 | 6.015 | 0.440 |
LO3 | −1.452 | 0.049 | −5.335 | 0.087 | −9.068 | 0.020 | −0.152 | 0.585 | 5.670 | 0.522 |
LO4 | −2.397 | 0.000 | −5.100 | 0.016 | −10.81 | 0.000 | −0.278 | 0.134 | 18.261 | 0.006 |
R
2
| 0.71 | | 0.43 | | 0.65 | | 0.05 | | 0.63 | |
An inspection of the co-efficients on the PC variables indicates that only one significant value is observed. The Creditor Rights variable (PC4) is positively associated with the percentage of women MPs in a country (co-efficient = 4.244, p value = 0.025). However, this may simply reflect the fact that in Scandinavian countries, creditor rights are protected to a greater extent and a larger percentage of MPs are women. None of the other PC measures constructed from the investor protection variables employed in La Porta et al. can significantly explain the social indicators of the countries being studied.
Finally, it is worth pointing out that three of the regression equations have relatively high explanatory power. Specifically, for the U5MR 01-04, Gini index and % Women MPs, the R
2 values are 0.71, 0.65 and 0.63, respectively. The only equation with a very low level of explanatory power is where Log Pr Pop is the dependent variable; in this instance, the R
2 is only 0.05 and none of the co-efficient values are statistically different from zero.