1 Legal Family Tree

Figure 1.1 shows the relative positions and percentages of disagreement of the 136 jurisdictions in a property-law family tree. The scale at the top shows the Gower distance, which is simply the percentage of disagreement. Among the 279 variables, if a pair of countries has the same values (both 0 or both 1) in 200 variables and different values in the remaining 79 variables, the percentage of agreement is 72% (=200/279) and the percentage of disagreement is 28% (=79/279). The Gower distance is thus 0.28. If two countries both copy, say, the Portugal Civil Code, their Gower distance is 0.

Figure 1.1 quantitatively summarizes dissimilarities of the 136 studied jurisdictions, arranged vertically. At the bottom (the left side) of the dendrogram, each jurisdiction is considered its own cluster. The horizontal axis (the scale at the top) of the dendrogram indicates the Gower dissimilar coefficients. The “height” of the 135 nodes (or joining points), visualized as vertical line in Figure 1.1, thus represents the Gower dissimilar coefficients when two clusters merge. Figure 1.1 reports average-linkage clustering, under which the intercluster distance equals the average distance between all intercluster pairs of jurisdictions. As one moves up the tree (from left to right), the groupings of merged jurisdictions become more dissimilar. Groupings continue until, at the far right of the dendrogram, all observations appear in a single supercluster. Figure 1.1 shows that the maximal dissimilarity between any dyad of jurisdictions in terms of property law is no more than 35%.

My quantitative approach can produce any number of legal families. This chapter arbitrarily picks ten groups for expositional purposes. Ten, or any number of, groups can be identified by moving a vertical line from the right-hand side and moving leftward until ten horizontal lines appear to the right of the vertical line. Below, the ten groups are also labeled according to conventional wisdom to make it easier to refer to.Footnote ³ The algorithm in no way knows or reveals that members in a certain group have been heavily influenced by, say, Germany or England.

In Figure 1.1, from top to bottom, the first cluster is the English group (numerically labeled as Group 3 in the data released along with this book). The English group contains all the well-known common-law jurisdictions, plus Scotland and Israel, which are less similar with the core common-law jurisdictions.

The second cluster is China alone (Group 10). In my previous work (Chang et al. Reference Chang, Garoupa and Wells2021), China’s 2020 Civil Code was found to be similar to Mongolia’s 1995 Civil Code. The similarity greatly attenuates after this book used the coding of Mongolia’s 2002 Civil Code. The following chapters demonstrate that China’s 2020 Civil Code is often idiosyncratic in content (though German and Taiwanese in structure) and stingy in spelling out many key property doctrines. Russian (USSR) legal scholarship was influential in China before the 1980s, though many Chinese scholars would be surprised that China is still closer to the Socialist family that contains many former USSR republics than to the German family. As conjectured in the following text, the similar lack of many standard property doctrines may explain the resemblance of the Chinese property law to the Russian property law.

The third cluster is the Socialist group (Group 1). Civil codes in several former USSR republics have been influenced by the Russian Civil Code (Burnham et al. Reference Burnham, Maggs and Danilenko2009: 353–354). Notable members also include the four Scandinavian countries: Denmark, Norway, Finland, and Sweden. They are closer to one another than to other countries. Note that many jurisdictions are grouped together in this cluster also because many variables take the value of 0, as the property doctrines do not exist. Inevitably, the design of my codebook is more affected by Western European, East Asian, and American laws, with which I am more familiar. Those doctrines may not exist in this group perhaps because (1) there has not been such disputes (e.g., the Scandinavian countries), (2) the regulatory density of property law is low (e.g., North Korea), or (3) disputes arising under those doctrines may be dealt with in completely different ways.

The fourth cluster, the biggest, is the French group (Group 2). France and its former colonies constitute a subcluster. Spain and many of the Spanish-speaking Latin American countries form another subcluster. The Code Napoleon was forced upon Spain in the early nineteenth century, but Spain broke away from the French code later on. The inspiration of property law in Latin American countries is complicated and beyond the scope of this chapter. However, it is worth noting that Spanish-American colonies achieved independence before Spain adopted a French-oriented code; thus, they received the French civil law through voluntary transplant (Berkowitz et al. Reference Berkowitz, Pistor and Richard2003b; Oto-Peralías and Romero-Ávila Reference Oto-Peralías and Romero-Ávila2014: 576). Here it is shown that Latin American countries are close to one another, and closer to Spain, than to France. Besides, several jurisdictions from the Middle East and North Africa form another subcluster. These countries emulate French law without having any colonial history (Klerman et al. Reference Klerman, Mahoney, Spamann and Weinstein2011), but do not buy the French property law wholesale. Notably, members of this group are the non-European part of the Ottoman Empire. Before its dissolution, the Ottoman Empire imported French commercial law and public law, though its private law is local and Muslim (Xu Reference Xu2007: 303–304). This history partly explains its French inclination. Indeed, reading civil codes of, for example, Iraq and the United Arab Emirates, a private law scholar familiar with French law will certainly smell champagne and cheese.

I call the fifth and sixth clusters “semi-French” (Group 9) and “quasi-French” (Group 5), respectively, as they are part of the larger supercluster influenced by the Napoleonic Code. Portugal and its former colonies are here, as are a number of countries that have learned from non-French sources. Quebec is a famous mixed jurisdiction, also influenced by English law. Brazil has borrowed from French, German, and English laws (Pargendler Reference Pargendler2012a: 810–812). The Netherlands is between (and beyond) the French and German systems.

The seventh cluster has a sole member, another famous mixed jurisdiction – South Africa (Group 8), which has both a Roman-Dutch tradition and an English legal heritage.

The final three clusters are all affected by German law. The Japanese cluster contains three other (South-)East Asian jurisdictions: Taiwan, South Korea, and Cambodia. The affiliation of Japan is more complicated than commonly thought. The Japanese Civil Code of 1898, in its drafting stage, was heavily influenced by the French Civil Code. While Japanese private law scholarship later turned to German jurisprudence for inspiration – note that the German Civil Code did not go into effect until two years later, in 1900 – the Japanese Civil Code itself retains the French rules (Note Reference North1906: 73; Belli Reference Belli1959: 137). When South Korea’s civil code went into effect in 1960, it was indirectly affected by the German code through the Japanese code and Japanese scholarship (Kwon Reference Kwon2013: 114). The Taiwanese Civil Code, on the other hand, was enacted in 1930, when the Nationalist Party ruled mainland China. The Taiwanese Civil Code’s Book of Things is a hybrid of traditional Chinese law, Swiss law, and German law. Japanese scholars played an instrumental role in the drafting of the 1930 code (Chang et al. Reference Chang, Chen and Chieh Wu2022: 23). When the Book of Things in the Taiwanese Civil Code was reformed between 2007 and 2010, the Japanese influence, partly due to the colonial experience in 1895–1945, is obvious (Chang Reference Chang, Chang, Shen and Wang2016a: 228). The 2007 Cambodia Civil Code was drafted with assistance from Japanese scholars (Upham Reference Upham2018: 109). Had I coded the 2020 Laos Civil Code, it would be likely to be grouped here as well, as Japanese scholars again played an instrumental role in the drafting stage of Laos’ code.

The penultimate cluster is Quasi-German (Group 6). The members are the Eastern European neighbors of Germany. The Austria Civil Code predates the German Civil Code and has its own style.

The final cluster is the German group (Group 4). Members contain several European neighbors of Germany, including Turkey. Turkey was the heart and soul of the Ottoman Empire, but in 1926, drafters of its new civil code under the Kemal administration adopted the Swiss Civil Code as their model (Xu Reference Xu2007: 314).

The big-picture, takeaway point is that common versus civil law divide is not the most salient dichotomy in property law. Chang et al. (Reference Chang, Garoupa and Wells2021), using 170 binary variables (most of which are included here), find that a French influence versus the lack thereof drives the classification. Here, the French supercluster is still salient. The (Quasi-/Semi-)German, Socialist, and English groups are parts of the large non-French supercluster in Chang et al. (Reference Chang, Garoupa and Wells2021). Here, with 279 binary variables, the German supercluster is deemed to be closer to the French supercluster. Still, standing on the other side of the German-French supercluster are civil-law Socialist countries spearheaded by Russia, Scandinavian countries, China, and English common-law countries. Civil law is plural, and common law is heterogeneous.

Figure 1.2 shows the ten groups in a world map. The three broadly speaking French clusters are in different shades of orange and yellow, whereas the three broadly speaking German clusters are in different shades of green (see also Table 1.1).

Figure 1.2 Legal families in world map

2 Distinctive Features

What are the key variables that drive the legal families? A supervised machine-learning method called “sparse linear discriminant analysis” (Gaynanova et al. Reference Gaynanova, Booth and Wells2016) produces Figure 1.3. The rows are the key variables that contribute to the classification of legal families. Numbers in column heads represent the numeric group labels. The large positive numbers in the cells mean that variables taking the value of 1 contribute greatly to a certain grouping. The large negative numbers in the cells mean that when variables take the value of 1, it reduces the likelihood of being classified into a certain grouping. The numeric scale on the right-hand side shows the minimal and maximal coefficients contained in the cell. Blue numbers are positive, whereas red numbers are negative. The larger the number, the darker the color. The absolute values in the cell do not mean anything; what matters is the relative size of the values.

Figure 1.3 Key factors driving the grouping

According to Figure 1.3, the English group (Group 3) is featured by the voidable rule and entrustment rule in the good-faith purchase doctrine (Chapter 10). China (Group 10) has many one-of-a-kind rules, and the algorithm picks out the co-ownership quorum rule for sales – China requires supermajority votes rather than unanimity.

The French group (Group 2) and the Quasi-French group (Group 5) have the right of use, disallow right of use and right of habitation to be inherited, and limit by mandatory or default rules the utilization of immovables under the right of habitation for personal needs (Chapter 3). The Semi-French Group (Group 9) is featured by the two variables regarding the right of habitation. All three groups tend to have the France §554 rule (Chapter 13).

South Africa (Group 8) is featured by the non-causa principle and attornment as constructive delivery of movables (Chapter 4), which are adopted in mostly jurisdictions affected by Germany. On the other hand, South Africa requires possessors to have animus possidendi, and only some English group members require the same. This demonstrates why South Africa is idiosyncratic.

As for the German supercluster, the Japanese group (Group 7) recognizes divided superficies (Chapter 3). In addition, like the two other groups influenced by Germany, under the specificatio doctrine (Chapter 12), as compared to others, bad-faith improvers are inclined to be compensated. The Quasi-German group (Group 6) also recognizes attornment as constructive delivery. In addition to dimensions already mentioned, the German group (Group 4) also requires prompt protest in building encroachment disputes (Chapter 6) and registration of ownership (rather than title) in terms of registration-based acquisitive prescription (Chapter 5).

The algorithm suggests that the most salient feature in the Socialist group (Group 1) is the France §554 rule and the right of use. This is puzzling, as very few jurisdictions in this group have these two rules.