Lexicons of Adjudicated Denials

The contents of divorce holdings exhibited astonishingly little variation, regardless of whether a plaintiff made a domestic violence allegation. Whether the trial was held in Henan or Zhejiang likewise made little, if any, difference. Judges drew from an extremely limited pool of stock words and phrases that referred overwhelmingly to breakdownism and rarely to faultism. They rendered their holdings mechanically, in highly standardized and scripted boilerplate.

In both provincial samples of first-attempt divorce decisions, plaintiffs used the word for “mutual affection” (感情) in 83% of their legal complaints. In both Henan and Zhejiang, only the words “marry” and “divorce” appeared in more legal complaints. Judges used the word for “mutual affection” even more frequently: it appeared in 93–94% of their holdings in the two samples. Whereas plaintiffs used it to claim that mutual affection had broken down, judges used it to hold the opposite. Defendants who were unwilling to divorce also challenged plaintiffs’ claims that marital affection had broken down. All participants in divorce litigation spoke the language of breakdownism. As we saw in Chapter 7, plaintiffs also spoke the language of faultism in their efforts to establish the breakdown of mutual affection.

By contrast, rarely did judges refer to fault-based standards of wrongdoing. Judges used terms such as “violence” (暴力), “odious habit” (恶习), “fault” (过错), and “Article 46” (四十六) or “46,” the provision in the Marriage Law on civil damages for wrongdoing, in their holdings in only 4% of first-attempt decisions in each of the two samples.Footnote ⁴ On the rare occasions judges used the language of wrongdoing, they did so only to say that it was not tantamount to the breakdown of mutual affection. Meanwhile, judges grounded their decisions in political and ideological discourse by using words such as “harmonious” (和谐 and 和睦), “stability” (稳定), and “civilized” (文明) in 25% and 13% of holdings in my Henan and Zhejiang samples, respectively. These politically salient words are similarly represented in subsamples of holdings involving allegations of domestic violence.

Before scrutinizing judges’ language at a more granular level, we can draw a couple of preliminary conclusions from this simple exercise. First, even in cases that involved wrongdoing, judges were averse to applying fault-based legal standards. They tended to apply the breakdownism standard in a one-size-fits-all manner. Second, we can also see that, within this general pattern that applies to Henan and Zhejiang alike, ideological discourse was almost twice as prevalent in Henan as it was in Zhejiang. Judges used political slogans as grounds for denying divorce petitions. Variations of “for the sake of maintaining harmonious and civilized marital and family relations” (维护和睦文明的婚姻家庭关系), “for the benefit of marital and family stability and of social harmony” (为有利于婚姻家庭稳定和社会和谐), and “in order to maintain family harmony and social stability” (维护家庭和谐、社会稳定) are prevalent throughout holdings in both samples, but somewhat more so in holdings from Henan.

Figure 8.1 depicts word clouds of the top 50 most frequent words across the two provinces in two types of holdings: (1) adjudicated denials and (2) cases involving allegations of domestic violence.Footnote ⁵ If every word were unique, there would be 200 words across all four word clouds (50 per word cloud). In fact, there are only 73 unique words because there is so much redundancy between holdings for adjudicated denials of cases without domestic violence allegations (Panels A and B) and holdings for domestic violence cases (Panels C and D). Clearly, judges attached little importance to domestic violence allegations. To judges, there was nothing about cases involving domestic violence allegations that merited their special attention or consideration. Redundancy between holdings from Henan (Panels A and C) and Zhejiang (Panels B and D) also underscores the extent to which domestic violence was similarly unimportant to judges in both provinces.

Figure 8.1 Word clouds of top 50 most frequently used words in judges’ holdings

Note: Limited to first-attempt adjudications. Words were scaled according to their frequencies in holdings. Their placement locations were optimized according to their sizes. For this reason, although Chinese words and their English translations are scaled identically, they are not located in the same places in their respective word clouds. Every adjudicated first-attempt divorce decision in my samples is included, except granted divorce petitions that did not contain domestic violence allegations. Court decisions are not used in more than one word cloud; each court decision is used in only one word cloud.

Source: Author’s calculations from Henan and Zhejiang provincial high courts’ online decisions.

All four word clouds share a common set of 35 words. Panels A and C of Figure 8.1 each contains seven unique words found in no other word cloud; Panel B contains three unique words; and Panel D contains only two unique words. Word clouds from the same province share even more in common. Henan’s two word clouds share 41 words, Zhejiang’s 45 words. Because some domestic violence cases led to actual divorces (28% and 14% in Henan and Zhejiang, respectively), the word clouds for domestic violence cases contain unique words pertaining to granting divorces. With respect to Henan, Panel C contains nine such words (in italics) that do not appear in Panel A: granting (予以) the divorce petition on the basis of the Marriage Law’s provision (Article 32), stipulating (规定) that divorce should be granted when mediation (调解) fails or when a physical separation (分居) test is satisfied, and thus also ruling on the disposition of premarital (婚前) property (财产), custody (抚养) of a child (孩子), and child support payments (抚养费). Similarly, in Zhejiang, Panel D contains five such words (in italics) that do not appear in Panel B: after affirming (认定) that mutual affection has indeed broken down and granting (予以) the divorce, some claims (主张) such as child custody (抚养) could be dealt with (处理), but other claims (typically concerning property) could not be dealt with (不予处理) or must be dealt with through separate litigation (另案处理 or 另行处理).

Although the word clouds for domestic violence cases contain words associated with granted divorces, their tiny sizes reflect their relatively rare usage owing to high denial rates. Notably absent are words associated with wrongdoing in general and domestic violence in particular, a point to which I will return later.

For reference purposes, Table 8.1 contains all 73 unique words across both provinces and case types. The first page of the table contains the 35 words shared by all four word clouds.

To support their holdings, judges cited Article 32 of the Marriage Law in 94% and 98% of their adjudications in the Henan and Zhejiang samples, respectively. Indeed, this was the sole legal provision judges cited in 32% and 44% of all holdings in the Henan and Zhejiang samples, respectively. Article 32 contains 12 words in the word clouds, three of which are particularly prominent: “marital” (夫妻), “affection” (感情), and “breakdown” (破裂). These three words account for only 4% of all 73 unique words but for 21% of all word frequencies in the four word clouds; at least one of them appeared in 95–96% of holdings in each provincial sample of first-attempt divorce adjudications. At least one of these three words was in the vast majority of holdings to grant divorces (94% and 91% in the Henan and Zhejiang samples, respectively) and almost universally found in holdings to deny divorces (98% in each sample).

Article 32 contains an additional nine words in the word clouds: “has indeed” (确已), “family” (or “domestic,” 家庭), “demand” (which can also mean “request,” 要求), “bring forward” (提出), “divorce” (离婚), “litigation” (诉讼), “ought to” (应当), “mediation” (调解), and “separation” (分居). Elsewhere in the Marriage Law are 30 additional words in the word clouds, the most prominent of which are “both sides” (双方), “live” (or “living,” “life,” 生活), “marriage” (婚姻), “request” (or “petition,” 请求), “together” (or “joint,” “common” 共同), “marry” (结婚), and “postmarital” (婚后). Some of these words – such as “marital,” “affection,” “breakdown,” “marry,” and “family” – were more likely to be used in holdings to deny divorces. Others – such as “divorce,” “live,” and “together” – were more likely to be used in holdings to grant divorces. The words “live” and “together,” for example, appeared in holdings to grant divorces when they referred to joint property, joint debt, child custody, and future living expenses.

Although Article 32 also contains the term “domestic violence” (家庭暴力) as grounds for affirming the breakdown of mutual affection, the language of violence rarely appears in judges’ holdings. As we have seen, it is conspicuously absent even in cases involving allegations of domestic violence. Violence words are therefore also conspicuously absent in all the word clouds, even those constructed from holdings of cases involving allegations of domestic violence. The word “violence” (暴力) appears in only 2% of the holdings in each provincial sample of first-attempt divorce adjudications, and in only 5% and 8% of holdings in cases involving allegations of domestic violence in the Henan and Zhejiang samples, respectively. The contracted form of “domestic violence” (家暴) appears in fewer than 70 out all the nearly 150,000 holdings in both samples of first-instance divorce adjudications. We also saw in Chapter 7 that judges’ holdings are largely devoid of other words related to domestic violence, such as “beat,” “hit,” “punch,” “kick,” and so on. Finally, when judges did use violence words, they often did so in the process of invalidating plaintiffs’ claims of abuse on evidentiary grounds or on the grounds that the incidents were not serious enough to constitute domestic violence.

A Typology of Judicial Discourse in Adjudicated Denials

While Marriage Law words in general and Article 32 words in particular formed the core of judges’ lexicon of adjudicated denials, they were similarly central to holdings to grant divorces. Words associated with Article 32 were therefore not uniquely constitutive of the language of adjudicated denials. Because they were not deployed specifically for the purpose of denying divorce petitions, they did not define a distinct discourse of adjudicated denial. Other words, however, did form four discourses strongly associated with adjudicated denials of divorce petitions: (1) ideological words, (2) Fourteen Articles words, (3) other therapeutic words, and (4) evidentiary words. I will discuss each in turn.

First, in contrast to the Marriage Law words I just described, ideological words were deployed for the specific purpose of denying divorce petitions and do define a distinct discourse of adjudicated denial. I already showed that four words with political valence in China – ”stability,” “civilized,” and two synonyms for “harmonious” – appeared in a lot of holdings, particularly in the Henan sample. Of these four ideologically salient words, only “stability” (稳定) and one of the words for “harmony” (和睦) were among the top 50 most frequently used words in any word cloud: They both appear in Panel A of Figure 8.1 for adjudicated denials. Judges were far more likely to use these words when denying divorces than when granting divorces. Judges often riffed on the ideological language of socialist morality by holding that divorce would be to the detriment of family unity and harmony; that litigants should stay together and work earnestly to protect harmonious and civilized marital and family relations; that for the sake of family and social harmony and stability, the litigants should try to reconcile; and so on.

China’s ongoing domestic relations trial reforms have renewed the supply of this discursive grist for the mill of adjudicated denials. We previously saw that judges parroted ideological discourse by holding that “the family is the cell of society” (also see Reference FincherFincher 2014:23–24). Divorce decisions are not merely legal matters but also political matters. The following adjudicated denial is a case in point:

This example also anticipates my discussion in the following pages about the importance of judges’ evidentiary discourse, which they used to invalidate plaintiffs’ legally valid evidence.

Second, the SPC’s 1989 judicial interpretation, known as the Fourteen Articles, is a major source of vocabulary in judges’ discourse of adjudicated denial. Each italicized word and phrase in the rest of this paragraph appears in the Fourteen Articles. Because the Fourteen Articles stipulates that inadequate premarital acquaintanceship constitutes grounds for affirming the breakdown of mutual affection, plaintiffs often claimed not to know the nature of the person they married until it was too late, whereas judges held that both sides had established a solid premarital acquaintanceship. Likewise, because the Fourteen Articles calls for consideration of the marital foundation when determining the state of mutual affection and marital relations, plaintiffs claimed a weak foundation, whereas judges claimed a solid foundation. Because the Fourteen Articles stipulates as grounds for divorce difficulty living together owing to wrongdoing – including maltreatment – and failure to establish or build marital affection, plaintiffs claimed that living together was impossible, whereas judges held that the litigants had already built definite marital affection in their years living together. Finally, because the Fourteen Articles stipulates that marital affection should be determined according to reconciliation potential, plaintiffs claimed the impossibility of reconciliation, whereas judges held that reconciliation remained possible.

Some of these italicized words and terms – most notably, “marital,” “affection,” “breakdown,” and “both sides” – also appear in the Marriage Law. But others – most notably, “establish” or “build” (建立), “foundation” (基础), “reconcile” (和好), and “possibility” (可能) – are unique to the Fourteen Articles. Judges put the words “reconcile” and “possibility” together to form “reconciliation potential.” These words and expressions should be familiar from the case examples in Chapter 7. Although the Marriage Law stipulates that the breakdown of mutual affection is grounds for divorce, it contains no legal test to help judges determine “whether mutual affection has indeed broken down” (part of the title of the Fourteen Articles). For this reason, these three words and terms intended to help judges assess the strength of mutual affection – namely “build,” “foundation,” and “reconciliation potential” – unique to the Fourteen Articles appeared in the clear majority of adjudicated denials in my samples.

Several courts in Zhejiang were particularly fond of quoting the Fourteen Articles almost verbatim: “After conducting a comprehensive analysis of the marriage’s foundation, postmarital affection, grounds for divorce, the current state of marital relations, and other aspects, the court’s holding is to confirm that marital affection between plaintiff and defendant has not completely broken down and that reconciliation is possible” (e.g., Decision #3417261, Yiwu Municipal People’s Court, Zhejiang Province, March 26, 2015).Footnote ⁷ When judges used the word “foundation” in a manner consistent with the Fourteen Articles, namely to refer to the foundation of marital affection, they sometimes added ideological words to the discursive mix. Another court in Zhejiang, for example, frequently held that “Marital and family relations should be constructed on a foundation of civilization, equality, and harmony” (e.g., Decision #3592860, Huzhou Municipal Wuxing District People’s Court, Zhejiang Province, June 4, 2015).Footnote ⁸

Judges sometimes used the word “foundation” to deny divorce petitions in an exclusively ideological manner inconsistent with the Fourteen Articles. In 86 of its holdings in my sample, Henan’s Taikang County People’s Court held that “mutual affection is essential for marital preservation, and family stability is the foundation of social stability.” Other courts similarly held that “harmonious and stable families are the foundation of a harmonious society” (e.g., Decision #1305470, Qi County People’s Court, Henan Province, October 31, 2014)Footnote ⁹ and “the family is the cell of society; harmonious and civilized marital and family relations are the foundation of a harmonious and civilized society” (e.g., Decision #1353450, Xia County People’s Court, Henan Province, January 13, 2015).Footnote ¹⁰ Notwithstanding some variation in usage, judges tended to invoke the word “foundation” in the context of the “marital foundation” as stipulated by the Fourteen Articles.

Because they were so frequently used by judges to deny divorce petitions in both provinces, the words “build,” “foundation,” “reconcile,” and “possibility” are in all four word clouds. Judges’ reliance on the Fourteen Articles helps us understand the basis of their impulse to privilege breakdownism over faultism. Both the 1989 Fourteen Articles and the 2001 Marriage Law provide fault-based grounds on which judges may affirm the breakdown of mutual affection. Neither source of law, however, stipulates that judges must affirm the breakdown of mutual affection when they affirm wrongdoing. Meanwhile, judges roundly ignored a separate source of law – the 2001 Interpretations of the SPC on Several Issues Regarding the Application of the Marriage Law – requiring them to grant divorces when a faultism test is satisfied (Article 22; Reference JiangJiang 2009b:18). Indeed, I found only ten cases (out of almost 150,000 first-instance divorce adjudications) in which judges explicitly cited this provision. (They granted the divorce in all ten of them.) Judges routinely affirmed the occurrence of domestic violence but nonetheless denied the divorce petition by holding that mutual affection had not broken down and that reconciliation remained possible. In essence, judges rendered adjudicated denials as if the SPC never issued its judicial interpretation in 2001 requiring them to privilege faultism over breakdownism.

Judges also contorted facts and evidence to ensure that faultism tests were not satisfied in the first place. As we saw in Chapter 7, judges routinely held that plaintiffs’ allegations of domestic violence did not meet the legal definition of domestic violence. Even in the face of overwhelming evidence of domestic violence, including confessions by defendants, judges routinely – and inexplicably – held that plaintiffs’ allegations could not be proven. Normalizing domestic violence by reducing them to the mere family trifles and petty squabbling that (they asserted) were unavoidable and intrinsic features of marriage was one of judges’ gaslighting strategies for privileging breakdownism over faultism. Contriving an inability to affirm wrongdoing, as judges so often did, helped them skirt the SPC’s 2001 judicial interpretation requiring them to grant divorces when they did affirm wrongdoing.

The ease and efficiency of denying a divorce petition simply by holding – on the ostensible basis of a comprehensive analysis of the current state of marital relations – that reconciliation is possible because husband and wife built a solid marital foundation helps explain the enduring allure to judges of the Fourteen Articles. And yet, despite its profound influence, judges cited the Fourteen Articles by name in less than 1% of their holdings in all first-attempt divorce adjudications in my samples.

Third, judges’ therapeutic discourse emerged from the vocabulary they used in their relationship advice to litigants. To be sure, judges also used Fourteen Articles vocabulary therapeutically. When they held that reconciliation remained possible, they expressed hope and encouragement to litigants, urging them to invest time and effort into strengthening the marital foundation they had already built in their years of life together.

Most of the words that formed their therapeutic discourse, however, are altogether outside the scope of the law. Judges normalized abuse by diminishing it to ordinary conflict (矛盾, literally meaning “contradiction”) and family trifles (琐事) that, with commitment and effort, could be overcome. They held that couples could overcome this minor and unavoidable friction by virtue of having already built a definite (一定) marital foundation. Although, as judges so often held, marital problems had occurred (发生), reconciliation was possible if both sides improved their communication (沟通) and worked to cultivate greater mutual (相互) understanding, care, and consideration. Each of these six italicized words appears in all four word clouds in Figure 8.1.

An additional ten words that helped form judges’ therapeutic discourse appear in at least one of the word clouds. The fact that both wife and husband registered their marriage voluntarily (自愿) after freely (自由) forming a romance (恋爱) was proof, according to judges, that the couple had built a good (较好) foundation of mutual affection. Although marital life inevitably produced (产生) some conflict, judges urged both sides to strengthen (加强) their communication skills, build mutual understanding and rapport (互谅), accommodate one another and compromise (互让), cherish (珍惜) their families, and maintain (维护) family harmony and stability for the sake of social harmony and stability. Therapeutic discourse accompanied the vast majority of adjudicated denials.

Fourth, judges grounded their adjudicated denials in evidentiary discourse. They held that plaintiffs failed to prove (证明) their claims because they failed to supply (提供) valid evidence (证据). Each of these three italicized words appears in all four word clouds. Judges also held that the evidence plaintiffs did submit (提交) was insufficient (不足) to support their claims. Judges’ frequent use of evidentiary discourse to deny divorce petitions is also reflected in a legal provision they were fond of citing: Article 64 of the Civil Procedure Law, which stipulates that “litigants are responsible for supplying evidence to support their claims.” This specific provision was cited in 33% and 12% of all first-attempt adjudicated denials in the Henan and Zhejiang samples, respectively, more than double the rate at which it was cited in first-attempt adjudicated approvals. Judges invoked this provision as a legal pretext for denying divorce petitions.

Even when plaintiffs did submit evidence, judges often held that it was circumstantial or otherwise insufficient to support their claims, as we saw in Chapter 7. Moreover, judges’ use of evidentiary discourse to deny divorce petitions belied judicial rules and interpretations issued by the SPC obliging them to relax this provision, shift the burden of proof to the defendant, and adopt an alternative “preponderance of evidence” standard in domestic violence cases (Chapter 2). This alternative standard – namely, Article 73 of the 2001 Several Provisions of the SPC Concerning Civil Procedure Evidence – appeared in only two out of the nearly 150,000 first-instance adjudicated divorce decisions in my samples. Judges almost always applied conventional standards of evidence to domestic violence cases as if the SPC had never issued special instructions for the purpose of extending the benefit of the doubt to vulnerable abuse victims.

So far I have identified the words and terms judges used most frequently in two sets of holdings: (1) adjudicated denials of cases that do not involve domestic violence allegations and (2) all adjudicated decisions (both to deny and to grant divorces) in cases that do involve domestic violence allegations. Vocabularies in these two sets of holdings were strikingly similar both because judges tended to deny the divorce petitions of plaintiffs who made domestic violence allegations and because judges’ gaslighting strategies were similar in cases that did and did not involve domestic violence allegations. Table 8.2 summarizes the key words and terms in each of the four judicial discourses of adjudicated denial. Frequency distributions of words, which are reflected in the word clouds in Figure 8.1, tell us how often judges used them in their holdings but tell us little about their contextual meanings. So far I have provided only selected examples of context for the words that formed four types of judicial discourse in holdings to deny divorce petitions. I will now present results from a hierarchical cluster analysis (HCA) of the 50 most frequently used words in each sample of holdings.

Patterns of Judicial Discourse in Adjudicated Denials

By identifying words that tended to be used in tandem and thus to cluster together, HCA is a useful tool for contextualizing words and teasing out discursive patterns in the entire corpus of holdings. I limit the scope of the HCA to domestic violence cases because one of my key tasks in this chapter is to assess the influence of plaintiffs’ domestic violence allegations on judges’ holdings and verdicts.

Let us now take a closer look at how judges used the words depicted in the word clouds for first-attempt divorce adjudications involving allegations of domestic violence. Figure 8.2 contains two dendrograms – one for each province – depicting average conditional probabilities of the co-appearance of words and clusters of words within individual holdings. That is, the unit of analysis is the holding; the results of the HCA show the words judges tended to use together in the same holding. The holdings I used to construct the word clouds in Panels C and D of Figure 8.1 are the same ones I used in the HCA. Like the word clouds, each dendrogram contains the 50 most frequently used words in its corresponding provincial sample of holdings. The words in Panels A and B of Figure 8.2 are thus the same as those in Panels C and D of Figure 8.1, respectively. Both dendrograms share 38 words in common (denoted by a heavier font). Every pairwise combination of words in each dendrogram exists in its corresponding sample of holdings. Put another way, every possible pair of words in a dendrogram can be found within at least some of the holdings from which it was constructed. Although all clusters in each dendrogram are therefore connected to one another, I removed weaker links in order to facilitate the identification of word clusters. More specifically, I removed links connecting clusters of words whose average conditional probability of co-appearance in the holdings was 0.2 or less. As I present findings from the HCA, I will illustrate key discursive patterns with examples both of phrases and sentences judges commonly constructed with words in the dendrograms and of synonyms judges commonly used in lieu of words in the dendrograms.

Figure 8.2 Hierarchical cluster analysis of top 50 words in domestic violence cases

Note: Limited to first-attempt adjudications. The words in these dendrograms in Panels A and B are identical to those depicted in Panels C and D of Figure 8.1, respectively, because they were derived from holdings in first-attempt divorce decisions containing allegations of domestic violence. Major clusters are numbered and minor clusters are lettered. Words common to both dendrograms are in a heavier font (38 words); unique words are in a lighter font (24 words in total, 12 words per dendrogram). Words are clustered using the farthest neighbor (or complete linkage) method according to Kulczynski’s similarity measure of the average conditional probability that Word B is present in a holding given that Word A is present in the same holding.

Source: Author’s calculations from Henan and Zhejiang provincial high courts’ online decisions.

We already know that well over 90% of holdings in first-attempt divorce cases contained the word “affection.” We can see in Figure 8.2 that this word almost always appeared together with the words “marital” and “breakdown” in both provincial samples (Cluster 1a). This is hardly surprising given that these three words appear together in both the Fourteen Articles and Article 32 of the Marriage Law. Nor should we be surprised that the words “deny” and “support” clustered together in both samples of holdings (Cluster 1b), given that judges tended to rule to “deny support of the plaintiff’s divorce petition.” Judges’ holdings to deny divorce petitions often referenced plaintiffs’ legal complaints: the court denies support of “the plaintiff’s litigation request demanding a divorce” (原告要求离婚的诉讼请求). For this reason, the words “demand,” “request,” and “litigation” clustered together in both Henan (Cluster 1c) and Zhejiang (Cluster 1d).

The dendrograms also contain clusters of words judges used to grant divorces. They are less prominent, however, because judges so rarely granted first-attempt divorce petitions even in cases involving allegations of domestic violence. Henan’s Cluster 2 and Zhejiang’s Cluster 3c contain the word “grant” as well as words associated with property division and child custody.

Given that the breakdownism test requires determining whether “mutual affection has indeed broken down,” we also know that the word “has indeed” was inextricably linked to the words “breakdown” and “marital affection” in both the Marriage Law and the Fourteen Articles. In judges’ holdings, however, “has indeed” was more closely linked to evidentiary discourse. “Has indeed” was closely linked to “evidence,” “supply,” and “prove” in both Henan (Cluster 1d) and Zhejiang (Cluster 3b). Many if not most adjudicated denials contain variations of a boilerplate sentence along the lines of: “The plaintiff failed to supply evidence proving her claim that mutual affection has indeed broken down.”

In Henan, judges tended to use the Fourteen Articles word “foundation” in conjunction with the word “definite” (Cluster 5d). They frequently held that litigants possessed a “definite marital foundation” (一定的婚姻基础 or 一定的感情基础). This word cluster was strongly associated with another Fourteen Articles word cluster (Cluster 5e) containing the words “reconcile” and “possibility.” Judges often held that reconciliation was possible because the marriage rested on a solid foundation. Judges expressed their optimistic prognosis for reconciliation using a variety of synonyms for “reconcile” (e.g., 可调和, 重归于好, 如初, 重新, 修复, 恢复, 有望). As we continue to review judges’ flimsy determinations of the existence of mutual affection and reconciliation potential, bear in mind that each case in the HCA involves an allegation of domestic violence.

In Zhejiang’s holdings, the words “definite” and “build” were part of a cluster of words that also included “give birth” (Cluster 2b). In addition to the word “build” (which can also be translated as “establish”) that appears in the Fourteen Articles, judges also used a few synonyms (e.g., 构建, 共建, 组建) to refer to building happy families, building a harmonious society, and so on. This cluster in turn was connected to another cluster that included “marry” and “register” (Cluster 2a). When judges affirmed the validity of the marriage by holding that it was lawfully registered, they did so not only to establish the court’s standing. Insofar as marrying under duress constitutes grounds for divorce according to the Fourteen Articles (Article 6), judges used these words to hold that litigants had registered their marriages on their own volition. Judges also regarded voluntary marriage registration as proof that a couple had “built definite marital affection” (e.g., 应视为已建立一定的夫妻感情). Applying the same logic, judges regarded childbirth as additional proof of marital affection. My samples of holdings are full of statements such as “Plaintiff and defendant, having voluntarily registered their marriage and given birth to a son and a daughter, should have a definite foundation of mutual affection.”

Judges in Henan, too, used childbirth as evidence that marital affection had not broken down. Consider the words in Cluster 4. Judges held that occasional anger and arguing over family trifles was unavoidable, and that giving birth was evidence that marital affection was once strong and could therefore be rebuilt. As judges often put it, anger had led to physical separation but had not caused mutual affection to break down completely. They often added that the time duration of separation was relatively short and thus did not satisfy the minimum two-year requirement. Judges also used defendants’ unwillingness to divorce as evidence of marital affection. Although the term “withhold consent” is not among the 50 most frequently used words in the holdings of domestic violence cases, it is in the word cloud in Panel A of Figure 8.1. Judges sometimes cited the “earnest” (诚恳) “wishes” (愿望) of defendants – even those who had committed domestic violence – to reconcile as evidence of the couple’s marital affection and reconciliation potential. Judges similarly cited abusive defendants’ “remorse” (悔改, 后悔) for their wrongdoing, promises to “rectify” themselves (改正, 改造, 改错, 改善), and hopes for another chance to reconcile as grounds for denying divorces. As we saw in Chapter 7, judges sometimes used defendants’ apology letters as grounds for their adjudicated denials. Judges also used the words “child” and “children” to justify denying divorces “for the sake of the healthy upbringing of children.”

Zhejiang’s Cluster 1 contains Fourteen Articles words. The word “foundation” is most closely linked to the word “postmarital” (Cluster 1e). Sentences in which judges used both words include the following few examples: both the marital foundation and postmarital affection are good; both the premarital foundation and postmarital affection are good; although the marital foundation is good, disputes that emerged in their postmarital life together have not been properly handled; both sides built a good premarital foundation, their postmarital life together has been long, and they gave birth to a daughter; and plaintiff and defendant possess a good foundation for their marriage, and in their postmarital life together have built definite marital affection. Judges used a variety of synonyms for “good” when describing the quality of the marital foundation (e.g., 一般, 尚好, 尚可).

In both Henan and Zhejiang, Fourteen Articles words were closely linked to therapeutic words. In Henan, the Fourteen Articles words “build,” “foundation,” “reconcile,” and “possibility” belong to a cluster of words that also includes the therapeutic words “both sides,” “live,” “together,” “family,” “trifles,” “conflict,” “occur,” “mutual,” and “communication” (Cluster 5). One of Zhejiang’s corresponding clusters that includes “foundation,” “reconcile,” and “possibility” also includes “both sides,” “live,” “together,” “family,” “trifles,” “occur,” and “cherish” (Cluster 1). Zhejiang’s Cluster 2 also contains the Fourteen Articles word “build” as well as the therapeutic words “rapport,” “conflict,” “produce,” “communication,” and “strengthen.”

Judges held that marital conflict “occurred” as a result of – or was “produced” by – “family trifles” (家庭琐事), and was thus a normal and even inevitable part of marriage. While “conflict” was their word of choice in this context, they had a repertoire of synonyms (e.g., 冲突) and words that similarly expressed marital discord, including “estrangement” or “grown apart” (隔阂) and “argument” or “dispute” (争执, 吵架, 纠纷, 分歧, 吵闹, 争议, 争吵). A common synonym for “trifle” was “friction” (摩擦). Judges also used synonyms for “occur” and “produce” in this context (e.g., 造成, 导致). When they trivialized and normalized domestic violence, judges often held that “occasional” (时有, 偶尔) “rifts” (裂痕) did not rise to the “level” (地步) of “major” (大的) or “fundamental” (根本性, 原则性) marital problems. Some judges responded to allegations of domestic violence by holding that “opposition to frivolous divorce is a fundamental spirit of China’s Marriage Law” (Decision #3128494, Quzhou Municipal Kecheng District People’s Court, Zhejiang Province, October 30, 2014).Footnote ¹¹ And they held that litigants’ marital issues were not serious enough to cause the breakdown of mutual affection and that plaintiffs had either failed to submit evidence in support of their claims otherwise or had submitted evidence that was “insufficient” (不足, 不充分) to prove their claims.

Judges’ acknowledgment of marital discord was the premise of their relationship advice. They held that marital damage could be mended – and even that documented domestic violence could be prevented in the future – if both sides “worked harder” (加强, 多加, 增强, 增进, 努力, 尽到, 坚持, 进一步) on their relationship skills. They advised litigants to fix their “shortcomings” (不足之处, 缺点), deal with conflict “correctly” (正确, 妥善), and “improve” themselves (改正, 改造, 改错, 改善). As judges so often explained in their holdings, litigants could overcome the “personality” (性格) differences at the root of their marital discord “provided” (只要) they learned to build “rapport and mutual understanding” (互谅, 体谅, 理解, 融洽); “compromise” or “give and take” (互让); “love, respect, and tolerate each other” (互敬, 互爱, 尊重, 宽容, 包容); “show care, affection, and consideration for one another” (关爱, 爱护, 呵护, 照顾, 体贴); “trust each other” (信任, 互信); “communicate” more regularly and effectively (沟通, 交流); be “honest, sincere, and loyal” (忠诚, 以诚相待); act “rationally” (理性) and with a “sense of responsibility” (责任感, 责任心); and “help and support each other” (互助, 扶助, 帮助, 扶持, 持家).Footnote ¹² Finally, judges also explained that their holdings to deny divorce petitions were for the purpose of providing an “opportunity” (机会) for litigants to “cool off” (冷静) and “reflect on” (思考, 考虑) their commitment to their families. Judges advised litigants to adjust their attitudes and to “cherish” (珍惜, 珍视) and to feel blessed to have their “happy” (美满, 美好, 幸福), “perfect” (完美), “complete” (完整), and “precious and hard-earned” (来之不易) families. We have seen variations of a common cliché in judges’ holdings: “If you cherish what you have you can restore marital harmony.”

Clearly, the top 50 words in each set of holdings are a fraction of the rich vocabulary judges used to deny divorce petitions. Ideological words are absent from both dendrograms in Figure 8.2 because they were not among the top 50 most frequently used words in either sample. This does not mean that ideological discourse was unimportant. As we saw at the outset of this section and can see again in Table 8.3, ideological words appeared in the holdings of 25% and 13% of all first-attempt adjudications in the Henan and Zhejiang samples, respectively. The other three types of judicial discourse were simply more prevalent. Indeed, therapeutic words that appear in at least one word cloud in Figure 8.1 dominated judges’ holdings. At least one of the 16 words comprising therapeutic discourse appeared in 75% and 83% of holdings in the Henan and Zhejiang samples, respectively. Each of the remaining two judicial discourses of adjudicated denials – Fourteen Articles discourse and evidentiary discourse – also appeared in the majority of holdings in first-attempt divorce adjudications in both samples. Finally, Table 8.3 shows that whereas judges’ ideological discourse was more prevalent in the Henan sample than in the Zhejiang sample, Fourteen Articles discourse and therapeutic discourse were more prevalent in the Zhejiang sample than in the Henan sample.

Perhaps the most disconcerting pattern in Table 8.3 is that domestic violence allegations did not reduce judges’ use of any of these types of judicial discourse. On the contrary, judges responded to domestic violence claims with even greater usage of these types of judicial discourse. Domestic violence claims also triggered evidentiary discourse. More often than not, domestic violence claims were invalidated on evidentiary grounds, as we have seen. Next I will show that judges were much more likely to invoke any of the four judicial discourses when the plaintiff was a woman than when the plaintiff was a man.

Gendered Exposure to Judicial Discourses of Adjudicated Denial

Table 8.4 contains two sets of comparisons of the prevalence of judicial discourses: (1) female versus male plaintiffs and (2) divorces denied versus divorces granted. With respect to the first comparison, all four types of judicial discourse in both provincial samples were more prevalent in judges’ holdings when the plaintiff was a woman. Given that domestic violence allegations so greatly increased the incidence of evidentiary discourse, we should not be surprised that the gender gap was greatest there: 7 and 6 percentage points in the Henan and Zhejiang samples, respectively. Judges’ disproportionate use of therapeutic discourse in cases filed by women was also pronounced: a gap of 3 and 5 percentage points in the two respective samples.

With respect to the second comparison, Table 8.4 brings into sharp relief the concentration of all four types of judicial discourse in adjudicated denials. All four types of judicial discourse represent the language judges tended to use to deny divorce petitions. In both samples, judges’ use of these discourses was vastly more likely in adjudicated denials than in holdings to grant divorces. The degree to which these judicial discourses were concentrated in adjudicated denials is truly striking. Differences in the Henan and Zhejiang samples were 21 and 8 percentage points, respectively, in the use of ideological words, 25 and 33 percentage points, respectively, in the use of Fourteen Articles words, 21 and 36 percentage points, respectively, in the use of other therapeutic words, and 40 and 44 percentage points, respectively, in the use of evidentiary words. The incidence of these judicial discourses in adjudicated denials was generally about double their incidence in holdings to grant divorces.

If women were at greater exposure to these four judicial discourses, and if these four judicial discourses were associated with adjudicated denials, then perhaps women’s greater risk of having their petitions denied explains why they were more exposed to these four judicial discourses. In other words, perhaps judges were more likely to invoke a judicial discourse of denial in cases filed by women simply because they were more likely to deny the divorce petitions of women. If this is true, then female and male plaintiffs who won their cases should have been similarly or identically exposed to these judicial discourses. Likewise, female and male plaintiffs who lost their cases should have had essentially the same level of exposure to these judicial discourses.

We can easily test this possibility with regression analysis. If women’s greater exposure to judicial discourses of denial was simply a function of their greater likelihood to lose their cases, then controlling for the outcome should erase gender differences in exposure to the four judicial discourses. The results presented in Table 8.5 entirely support this expectation. The regression models in Table 8.5 also allow for a comparison of gender gaps between rural and urban courts. Rural courts are those in counties and county-level cities, and urban courts are those in urban districts (Chapter 4).

To facilitate the interpretation of the regression models in Table 8.5, I converted regression coefficients into average marginal effects (AMEs), which are interpreted simply as changes in the probability of a given outcome associated with changes in a given explanatory variable. For example, in Model 1 for Henan’s rural courts, changing the plaintiff from male to female is associated with a .03 increase in the probability that at least one ideological word appeared in the holding. Zhejiang’s corresponding AME was identical. AMEs for female plaintiffs in models for judges’ use of Fourteen Articles words in rural courts (Model 3) were identical or nearly so. The magnitude of AMEs for female plaintiffs was even greater in models for the appearance of other therapeutic words and evidentiary words in rural courts’ holdings (Model 5).

Note, however, that AMEs for female plaintiffs are only positive and statistically significant (at the conventional level of P < .05) in models for rural courts. Female plaintiffs in urban courts, by contrast, were no more exposed than male plaintiffs to these four judicial discourses (with the possible exception of therapeutic discourse in Zhejiang, where the AME is only marginally statistically significant).

Once the verdict to grant or deny the divorce petition is introduced into the models for rural courts, the effect of plaintiff sex almost or entirely disappears. The interpretation of this pattern is clear: what originally appeared to be a gender difference in exposure to judicial discourses of denials was actually the effect of a gender difference in adjudicated denials. Judges were more likely to use judicial discourses of denials in their holdings when the plaintiff was a woman because they were more likely to deny the divorce petitions of women. At the same time, this pattern is limited to rural courts. In urban courts, by contrast, for the simple reason that judges were equally likely to deny the divorce petitions of female and male plaintiffs, judges were also equally likely to infantilize female and male plaintiffs with holdings ordering them to stay together for the sake of society, the nation, their children, and their own happiness. Women were disproportionately targeted with judicial discourses of adjudicated denial only because they were disproportionately targeted for adjudicated denial. Judges were more likely to gaslight women because they were more likely to deny their divorce petitions.

Descriptive Correlates of Adjudicated Denials

Reading and manually coding even a small fraction of documents in a corpus of this size is infeasible. Chapter 4 documents my methods of machine-coding the contents of court decisions into measures that I use to analyze divorce verdicts in the remainder of this chapter. My dependent variable – my object of inquiry, the outcome I try to explain – consists of verdicts in first-attempt divorce trials. There are only two possible outcomes in this analysis of judicial decision-making: a granted or denied divorce petition.

China’s judicial clampdown on adjudicated divorce has been achieved in no small part on the backs on women. Figure 8.3 disaggregates by plaintiff sex the long-dash lines in Figure 6.2 (Panels B and C) depicting China’s judicial clampdown on divorce. Panels A and B of Figure 8.3 show a wide gap between female and male plaintiffs in the probability of an adjudicated denial. The overall gender gaps were 11–12 percentage points in the two samples. In Henan, the gender gap widened over time from 2 percentage points in 2009 to 13 percentage points in 2015. Among first-attempt divorce adjudications, the probability of a denial increased from 43% to 66% for men and from 45% to 78% for women (Panel A). As we saw in Figure 6.2, trends were flatter in Zhejiang, particularly after 2009. Zhejiang’s gender gap remained stable at 12–13 percentage points from 2010 to 2016 (Panel B). By 2015, Henan’s denial rates had almost caught up with Zhejiang’s. In 2015, 75% and 78% of adjudicated first-attempt divorce petitions were quashed in the Henan and Zhejiang samples, respectively. Meanwhile, in the same year, among female plaintiffs, denials accounted for 78% and 82% of all first-attempt adjudications in the two respective samples.

Figure 8.3 Proportion of first-attempt divorce petitions (%) denied

Note: n = 54,200 and n = 8,626 first-attempt adjudicated decisions (granted or denied) from Henan and Zhejiang, respectively. All sex differences are statistically significant (χ², P < .01). Panels A and B are smoothed with moving averages. For more information on scatterplot points, see the note under Figure 4.5.

Source: Author’s calculations from Henan and Zhejiang provincial high courts’ online decisions.

In short, women’s divorce requests were far more likely than men’s to be denied on the first attempt. Women’s disproportionate burden was compounded by five factors. First, as we can also see in Figure 8.3, the gap between female and male plaintiffs in first-attempt divorce denial rates was widest in rural areas where most divorce petitions were filed. In the two samples, the average gender gap (the female denial rate minus the male denial rate) was 14–15 percentage points in rural courts, which we know contained the majority of both people and divorce petitions. Figure 8.3 also shows that urbanization not only shrank the gender gap, but also, at least in the case of Henan, reversed it. No different from the gender gap in exposure to judicial discourses of denial discussed in the previous section, the gender gap in adjudicated outcomes appears also to have been largely a rural phenomenon. In urban courts, the average gender gap flipped to −3 percentage points in Henan and shrank to 5 percentage points in Zhejiang.Footnote ¹³ Consider Henan’s provincial capital of Zhengzhou, labeled in Panel C. In Zhengzhou’s seven district courts, which together supplied 953 decisions, female plaintiffs were more likely than male plaintiffs to win the divorces they sought. The rate at which divorce petitions were denied was 80% and 88% among female and male plaintiffs, respectively. The positive impact of urbanization, however, was relatively limited, as only 16% and 33% of first-attempt divorce adjudications in the Henan and Zhejiang samples, respectively, were filed in urban courts.

Second, Table 8.6 reproduces what we already saw in Figure 7.1, namely that women were disproportionately exposed to marital violence. Table 8.6 presents descriptive characteristics of key explanatory variables in the regression analysis later in the chapter. In both samples, female plaintiffs made allegations of domestic violence in almost 40% of their petitions. The remarkable discursive similarities we saw between judges’ holdings in adjudicated denials and judges’ holdings in domestic violence cases suggests that allegations of domestic violence did not sway judges to grant divorces. Indeed, because domestic violence allegations increased the likelihood that judges invoked a judicial discourse associated with adjudicated denials, we should not be surprised to discover that domestic violence allegations were also positively associated with adjudicated denials.

Third, the improbability of obtaining an adjudicated divorce on the first attempt disproportionately impacted women in part because male defendants were more likely than female defendants to withhold consent. Table 8.6 shows that cases in which the defendant withheld consent represent by far the largest category of the “defendant consent and absenteeism” variable and account for at least one-half of all first-attempt divorce trials in each sample. Within this category, public notice trials were far rarer in Zhejiang than in Henan, perhaps because they reduce judicial efficiency by virtue of the requirement that they be conducted according to the ordinary civil procedure (Chapters 2 and 5). Meanwhile, defendants in only a small proportion of cases (15% and 14% in the Henan and Zhejiang samples, respectively) consented to divorce. Table 8.6 also shows that female plaintiffs were more likely than male plaintiffs to face defendant obstructionism and less likely than male plaintiffs to have defendant consent. We know that judges used defendants’ unwillingness to divorce as evidence of mutual affection and thus as grounds for denying divorce petitions. We also know that judges fear provoking the violent wrath of disgruntled male defendants.

Figure 8.4 shows that the gender gap in defendant obstructionism remained fairly stable over time. It also shows how defendant obstructionism varied by urbanization. Female plaintiffs’ disadvantage was another exclusively rural phenomenon in both samples. Whereas female plaintiffs were more likely than male plaintiffs to face defendant obstructionism in rural areas, the opposite was true in urban areas (although the small difference in Zhejiang’s urban courts was not statistically significant). Once again, Henan’s capital of Zhengzhou is illustrative of urban courts more generally: whereas the husbands of 55% of female plaintiffs withheld consent, the wives of 71% of male plaintiffs withheld consent. On the whole, female plaintiffs’ advantage in urban courts such as those in Zhengzhou was far overshadowed by their disadvantage in rural courts, where the vast majority of divorce adjudications occurred.

Figure 8.4 Proportion of defendants (%) who withheld consent to divorce

Note: Lines for females and males refer to plaintiffs. Lines for female plaintiffs are interpreted as the proportion of defendants who withheld consent to divorce when the plaintiff was female. n = 54,200 and n = 8,626 first-attempt adjudicated decisions (granted or denied) from Henan and Zhejiang, respectively. With the exception of urban courts in Zhejiang, all sex differences are statistically significant (χ², P < .001). Panels A and B are smoothed with moving averages. For more information on scatterplot points, see the note under Figure 4.5.

Source: Author’s calculations from Henan and Zhejiang provincial high courts’ online decisions.

Fourth, as we can also see in Table 8.6, female plaintiffs were less than half as likely as male plaintiffs to have public notice trials. Because, as we will see, judges were relatively inclined to grant divorces when defendants were unable or chose not to participate in first-attempt trials, women’s lower chances of success in their attempts to divorce are explained in part by men’s vast overrepresentation among plaintiffs in public notice trials.Footnote ¹⁴

Figure 8.5 depicts patterns with respect to courts’ utilization of the ordinary procedure. Note that Henan’s time trend – its turn away from the ordinary procedure in favor of the simplified civil procedure beginning in 2012 – reflects what we already saw in Figure 5.1. Zhejiang’s far greater aversion to the ordinary procedure also reflects what we saw in Figure 5.1. Figure 8.5 disaggregates the application of the ordinary civil procedure according to plaintiff sex. Men who filed for divorce were much more likely than women to have their cases tried according to the ordinary civil procedure. This gender gap narrowed or altogether disappeared with urbanization; like other gender gaps, it was primarily a rural phenomenon.

Figure 8.5 Ordinary civil procedure utilization rate (%) in first-attempt divorce trials

Note: n = 54,200 and n = 8,626 first-attempt adjudicated decisions (granted or denied) from Henan and Zhejiang, respectively. With the exception of urban courts in Henan, all sex differences are statistically significant (χ², P < .001). Panels A and B are smoothed with moving averages. For more information on scatterplot points, see the note under Figure 4.5.

Source: Author’s calculations from Henan and Zhejiang provincial high courts’ online decisions.

As mentioned in Chapter 4, the ordinary civil procedure was much more common when the defendant was in absentia because its application is legally mandated in public notice trials. Courts applied the ordinary civil procedure in practically 100% of public notice trials, as required by the SPC. Because my measure for civil procedure and my measure for defendant consent and absenteeism are therefore partially redundant (i.e., we already know the type of civil procedure that courts used in public notice trials), in the regression analyses later in the chapter I removed all public notice trials from the measure of the ordinary civil procedure in order to prevent multicollinearity.

The gender gap in courts’ use of the ordinary civil procedure was partially an artifact of a gender gap in defendant absenteeism. In other words, gender gaps in public notice trials depicted in Figure 8.6 mirror gender gaps in courts’ application of the ordinary civil procedure in Figure 8.5. To some extent, female plaintiffs were less likely to have their cases tried according to the ordinary civil procedure because they were less likely to have public notice trials. Indeed, as we can see in Table 8.6, once we remove public notice trials from my measure of ordinary civil procedure, the gender gap narrowed (Zhejiang) or disappeared (Henan). The wives of male plaintiffs were a lot more likely than the husbands of female plaintiffs to be summoned by public notice and absent from their trials. And this is an important reason why male plaintiffs were more likely than female plaintiffs to have their cases tried according to the ordinary civil procedure. Once again, this gender gap with respect to public notice trials was limited almost entirely to rural courts.

Figure 8.6 Public notice trials (%) among all first-attempt divorce trials

Note: n = 54,200 and n = 8,626 first-attempt adjudicated decisions (granted or denied) from Henan and Zhejiang, respectively. All sex differences are statistically significant (χ², P < .001). Panels A and B are smoothed with moving averages. For more information on scatterplot points, see the note under Figure 4.5.

Source: Author’s calculations from Henan and Zhejiang provincial high courts’ online decisions.

Altogether, about one-third of plaintiffs in each sample did not face defendants in their first-attempt trials. A far greater share of defendants was AWOL in Henan than in Zhejiang undoubtedly because Zhejiang is far more urbanized than Henan. We can see in Figure 8.6 that defendants whose whereabouts were allegedly unknown or who opted out of court proceedings for other reasons were overrepresented in rural areas. In the rural courts in my samples, 35–40% of first-attempt trials were held without the participation of defendants (including but not limited to public notice trials), compared with only 25–30% in the urban courts in my samples.

Fifth, the foregoing dynamics that militated against women’s efforts to divorce were multiplied by women’s disproportionate representation among first-attempt divorce petitioners in court. Table 8.6 shows that, as previously reported in Chapter 4 (Figure 4.5), and consistent with previously published estimates reported in Chapter 2, women accounted for 66% and 67% of all plaintiffs filing first-attempt divorce petitions in the Henan and Zhejiang samples, respectively.

The remaining variables in Table 8.6 also pertain to divorce verdicts. Female plaintiffs were more likely than male plaintiffs to submit evidence in support of their legal complaints. As we will see, whether judges treated women’s evidence as seriously as men’s evidence is another matter. The cases of female plaintiffs were more likely than those of male plaintiffs to involve both children and marital property. Judges’ aversion to ruling on complex and contentious matters such as these may have therefore contributed to female plaintiffs’ higher risk of adjudicated denial. Physical separation of at least two years is statutory grounds for divorce that should reduce the probability of an adjudicated denial. Over 40% of cases in Henan and a little over half of cases in Zhejiang involved claims of physical separation. Finally, plaintiffs sometimes gave up claims to property or child custody as a means of obtaining defendants’ consent to divorce and thus of boosting their chances of obtaining a divorce verdict. Women were more likely than men to make this sort of concession in Henan but not in Zhejiang. Plaintiffs’ concessions on property or child custody were concentrated in cases in which the divorce was granted. In Henan, almost one out of five female plaintiffs – but only about one out of 10 male plaintiffs – whose divorce petitions were granted had “voluntarily” given up property or child custody claims. The same pattern, albeit a more muted one, emerges from the Zhejiang sample.

Descriptive characteristics of additional variables not included in Table 8.6 are available elsewhere (Reference MichelsonMichelson 2019c). Let us now turn to a multivariate assessment of the relative importance of domestic violence allegations as grounds for granting divorces (faultism) and defendants’ unwillingness to divorce as grounds for denying divorces (breakdownism).

Multivariate Correlates of Adjudicated Denials

I use AMEs to assess the impact of plaintiff sex, domestic violence (faultism), and defendant consent (breakdownism) on courts’ verdicts, net of control variables included in the regression models. A marginal effect – also known as a first difference – is the difference between the predicted probabilities for each group. An AME is the average of all marginal effects computed for each observation in the sample. The interpretation of AMEs is highly intuitive. An AME can be interpreted as the effect of a variable (say, of changing the value of plaintiff sex from male to female) on the probability the outcome of interest occurs, holding all remaining variables at observed values (Reference Long and FreeseLong and Freese 2014:242–46; Reference MizeMize 2019:85–87).Footnote ¹⁵ An AME of .05 for “female plaintiff” thus means that the probability a court denied a divorce petition was .05 higher for women who filed for divorce than for men who filed for divorce. The difference between two AMEs is known as a second difference.

I will proceed in two steps. In the first step, I will present overall AMEs for plaintiff sex, domestic violence (faultism), and defendant consent (breakdownism). Comparing the magnitudes of these effects allows me to conduct two empirical assessments: (1) the magnitudes of and reasons for gender differences in adjudicated outcomes of divorce trials and (2) the importance judges attached to the faultism and breakdownism standards for granting divorces. In the second step, I will present AMEs for domestic violence (faultism), defendant consent (breakdownism), in absentia trials (defendant absenteeism), and other explanatory factors that vary by plaintiff sex. For example, comparing the effects of defendant consent by plaintiff sex allows me to test whether defendant consent interacted with plaintiff sex.Footnote ¹⁶ Recall from Chapter 4 that I combined defendant consent and defendant absenteeism into a single variable in order to isolate affirmative consent (and affirmatively withholding consent) from failure to withhold consent by virtue of being an absentee defendant and not participating in the litigation process.

I include fixed effects for the court that adjudicated the case in order to account for unobserved heterogeneity across contexts. Because basic-level court jurisdictions correspond to rural counties and urban districts, court fixed effects (court dummy variables) serve the function of controlling for unobserved characteristics of both courts and the contexts in which they are embedded. For example, court fixed effects control for court-level variation in caseloads, an issue at the heart of Chapters 5 and 6. Court locations also reflect and therefore control for the social origins of divorce litigants. Divorce litigants who hailed from rural areas were overwhelmingly at the mercy of rural courts. As we saw in Chapter 4, migrants from rural areas rarely filed their divorces in urban courts. Rural courts tended to serve rural residents, and urban courts tended to serve urban residents.

The models also include additional control variables. Control variables are essential in order to minimize the possibility that an observed effect is an artifact of an omitted correlate. In order to assess the effects of the variables of central interest among otherwise similar cases, I control for the year of the decision, whether or not the plaintiff submitted evidence, whether or not the plaintiff gave up marital property or child custody, a physical separation claim, the participation of one or more female judges, the civil procedure adopted (ordinary or simplified), marital duration, marital property, children, and the participation of legal counsel. We will see that some of these control variables are important in their own right. Details on the construction of all measures are in Chapter 4.

Table 8.7 contains AMEs estimated from models of judges’ decisions to deny first-attempt divorces in the Henan and Zhejiang samples. In Model 1, which controls only for decision year and court, the AMEs for the gender gap in the probability of an adjudicated denial – average female predicted probabilities minus average male predicted probabilities – are .09 in Henan and .11 in Zhejiang. Model 2 adds claims of domestic violence. In both Henan and Zhejiang, a claim of domestic violence increased the probability of a first-attempt adjudicated denial by .09 and .04 in the Henan and Zhejiang samples, respectively. Model 2 also shows that the highly statistically significant gender gap is reduced by controlling for claims of domestic violence. As we will see, this effect of domestic violence in Model 2 is largely an artifact of (1) an overrepresentation of domestic violence claims in cases in which defendants withheld consent (because judges rarely granted divorces in such cases) and (2) an underrepresentation of domestic violence claims in public notice trials (because judges were reasonably likely to grant divorces in such cases). In short, defendant consent and defendant absenteeism are driving what appears to be an effect of domestic violence.

When “defendant consent and absenteeism” is introduced in Model 3, the gender gap shrinks yet again, suggesting that some of the gender gap in the probability of an adjudicated denial on the first attempt was due to female plaintiffs’ greater exposure to domestic violence and spousal obstructionism, and to the more limited use of public notice trials for their cases. Indeed, defendant consent and absenteeism accounts for the majority of the gender gap in the Henan sample. Just as striking is the sheer magnitude of the effect of defendant consent and absenteeism. Recall from Chapter 2 that statutory grounds for the breakdown of mutual affection can be established relatively straightforwardly both in public notice trials and when the defendant consents. Regression results show that only under these two circumstances was a court reasonably likely to grant a plaintiff’s divorce request. In the full model (Model 4), a spouse’s unwillingness to divorce increased the probability of an adjudicated denial by .51 and .47 in the Henan and Zhejiang samples, respectively. The effect of defendants’ withholding consent was greater than the effect of plaintiffs’ domestic violence allegations on divorce outcomes by dozens of orders of magnitude. Similarly, in absentia trials in which defendants were not served by public notice (because they were not alleged to be missing) increased the probability of an adjudicated denial by .27 and .35 in the Henan and Zhejiang samples, respectively. The relatively minor importance of faultism standards and the major importance of breakdownism standards are also reflected in the minor change in pseudo-R² values between Model 1 and Model 2 and the major change in pseudo-R² values between Model 2 and Model 3.Footnote ¹⁷

In both samples, defendant consent and absenteeism substantially reduced the effect of making a domestic violence claim. When controlling for plaintiff sex, defendant consent, and defendant absenteeism in Model 3, the effect of an apparent domestic violence allegation approached irrelevance in both samples (.02 in Henan and .01 in Zhejiang). Note that with the introduction of additional variables in subsequent models, the effect of a claim of domestic violence almost entirely disappeared (although its effect of .01 remained statistically significant in the Henan sample). Defendant consent and absenteeism explained away most of the effect of domestic violence claims for two reasons: (1) defendants who did not consent to divorce were disproportionately accused of perpetrating domestic violence and (2) plaintiffs were relatively unlikely to make claims of domestic violence in public notice trials. One obvious interpretation is that abusers also tended to be obstructionists. However, given limitations in the data, we cannot entirely rule out an alternative possibility that abuse claims were endogenous to spousal consent: some plaintiffs may have made abuse claims because their spouses were unwilling to divorce. Similarly, although it seems highly plausible that plaintiffs were at much lower risk of domestic violence when their spouses were missing, we cannot entirely rule out an alternative possibility that missing spouses obviated plaintiffs’ perceived need to make abuse claims.

We now have clues that help explain why male plaintiffs were more likely than female plaintiffs to succeed in their efforts to divorce on the first attempt. Women’s sizeable disadvantage in the probability of obtaining an adjudicated divorce stemmed from a triple whammy of gender differences in the incidence of domestic violence, defendant obstructionism (in the form of withholding consent), and missing defendants. We also know that these costs were further amplified by a huge overrepresentation of women among plaintiffs who filed for divorce in court.

The average marginal effect of a claim of abuse in the full model (Model 4) was tiny in both samples, and below the threshold of statistical significance in the Zhejiang sample. Meanwhile, the effect of defendant consent and absenteeism remained immense: it alone contributed more to pseudo-R² values than did all remaining control variables combined, including the court dummies. The effect of defendant consent and absenteeism towers above that of everything else in the model.

Female plaintiffs’ disadvantage persisted in Model 4 net of controls. Even among plaintiffs who were otherwise similar in terms of defendant consent, defendant absenteeism, domestic violence claims, and an array of controls, women were still less likely than men to obtain an adjudicated divorce on the first attempt. I disaggregated rural and urban verdicts by applying the same model separately to rural (Model 5) and urban courts (Model 6). Separately modeling rural and urban court decisions shows that women’s net disadvantage was limited to rural areas in both provinces. We already saw descriptive findings showing that urbanization reduced and erased gender differences in verdicts as well as in two of their key determinants, namely defendant consent and defendant absenteeism. Model 5 shows that, net of controls, the probability of an adjudicated denial in a rural court was statistically significantly higher for female plaintiffs than for male plaintiffs in both provinces (by .04 in Henan and .08 in Zhejiang). Model 6, by contrast, shows no gender difference whatsoever in urban courts in either province. The absence of a gender difference in urban courts is not a function of control variables, as the effect of plaintiff sex remained statistically insignificant even after stripping them out to create an urban-only version of Model 1. For this reason, had Model 1 been limited to urban courts, the AMEs for female plaintiffs would have been much smaller and statistically insignificant in both provinces (details omitted).

We already know that public notice trials, which must be held using the ordinary civil procedure, were associated with relatively low adjudicated denial rates. Regression results also show that the application of the ordinary civil procedure outside the scope of public notice trials reduced the probability of an adjudicated denial by .09 in Henan and by .21 in Zhejiang compared to when judges applied the simplified civil procedure. Which civil procedure judges used was generally a good predictor of its verdict. Judges tended to reserve the ordinary civil procedure for when they granted divorces.

Not surprisingly, courts were less likely to deny the divorce petitions of plaintiffs who reportedly submitted evidence in support of their petitions. Compared to plaintiffs who apparently did not submit evidence, the probability of an adjudicated denial among those who apparently did submit evidence was .05 and .02 lower in the Henan and Zhejiang samples, respectively.

In both provinces, cases that involved both children and property were much more likely than cases that involved only one or the other, or neither, to result in an adjudicated denial. In other words, cases that involved marital property but no children, children but no marital property, or neither, were much less likely than cases that involved both marital property and children to result in an adjudicated denial. Ruling on property division and child custody can be both time-consuming and fraught. In their efforts to maximize judicial efficiency and minimize the possibility of complaints, petitioning, and other “extreme incidents” antithetical to the political imperative of maintaining social stability, judges shied away from granting divorces in relatively complex cases, including those involving domestic violence claims. Moreover, as we have seen in both Chapter 7 and this chapter, judges used children as evidence of mutual affection and thus as grounds for denying divorce petitions.

Owing to their disinclination to grant divorces in cases involving marital property and children, judges’ likelihood of applying the simplified civil procedure perversely increased commensurately with case complexity. Although judges are supposed to use the simplified civil procedure only when “the facts are clear, rights and obligations are unambiguous, and the dispute minor” (see Chapter 5), they tended to do precisely the opposite. In Henan, 34% of lower-complexity cases involving neither marital property nor children, 37–39% of cases involving one but not the other, and 42% of higher-complexity cases involving both marital property and children were tried according to the simplified civil procedure. In Zhejiang, simplified civil procedure utilization rates also increased with case complexity: 65% in lower-complexity cases, 76% in cases involving one item but not the other, and 85% in higher-complexity cases involving both items. Judges averse to wading into the thicket of property division and child custody determinations fast-tracked for adjudicated denial cases involving such matters by designating them as “minor disputes” in which “the facts are clear” in order to apply the simplified civil procedure. In so doing, as we saw in Chapter 5, they deprived litigants of their due process rights. By contrast, lower-complexity cases that did not involve marital property, children, or both were more likely to be tried according to the ordinary civil procedure, because judges were more willing to grant divorces when they could do so without the hassle and risk of ruling on such matters. Domestic violence allegations, which are often complex from an evidentiary standpoint, were similarly associated with higher simplified procedure utilization rates.

Claims of physical separation reduced the probability of an adjudicated denial by .10 in Henan and by .08 in Zhejiang, which comes as no surprise given that physical separation constitutes grounds for divorce in both the Fourteen Articles and the Marriage Law. Nonetheless, a physical separation claim far from guaranteed a successful divorce. Judges were often unconvinced that litigants met the statutory minimum separation period or that separation was the result of the breakdown of mutual affection.

Because withholding consent had the practical effect of preserving a marriage or at least prolonging a divorce petition, defendants weaponized consent by withholding it until plaintiffs made concessions on the terms of the divorce. Even when defendants wanted out of their marriages, they often strategically withheld consent in order to gain an advantage vis-à-vis property division and child custody. Such a tactic forced some plaintiffs to exchange their legal rights for their freedom (Reference LiLi 2022). When plaintiffs gave up their rights, defendant consent magically increased. As we can see in Table 8.6, 50% and 56% of defendants in the Henan and Zhejiang samples, respectively, withheld consent. By contrast, in cases in which the plaintiff gave up property or child custody, only 26% and 44% of defendants in each respective sample withheld consent. Not surprisingly, therefore, abandoning a claim on marital property, child custody, or both dramatically reduced the probability of an adjudicated denial by .32 in Henan and .23 in Zhejiang.

Such concessions were also captured by my case complexity measure. Rather than explicitly stating that they forwent a claim to marital assets (variations of “the plaintiff gives up” identified in Chapter 4), plaintiffs could simply claim that there was no marital estate to contest (variations of “there is no common property” also identified in Chapter 4). Either way increased plaintiffs’ chances of winning their bid for divorce.

Gender Gaps in Adjudicated Denials

My final empirical task in this chapter is to assess conditions under which playing fields were relatively even and uneven between female and male plaintiffs. Table 8.8 goes beyond the contents of Table 8.7 by presenting predicted probabilities from which AMEs in Table 8.7, Model 4, were calculated. In Table 8.7, the AME of .01 for making a domestic violence claim in Model 4 for Henan corresponds to the overall difference in Table 8.8 between plaintiffs in Henan who made claims of domestic violence (.64) and those who did not (.63). Table 8.8 further shows that this slightly positive effect of making a claim of domestic violence is limited to female plaintiffs in the Henan sample. Among female plaintiffs in Henan, the difference between those who made claims of domestic violence (.66) and those who did not (.63) is statistically significant. A claim of domestic violence had no effect among female plaintiffs in Zhejiang or among male plaintiffs in either sample. Let us now consider gender gaps among plaintiffs who made allegations of domestic violence before turning to gender gaps among plaintiffs who shared other characteristics.

In Henan, claims of domestic violence widened the gender gap considerably. Whereas the gender gap was only .02 among plaintiffs who did not make such a claim, it was a much wider .06 among plaintiffs who did make such a claim. The difference between these two gender gaps (a test of second difference) is statistically significant. Thus, the effect of making a claim of domestic violence was greater for female plaintiffs than for male plaintiffs. The obvious interpretation of this pattern is that judges treated men’s domestic violence claims more seriously than women’s domestic violence claims; they more readily dismissed women’s domestic violence claims as unimportant or fabricated. In the Zhejiang sample, by contrast, domestic violence claims were equally irrelevant to women and men alike.

In Table 8.8, we see once again that plaintiffs’ divorce prospects were highest when they passed the breakdownism test with either a public notice trial or defendant consent. The overall predicted probability of an adjudicated denial in a public notice trial (.23 in Henan and .53 in Zhejiang) was far less than the overall probability (.63 in Henan and .80 in Zhejiang). In Henan, plaintiffs’ chances of getting denied were even lower in this type of trial (.23) than in trials in which the defendant expressed consent to divorce (.34). By contrast, when defendants failed to participate in court proceedings for other reasons, overall predicted outcomes were about the same as in all trials taken together. Public notice trials and mutual consent were by far the most realistic pathways to divorce in terms of likelihood of success. They were also the least common pathways, together accounting for only 28% and 20% of first-attempt divorce trials in the Henan and Zhejiang samples, respectively.

Although plaintiffs as a whole benefitted from in absentia trials only when an allegedly missing defendant was served by public notice (thus satisfying the breakdownism standard), male plaintiffs enjoyed a large and statistically significant advantage over female plaintiffs when defendants failed to participate in court proceedings for any reason. Gender differences in the probability of a divorce on the first attempt, ranging from .09 to .23, were massive and statistically significant in both samples when defendants were AWOL, regardless of how the court served the defendant. Indeed, among male plaintiffs in the Henan sample, divorce approached a forgone conclusion (.84) in public notice trials. Tests of second difference show that, among all in absentia trials, public notice trials narrowed the gender gap in Henan (.16 versus .09, a statistically significant difference) and widened the gender gap in Zhejiang (.16 versus .23, a statistically insignificant difference).

Women’s severe disadvantage in the context of in absentia trials – both as plaintiffs and as defendants – is consistent with patriarchal cultural beliefs about women as less credible and less deserving than men. My empirical findings suggest that judges, who themselves were mostly men, took claims about missing spouses more seriously and treated them as more credible when they were made by male plaintiffs. The court decisions in my samples reflect cultural narratives not only about female plaintiffs making false claims of domestic violence in illicit efforts to abscond with marital property and child custody (Reference Epstein and GoodmanEpstein and Goodman 2019), but also about female plaintiffs who, for the same reasons, falsely conceal the whereabouts of their husbands. The court decisions may further reflect judges’ implicit belief that missing female defendants, particularly those they suspected were in illicit extramarital relationships, were less deserving than male defendants of the legal protections and due process rights they lost when they were absent from trials. In short, judges were far more inclined to protect husbands than to protect wives from getting “unwittingly divorced” (Chapter 3).

Courts were also relatively inclined to grant divorces to plaintiffs when defendants expressed their consent. In neither sample were female plaintiffs disadvantaged when both sides agreed to part ways. In the Zhejiang sample, mutual consent put women at a small but statistically insignificant disadvantage (.03). Quite the contrary in the Henan sample, where courts were more inclined to grant divorces to female plaintiffs than to male plaintiffs in the context of mutual consent, albeit to only a small (−.02) and marginally statistically significant extent.

Finally, an adjudicated divorce was highly improbable in the absence of spousal consent. According to Model 4, the average predicted probability of an adjudicated denial when the defendant withheld consent was .85 in Henan and .93 in Zhejiang. In other words, the probability of obtaining a unilateral divorce among plaintiffs whose spouses withheld consent was only .15 in Henan and .07 in Zhejiang. In the context of defendants who withheld consent to divorce, female and male plaintiffs were on a playing field that was similarly harsh to everyone. Adjudicated unilateral divorce prospects were slim for female and male plaintiffs alike on the first attempt. The chances of female and male plaintiffs seeking unilateral divorces in Zhejiang were identical. However, female plaintiffs in Henan had a small (−.02) but statistically significant advantage over male plaintiffs when defendants withheld consent.

Judges likewise treated evidence submitted by male plaintiffs more seriously than evidence submitted by female plaintiffs. Although female plaintiffs were more likely than male plaintiffs to submit evidence (Table 8.6), the probability a court denied a divorce to a female plaintiff who submitted evidence was .05 and .07 greater than it was for a male plaintiff who submitted evidence in the Henan and Zhejiang samples, respectively. By contrast, the probabilities of adjudicated denials were more similar for female and male plaintiffs who did not appear to support their claims with evidence (gender gaps of only .02 in both samples). These patterns suggest that plaintiffs’ failure to submit evidence levelled the playing field and that their submission of evidence widened the gender gap in adjudicated divorce outcomes in both samples. Judges attached greater weight to evidence submitted by male plaintiffs and discounted evidence submitted by female plaintiffs. Evidence benefitted male plaintiffs far more than it benefitted female plaintiffs, strongly suggesting that courts treated men’s claims more seriously than women’s.

Finally, with respect to case complexity, the gender gap in the probability of obtaining an adjudicated divorce was relatively wide in cases that involved marital property but did not involve children. In both samples, female and male plaintiffs were more similarly likely to obtain an adjudicated divorce in cases involving both children and marital assets as well as in cases involving neither children nor marital assets. Male plaintiffs’ significant advantage over female plaintiffs in cases involving marital assets but no children suggests that property division was more contentious than child custody and that courts protected the financial interests of male litigants (both plaintiffs and defendants) more than they protected the financial interests of female litigants (both plaintiffs and defendants). Previous research shows that judges, in their property division rulings in divorce cases, favored men over women (Reference LiLi 2015b). Although an analysis of courts’ rulings on property division is beyond the scope of this book, my findings nonetheless reveal that female plaintiffs were significantly less likely than male plaintiffs to be granted divorces in cases involving marital property. Judges, consciously or unconsciously, were more likely to deny divorce requests when they were made by women who threatened the integrity of marital estates, over which men tend to exercise control, particularly in rural areas. Insofar as judges in both Henan and Zhejiang, when deciding cases involving marital assets but not involving children, were more likely to preserve the marriage if the plaintiff was a woman, they acted, wittingly or unwittingly, to preserve the assets – such as housing and farmland – owned, controlled, or used by men and their families.