Caseloads Were Strongly Associated with Both Population and Economic Conditions

Table 6.1 contains regression models predicting the annual volume of court cases at the provincial and court levels.Footnote ³ At the provincial level, Model 1 shows that Zhejiang closed 309,000 more cases than Henan did in 2011. At the court level, Model 1 shows that Zhejiang’s courts on average closed almost 8,000 more cases than Henan’s courts did. Model 2 introduces per capita GDP. When per capita GDP is held constant in Model 2, the gap between Zhejiang and Henan shrinks, suggesting that Zhejiang’s heavier caseloads were a consequence of its more dynamic economic conditions (and that Henan’s lighter caseloads were a consequence of its weaker economic conditions). Model 3 introduces population size. The huge boost it gives to R² values suggests that population is the more important determinant of caseload size. Its introduction also reopens the gap between Zhejiang and Henan (albeit only slightly in the court-level model). Given that Zhejiang’s population is considerably smaller than Henan’s, Model 3 tells us that the gap in case volumes between the two provinces would have been even greater if their populations had been the same size.Footnote ⁴ In all models, including separate ones for courts in Henan and Zhejiang (Models 4 and 5, respectively), both per capita GDP and population size significantly contributed to case volumes. Doubling per capita GDP (comparable to the difference between the two provinces) was associated with an increase of 157,435 cases at the provincial level and an increase of 2,073 cases at the court level.Footnote ⁵

These effects are unsurprising. More people means more litigation. More economic activity also means more litigation. First-instance civil cases have far outnumbered criminal and administrative cases, and have accounted for the vast majority of all cases in China’s court system. Growth in the volume of civil cases has been driven by contract disputes in general and debt disputes in particular. In my collections of court decisions, contract disputes as a share of all first-instance cases were far more numerous in Zhejiang than in Henan (60% and 37%, respectively in 2014). Within the category of contract disputes, debt was by far the largest subcategory in both provincial collections.Footnote ⁶

The Supply of Judges Was Only Weakly Associated with Caseloads

According to the SPC’s 2002 Several Opinions Concerning Strengthening the Construction of the Profession of Judges, slots for judges in the state personnel system are supposed to be allocated “on the basis of a comprehensive consideration of China’s circumstances, caseloads, land areas and population sizes of court jurisdictions, levels of economic development, and other factors” (Reference TangZ. Tang 2014:45; Reference Xu, Huang and LuXu, Huang, and Lu 2015:133). This loose formula was reaffirmed in the SPC’s 2004–2008 second five-year outline for court reform (Reference LiuLiu 2019:104) and added to the 2018 amended Organic Law of People’s Courts (Article 46). In practice, however, judge quotas for courts have been determined primarily according to the number of people in their jurisdictions. To be sure, judge quotas have also been associated with caseloads, but largely because caseloads themselves have been associated with population size. Population size has been the primary determinant of the number of judges. For this reason, judges in China’s more developed regions have been burdened with heavier per capita caseloads. Further compounding the problem, personnel slots for judges have been typically determined according to the officially registered population (户籍人口) and not according to the population of actual residents. Such a method punishes courts in more prosperous areas such as Zhejiang that receive migrants and rewards courts in less prosperous areas such as Henan that send migrants (Reference Fan and JinFan and Jin 2012:99).

Table 6.2 contains regression models predicting judge population. Model 1 shows that Zhejiang had 5,731 fewer judges than Henan at the provincial level and that Zhejiang’s courts averaged 5.65 fewer judges than Henan’s courts did. Model 2 introduces the volume of closed cases, which appears to be positively and strongly associated with the population of judges. Controlling for – i.e., holding constant – the volume of closed cases also widens differences between Henan and Zhejiang in numbers of judges at both provincial and court levels. Because caseloads have been so much heavier in Zhejiang than in Henan at both provincial and court levels, the coefficients in Model 2 tell us that Henan’s judges would have outnumbered Zhejiang’s by an even greater margin had the volume of cases been the same in the two provinces. In Model 3, the effect of general population size wipes out all other effects at the provincial level and greatly shrinks them at the court level. Because, as we saw from the previous analysis, population size is a key determinant of case volume, the former explains away the effect of the latter at the provincial level and attenuates the effect of the latter at the court level. Among courts that were otherwise seemingly identical in terms of population and economic conditions, an increase of 1,000 cases was associated with an average increase of only 1.6 judges. Courts averaged about 4,000 cases in Henan and about 11,500 cases in Zhejiang. According to Model 4, a caseload increase of 1,000 (or about 25%) would have added only about seven judges in Henan. According to Model 5, a caseload increase of 3,000 (or about 25%) would have added only about six judges in Zhejiang. In short, large differences in caseloads were associated with only small differences in numbers of judges.

The models indicate that Henan’s larger population was the main reason why it had more judges than Zhejiang even though Zhejiang closed 60% more cases than Henan (825,110 versus 515,966). At the provincial level, population was the key determinant of personnel slots for judges. Population growth of 1 million people was associated with an extra 126 judges (Model 3). Put another way, for a province to gain an additional 1,000 judges, it would have needed a population increase of 8 million. At the level of the court, an additional population of 1 million was associated with an additional 38 judges (Model 3). Within the two provinces, however, a population difference of 1 million between court jurisdictions was quite rare. For this reason, we should instead interpret the effect in terms of a population increase of half a million, which would have yielded an average of 19 more judges. This is a substantively large effect given that, in both provincial samples, courts averaged about 60 judges and that court jurisdictions averaged about 600,000 people. A population increase of this amount was associated with an additional 17 judges in Henan (Model 4) and 14 judges in Zhejiang (Model 5).

In short, the supply of judges has had little to do with demand for their services. Judges have been allocated mechanically primarily according to population size, without due consideration of variation in caseloads among provinces and court jurisdictions with similar populations. Figure 6.3 depicts the sheer extent of the correlation between the judge counts and population size at both provincial and court levels: .93 at the provincial level (Panel A) and .78 at the court level (Panel B). Consistent with the regression results, correlations between the number of judges and caseloads were weaker: .74 at the provincial level and .51 at the court level (scatterplots omitted).

Figure 6.3 Association between judge population and general population

Note: Panel A, N = 31 provinces. Panel B, n = 150 basic-level courts (80 in Henan and 70 in Zhejiang). In Panel B, Henan r = .81, Zhejiang r = .74. On sources of numbers of judges, see Chapter 4. Court codes in Panel B are listed with their corresponding court names in supplementary online material at https://decoupling-book.org/. Panel B excludes courts of special jurisdiction and economic and technological development district courts.

Source: Court work reports; Basic Level Legal Artisan 2016a, 2016b, 2016c; Henan Provincial Bureau of Statistics 2015; Zhejiang Provincial Bureau of Statistics 2015; author’s calculations from Henan and Zhejiang provincial high courts’ online decisions.

Henan and Zhejiang in 2011 illustrate the primacy of population size. At the provincial level (Panel A), Henan had the most judges in China because it was so populous (ranked third) even though its volume of closed cases was somewhat less remarkable (ranked seventh). Meanwhile, Zhejiang was in the middle of the pack in terms of judge counts (ranked 12th) because it was also in the middle of the pack in terms of population size (ranked tenth) even though its volume of closed cases was somewhat more remarkable (ranked fourth). Similarly, Sichuan had slightly more judges than Jiangsu because its population was slightly larger, even though Jiangsu’s courts closed far more cases than Sichuan’s. Finally, Chongqing had more judges than Shanghai because its population was larger even though Shanghai’s courts closed far more cases than Chongqing’s.

These provincial patterns extend to the court level (Panel B). Consider one of the busiest courts in Henan, Zhengzhou’s Erqi District People’s Court (1c). Other courts in Henan that closed fewer cases had more judges simply because they had larger populations.Footnote ⁷ Consider the busiest court in Zhejiang, Hangzhou’s Xiaoshan District People’s Court (1h), which also happened to be a relatively populous court jurisdiction. Other courts in Zhejiang with considerably lighter dockets nonetheless had similar numbers of judges because they were similarly populous.Footnote ⁸

In areas such as Erqi and Xiaoshan with disproportionately heavy caseloads relative to their population sizes, an imbalance between cases and judges resulted in heavy average workloads. Judge-level caseloads by definition are determined by the number of judges and the number of cases. They are simply calculated as the number of closed cases divided by the number of judges. Critics have advocated for a system that better calibrates the supply of judges to the actual work of courts (Reference ZhangZhang 2016b:60). Across contexts of similar population size, similar numbers of judges despite wild variation and explosive growth in cases is the crux of the “many cases, few judges” problem. A widening imbalance between cases and judges has resulted in far heavier average caseloads per judge in Zhejiang than in Henan. Among all basic-level courts with available data, the average caseload per judge was 65 in Henan (26 courts) and 200 in Zhejiang (70 courts). The average volume of closed cases in Henan’s basic-level courts (3,582 cases among 57 courts with available data in 2014) was only about one-third the average annual volume of closed cases in Zhejiang (10,334 cases among 87 courts with available data in 2012–2014). And yet, among the 150 basic-level courts depicted in Figure 6.3, the average number of judges per court was nearly identical in Henan and Zhejiang (57.4 and 57.8, respectively) because the average population in their jurisdictions was also nearly identical (575,754 and 592,746, respectively).Footnote ⁹ Numbers of judges and volumes of closed cases for all courts depicted in Panel B are available with the supplementary online material at https://decoupling-book.org/.

Table 6.3 contains models that are tautological insofar as the outcome (cases per judge) is defined by two regressors (closed cases and judge counts). The key point of Table 6.3 is to demonstrate that differences between Henan and Zhejiang in average cases per judge are attributable to differences in caseloads between the two provinces. At the provincial level, Model 1 shows that judges in Zhejiang handled an average of 71 more cases than judges in Henan did in 2011. At the court level, Model 1 shows the difference to be 132 a few years later. The introduction of judge counts in Model 2 did not change anything: on its own, the number of judges was unrelated to cases per judge and did not explain away any of the gaps between Henan and Zhejiang at either the provincial level or the court level. The insignificance – and even irrelevance – of judge counts is unsurprising given that they varied so little between areas with similar populations. Cases per judge were driven not by judge counts but rather by caseloads. When per capita GDP and population size – which we know from Table 6.1 are key determinants of caseloads – are introduced in Model 3, the gap between Henan and Zhejiang shrinks and the effect of judge count becomes statistically significantly negative. Among provinces and courts with seemingly identical populations and economic conditions, those with fewer judges have more cases per judge simply because per capita GDP and population size are proxies for caseloads. For the very same reason, the effect of per capita GDP shrinks dramatically and the effect of population disappears when caseloads are added to Model 4.Footnote ¹⁰ At the same time, caseloads explain away the gap between Henan and Zhejiang in its entirety at the provincial level and almost entirely at the court level. Among provinces that were otherwise seemingly identical in terms of judge counts, population size, and economic conditions, a caseload increase of 300,000 cases (roughly the difference between Henan and Zhejiang in 2011) was associated with an average increase of 33 cases per judge (300 × .11 = 33; Model 4). Among courts that were otherwise seemingly identical in the same ways, an increase of 7,500 cases (roughly the average difference between courts in Henan and courts in Zhejiang included in the analysis) was associated with an average increase of 89 cases per judge (7.5 × 11.93 = 89; Model 4). The same patterns persisted in separate models for each province (Models 5 and 6). Next we will see that differences between Henan and Zhejiang in their use of the divorce twofer stemmed to a significant degree from differences in average caseloads per judge.

Judges’ Tendency to Grant Divorces Was Negatively Associated with Their Caseloads

The final set of regression models presented in Table 6.4 strongly suggests that China’s judicial clampdown on divorce was part of the mix of coping strategies courts adopted to deal with an acute imbalance between the supply of and demand for judges. Model 1 assesses the association between divorce rates and adjudicated divorce outcomes.Footnote ¹¹ Although these were highly correlated over time at the national level (albeit only coincidentally, Figure 6.1), they were uncorrelated at the provincial level in 2011 (Table 6.4, Model 1). A change of 1 per-mille point in the crude divorce rate – a massive change given that the crude divorce rate was 2.1‰ in 2011 – was associated with a 1.1 percentage point increase in the rate at which courts granted divorce petitions. No subsequent model includes the provincial divorce rate because it remained irrelevant even when I did include it (details omitted). In short, regional variation in divorce rates does not map onto regional variation in the extent of the judicial clampdown on divorce.

Model 2 shows that per capita GDP was strongly and negatively associated with courts’ tendency to grant divorce petitions at both provincial and court levels. Model 3 shows the gap between Henan and Zhejiang in the rate at which divorce petitions were granted: 26 percentage points at the provincial level in 2011 (all first-instance petitions) and 15 percentage points at the court level in 2009–2016 (first-attempt petitions). Subsequent models show that both the effect of per capita GDP and the gap between the two provinces were attenuated by average caseloads per judge. In Model 4, the effects of per capita GDP and the dummy variable for Zhejiang are both attenuated by their mutual presence simply because Zhejiang’s per capita GDP is so much higher than Henan’s. The effects of both variables, however, are obliterated by the introduction of average cases per judge in Model 5. The introduction of average cases per judge in Model 5 explains away both the effect of per capita GDP and the gap between Henan and Zhejiang. These results tell us that the judicial clampdown in divorce would have been a lot more similar in Henan and Zhejiang had their average caseloads per judge been the same. In other words, an important reason why divorce was so much harder to obtain in Zhejiang’s courts was because its judges were so much more overworked.

At the provincial level, an increase of 50 cases per judge was associated with a decrease of 13 percentage points in the rate at which divorce petitions were granted (0.5 × −26.07 = −13). At the court level, an increase of 100 cases per judge was associated with a decrease of 5 percentage points in the rate at which divorce petitions were granted (1 × −5.26 = −5). Because Henan’s caseloads were so much lighter than Zhejiang’s, there were no two courts in Henan that differed by 100 cases per judge. Differences of 30 between courts in Henan were more common. According to Model 6 (for Henan), an increase of 30 cases per judge was associated with a decrease of 9 percentage points in the rate at which divorce petitions were granted (0.3 × −31.33 = −9). In Zhejiang, by contrast, differences of as much as 200 between courts were not uncommon. According to Model 7 (for Zhejiang), an increase of 200 cases per judge was associated with a decrease of 9 percentage points in the rate at which divorce petitions were granted (2 × −4.74 = −9).

Figure 6.4 depicts the bivariate relationship between divorce petition grant rates and average caseloads per judge both among provincial-level units (Panel A) and courts (Panel B). Clearly visible in Panel A is a cluster formed by Zhejiang, Jiangsu, and Shanghai with the lowest rates in the country at which courts granted divorce petitions. They were among China’s provincial-level units – which also included Beijing, Chongqing, and Guangdong – with the heaviest caseloads per judge. These six places were the very same to top the list of heaviest caseloads per judge in 2019 (Reference SunSun 2019). But it was the three provincial-level units in the prosperous coastal Yangtze Delta region that formed a sort of judicial cabal by clamping down on divorce as a coping strategy. When all 31 provincial-level units are ranked according to rates at which they granted first-instance divorce petitions, Zhejiang, Jiangsu, and Shanghai occupied the bottom three spots in every year between 2002 and 2012 (Ministry of Civil Affairs of China, various years). They occupied the top three spots in a ranking of “the marketization level of each province from 2008 to 2014” (Reference LiuTang and Liu 2019:21, 26). Consequently, their judges had the heaviest caseloads and dealt with them in part by denying a considerable majority of divorce petitions.Footnote ¹² The orderly pattern depicted in Panel A gradually weakened after 2012 when courts across the country began to experience caseload spikes and as China’s domestic relations trial reforms got underway (Chapter 3). As a consequence of both processes, courts began to converge in their treatment of first-attempt divorce petitions, thus forming the nationwide clampdown depicted in Figures 6.1 and 6.2. Zhejiang, Jiangsu, and Shanghai were all early adopters of the divorce twofer as an innovative response to heavy caseloads. Over time, many provinces, including Henan, expanded the three-province judicial cabal into a national one.

Figure 6.4 Association between percentage of divorces granted and cases per judge

Note: Axes are scaled according to the natural logarithm of values. All data depicted in Panel A are from 2011. Panel B Y-axis data were aggregated from 2009 to 2016. Panel B X-axis data are from various years. On sources of closed cases per judge, see Chapter 4. Panel A N = 31. Panel B n = 95 basic-level courts (26 in Henan and 69 in Zhejiang). In Panel B, Henan r = –.47 (p = .02), Zhejiang r = –.29 (p = .02). Court codes in Panel B are listed with their corresponding court names in the supplementary online material at https://decoupling-book.org/. Panel B excludes courts of special jurisdiction and economic and technological development district courts.

Source: Court work reports; Basic Level Legal Artisan 2016a, 2016b, 2016c; author’s calculations from Henan and Zhejiang provincial high courts’ online decisions.

In Panel B, as in the regression results, the relationship between judges’ average caseloads and the rate at which divorces were granted emerges across Henan and Zhejiang combined and within each province separately. Consider, for example, the contrast between two courts in Henan: Luoyang’s Xigong District People’s Court (3c) and Xinxiang’s Fengqiu County People’s Court (7j). They closed about the same number of cases in 2014 (3,700 cases). However, because the population of Fengqiu County was about double the population of Luoyang’s Xigong District, Fengqiu’s court had about twice as many judges as Xigong’s. With an average caseload per judge almost double that of Fengqiu (107 versus 61), Xigong’s rate of granting first-attempt divorce petitions (22.5%) was less than half of Fenqiu’s (56.7%). Now consider the contrast between two courts in Zhejiang: Taizhou’s Luqiao District People’s Court (10d) and Wenzhou’s Yongjia County People’s Court (3f). Luqiao’s court closed more cases (about 12,100) than Yongjia’s (about 9,200). However, because Yongjia’s population was about double Luqiao’s, its court had considerably more judges than Luqiao’s (about 75 and 48, respectively). Consequently, Luqiao’s average caseload per judge (251) was double Yongjia’s (123). Not surprisingly, therefore, Luqiao’s rate of granting first-attempt divorce petitions (9.1%) was only a small fraction of Yongjia’s (47.1%). The supplementary online material (available at https://decoupling-book.org/) contains all numerical values in this paragraph as well as those available for every other court in the two-province sample.

We can also see in Panel B that variation between Henan and Zhejiang is greater than variation within each province. Some of the differences between the two provinces in average caseloads per judge could be an artifact of differences in how judge counts were reported. Henan’s judge counts may have encompassed all court personnel bearing the title of judge, whereas Zhejiang’s courts may have restricted counts to frontline judges who conducted trial work. Such reporting differences, which were certainly not universal, would account for only a small portion of the massive gap. More specifically, even if we make the false assumption that every judge count in Henan included all judges and every judge count in Zhejiang was limited to frontline judges, a reasonable upward adjustment to Zhejiang’s judge counts would reduce the average caseload per judge in the sample from 197 to about 150, which is still more than double Henan’s average caseload per judge of 65. To illustrate further, now consider two typical courts. The jurisdictions of Henan’s Fengqiu County People’s Court (7j) and Zhejiang’s Shengzhou Municipal People’s Court (6g) were of nearly identical population size in 2014 (733,147 and 731,200, respectively). For this reason, their judge counts were similar: 61 and 53, respectively. However, because Shengzhou’s court closed more than four times as many cases as Fengqiu’s, their average caseloads per judge were vastly different (162 and 61, respectively). Adjusting Shengzhou’s frontline judge count upward would yield an estimated total judge count of 71. Even after making such an adjustment, Shengzhou’s average caseload per judge would have still been double that of Fengqiu (121 and 61, respectively).Footnote ¹³ As we would expect, and as we can see in Figure 6.4, Shengzhou’s rate of granting first-attempt divorces (32.5%) was only a little over half that of Fengqiu (56.7%).

Figure 6.5 brings together in the form of a path model the key findings from the preceding series of regression analyses. It shows both the direct effect of average caseloads per judge and the indirect effects of judge counts, case counts, economic conditions, and general population size. And it tells us why Zhejiang’s courts clamped down harder on divorce than Henan’s courts did. Average caseloads per judge, a key determinant of the rate at which courts granted divorce petitions, were far heavier in Zhejiang than in Henan because courts in both provinces had similar numbers of judges even though Zhejiang’s courts handled far more cases than Henan’s courts did. Courts in both provinces had similar numbers of judges despite vastly different volumes of litigation because judge counts were determined primarily according to population size. Finally, Zhejiang’s higher volume of civil litigation responsible for its heavier caseloads stemmed from its more dynamic economic conditions.

Figure 6.5 Path model of factors contributing to the judicial clampdown on divorce

The remainder of this book is devoted primarily to the consequences of the divorce twofer – the dashed arrows stemming from the “judicial clampdown on divorce” box in Figure 6.5. The next section of this chapter follows the dashed arrow to the “judges’ gains” box by demonstrating several concrete ways in which the divorce twofer helped judges clear their dockets. The last section of this chapter and all subsequent chapters follow the dashed arrow to the “women’s costs” box.

The Divorce Twofer Made Cases Disappear

Figures 6.6 and 6.7 depict as pyramids the outcomes of first-instance divorce filings in court. Unlike my earlier analyses of divorce outcomes, which were limited to adjudications, the pyramids also include petitions withdrawn by plaintiffs. I limit the pyramids to decisions made after 2013 because this is when Henan’s courts consistently posted caiding decisions to grant plaintiffs’ requests to withdraw petitions. Withdrawn petitions are vastly underrepresented in prior years (Chapter 4). Recall from Chapter 1 that mediation is the first step after a divorce petition is filed. The primary purpose of mediation at this stage is to achieve marital reconciliation, which, if successful, would typically result in a petition withdrawal. Some plaintiffs withdraw their petitions at the trial stage. Judges also engage in mediation for the purpose of hammering out the terms of a divorce. Of all first-instance divorce cases handled by courts in recent years, about one-third resulted in mediation agreements. Despite the prevalence of mediation agreements in divorce litigation, they are exceedingly scarce in my samples of court decisions because courts are prohibited from posting them online (Chapter 4). Mediations are therefore not included in the pyramids. Divorces were granted in the vast majority (about 90%) of cases concluded by mediation agreements in recent years; about two-thirds of all divorces granted by courts resulted in mediation agreements and not in adjudicated verdicts. Consequently, the absence of mediation agreements skews the appearance of the pyramids in a few ways: the pyramids undercount both the total number of divorce petitions and the total number of granted divorces, and therefore misrepresent the overall rate at which courts granted divorce petitions. Had Henan’s 56,142 mediations been included in Figure 6.6, the overall rate at which courts granted divorces would have appeared as 47% in Panel A. Likewise, had Zhejiang’s 55,944 mediations been included in Figure 6.6, the overall rate at which courts granted divorces would have appeared as 48% in Panel C (Ministry of Civil Affairs, various years).

Figure 6.6 Outcomes of divorce petitions by source of data (official government statistics versus online decisions)

Note: Data are limited to first-instance decisions. Cases concluded by mediation agreements are excluded. Data are also limited to 2014–2015 for Henan and 2014–2016 for Zhejiang. Percentages do not always total 100% (and the sum of “denied” and “withdrawn” do not always equal “still married”) owing to rounding error.

Source: Ministry of Civil Affairs of China, various years; author’s calculations from Henan and Zhejiang provincial high courts’ online decisions.

Figure 6.7 Outcomes of divorce petitions by attempt (first versus subsequent)

Note: Data are limited to first-instance decisions. Cases concluded by mediation agreements are excluded. Data are also limited to 2014–2015 for Henan and 2014–2016 for Zhejiang. Percentages do not always total 100% (and the sum of “denied” and “withdrawn” do not always equal “still married”) owing to rounding error.

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

In other words, a little over half of all people who filed for divorce in court remained married after the conclusion of the process. As we can see in Figure 6.6, after excluding cases concluded by mediation agreements, about three-quarters of remaining divorce-seekers left court still married. Note the remarkable consistency between official statistics and the online court decisions in my samples – yet another indication of the representativeness of my data.

For all the reasons discussed earlier, judges would benefit from the divorce twofer even if every single petition they denied came back like a boomerang after the statutory six-month waiting period. Judges would benefit even more if only a portion of them returned. As it turns out, divorce courts were leaky. Only a small proportion of plaintiffs whose petitions were denied returned for another attempt. Some of them may have given up and stayed married, often but not always while physically separated. Others may have pursued “divorces by agreement” in the Civil Affairs Administration after “agreeing” to give up child custody, property claims, or both (see Chapter 9). According to one scholar, “the most important reason why women choose to divorce in the Civil Affairs Administration is because they want to get rid of their husbands and their insufferable marriages as quickly as possible. … For many women, giving up their rights to economic compensation is the cost of freedom from an agonizing marriage” (Reference ZhangZhang 2011:78; for the same argument, see Reference LiLi [2022] and Reference WangWang [2013:175]). To be sure, divorce cases also moved in the opposite direction. Many plaintiffs indicated in their divorce petitions that they filed for divorce in court only after their negotiations for an uncontested “divorce by agreement” in the Civil Affairs Administration had broken down. Women in both scenarios commonly discounted their claims to child custody and marital property as the price of divorce (see Chapter 10).

Official government statistics on divorce depicted in Figure 6.6 (Panels A and C) do not disaggregate first attempts from subsequent attempts. Because they include all divorce decisions, the pyramids in Figure 6.6 double-count a certain number of litigants who appear in both first-attempt and subsequent-attempt cases. They also include repeat players who are not double-counted. Some of the subsequent-attempt cases included in Figure 6.6 were spawned by prior cases that had been concluded prior to 2014. I can, of course, identify repeat players in my samples of court decisions. Among the cases represented by Panel B (for Henan) and Panel D (for Zhejiang), 14% and 21%, respectively, of the litigants were returnees.Footnote ¹⁴

Figure 6.7 showcases the docket-shrinking property of the divorce twofer. Panels A and B depict first-attempt decisions and thus contain no repeat players. Panels C and D depict subsequent-attempt decisions and thus consist exclusively of repeat players. Figure 6.7 thus disaggregates the pyramids in Figure 6.6 (Panels B and D) constructed from my samples of court decisions.Footnote ¹⁵ Subsequent attempts are not limited to second attempts. Some plaintiffs succeeded or gave up after three or even more attempts.

Two noteworthy patterns emerge from Figure 6.7. First, plaintiffs tended to remain married in first-attempt adjudications and to accomplish their mission to divorce in subsequent-attempt adjudications. This, of course, is a defining characteristic of the divorce twofer.Footnote ¹⁶ According to a study of divorce cases in a basic-level court in Zhejiang’s provincial capital of Hangzhou, 81% of plaintiffs who returned after a failed first attempt were successful on the second attempt (Reference LiuLiu 2012:81; Reference ZhangZhang 2013). Second, courts are sieves for divorce cases. Among the first-attempt petitions judges denied, most never returned to court. Courts would not be leaking divorce cases if, in Figure 6.7, the bottom bars representing subsequent-attempt divorce cases (Panels C and D) were equal to the top bars representing all the first-attempt divorce cases in which litigants failed to divorce and thus remained married (Panels A and B). Even if we exclude petitions withdrawn by plaintiffs, the number of returnees (7,186 in Henan and 12,531 in Zhejiang) pales in comparison to the number of plaintiffs whose petitions had been denied in first-attempt adjudications (18,572 in Henan and 23,593 in Zhejiang). These numbers make sense only if a sizeable share of petitioners never returned to court.Footnote ¹⁷ Indeed, according to a study of divorce cases in a basic-level court in Zhejiang’s provincial capital of Hangzhou, only 22% of plaintiffs whose first-attempt petitions were denied returned to file a second first-instance petition (Reference LiuLiu 2012:81; Reference ZhangZhang 2013).

Table 6.5 shows once again that courts denied the majority of first-attempt divorce petitions and granted the majority of subsequent-attempt divorce petitions.Footnote ¹⁸ More importantly, it shows the uniquely low rates at which first-attempt divorce petitions were granted, not only compared to subsequent-attempt divorce petitions, but also compared to other types of civil cases. Successful first-attempt divorce cases were truly exceptional. Courts in both provinces granted plaintiffs’ petitions in first-attempt divorce cases at rates that were less than half of the win rates enjoyed by plaintiffs in every other type of case except administrative litigation.Footnote ¹⁹

First-attempt divorce trials were also exceptional in their sparing use of collegial panels. As low-hanging fruit for quick-and-easy adjudicated denials using the simplified procedure, first-attempt divorce petitions have helped courts to economize on scarce judicial resources. Simplified procedure utilization rates in Table 6.5 suggest that, among all the types of cases they handled, courts devoted the fewest judicial resources to first-attempt divorce cases. By contrast, contract disputes in both provinces were much more likely to be tried according to the ordinary civil procedure. Compared to how they treated divorce cases, courts attached far more importance to contract disputes. As we will see shortly, contract disputes happened to be filed overwhelmingly by men, whereas, as we already know, divorces were filed overwhelmingly by women.

Subsequent-attempt divorce trials more closely resembled the civil trials of other types of cases in terms of both win rates and the use of the simplified procedure. Note also that in Table 6.5 administrative litigation cases were practically never tried according to the simplified procedure. The original 1989 Administrative Litigation Law required the use of collegial panels (Article 46). Only after the amended version of the law took effect on July 1, 2017, at the very end of the period of analysis, were solo judges permitted to apply the simplified procedure in administrative litigation cases (Article 83).

The Divorce Twofer Reduced Judge Effort

The divorce twofer was a docket-shrinking machine not only by making cases disappear. By allowing judges to deny so many first-instance divorce petitions in a relatively perfunctory manner, the divorce twofer also reduced the effort they put into first-instance divorce trials, and thus freed up time and effort for them to deal with other kinds of cases they deemed more important.

Table 6.6 contains mean and median durations of time from when civil cases were filed to when judges rendered adjudicated verdicts. Note the generally high degree of consistency between means and medians, suggesting that time distributions are not overly skewed one way or the other. I disaggregate times to close cases by type of procedure (ordinary versus simplified) because of Zhejiang’s dramatically higher simplified procedure utilization rates (Chapter 5). Overall case closing times were slower in Henan (in the “total” column) only because its simplified procedure utilization rates were so much lower. Across the two provincial samples, case closing times among cases of the same type tried according to the same procedure were generally similar.

Thanks to the divorce twofer, judges in both provinces closed first-attempt divorce cases faster than they closed any other type of civil case (according to means/medians in the “total” column). Once again, Zhejiang’s relatively enthusiastic embrace of the divorce twofer is evident in its remarkably swift adjudicated denials. The mean/median time from initial case filing to adjudicated denial using the simplified procedure was 52/51 days in Henan and 37/33 days in Zhejiang. Among all first-attempt divorce cases tried according to the ordinary procedure, decisions to deny petitions were about 20 days faster in Henan and, strangely, a few days slower in Zhejiang than decisions to grant petitions.Footnote ²⁰ Compared to divorces granted using the ordinary procedure, first-attempt denials using the simplified procedure cut the duration of time to an adjudicated decision by over one-half (about 60 days) in Henan and by two-thirds (about 80 days) in Zhejiang. Finally, among all first-attempt divorce cases tried according to the simplified procedure, the mean/median amount of time saved by denying petitions compared to granting them was one week per case in Henan and about two weeks per case in Zhejiang.

The aggregate time savings from the divorce twofer has therefore been immense in both provinces. Multiplying mean or median savings in time gained from denying first-attempt divorce petitions by the volume of adjudicated denials of first-attempt divorce petitions in the two samples yields millions of days in total time savings. Of course, the time necessary to adjudicate subsequent-attempt divorce petitions spawned by first-attempt denials must be subtracted from this amount of time savings. However, as we know, most first-attempt adjudicated denials never come back. Those that do return are generally both less burdensome, and thus benefit judges’ volume and efficiency scores, and are at lower risk of leading to social unrest or “extreme incidents” that would hurt judges’ performance evaluations (Chapter 3).

To any estimate of the divorce twofer’s total net time savings must be added the amount of time additional judges and lay assessors – whose efforts were spared by the solo judges who single-handedly denied first-attempt divorce petitions using the simplified procedure – would have otherwise spent as members of collegial panels in trials granting first-attempt divorces. Needless to say, judges’ labor accounts for only a small portion of the duration of time from case filing to adjudicated decision. We do not know the true time savings gained by the divorce twofer, because we do not know how much time judges actually devoted to cases. We know only when cases were filed and when decisions were finalized; the court decisions do not reveal exactly how judges allocated their time in the interim. Surely, though, the divorce twofer benefitted judges not only by conserving a real measure of their scarce supply of time and effort but also by increasing their simplified procedure utilization rates, shortening their case closing times, and thereby boosting their performance evaluation scores.

The relative brevity of judges’ written decisions in first-attempt divorce adjudications further suggests that the amount of time and effort saved was substantial. Table 6.7 contains mean/median numbers of characters – almost all Chinese, of course, but also some Latin letters, Arabic numerals, and symbols – per court decision. I disaggregate length of court decision by type of procedure for the same reason I disaggregated time to decision in the previous analysis. The written decisions of cases closed using the ordinary procedure were generally longer than those of cases closed using the simplified procedure. Regardless of how judges ruled and which procedure they used, their first-attempt divorce decisions were shorter than those of any other type of civil case. In particular, their decisions to deny first-attempt divorce petitions were shorter than their decisions in any other type of case. In both provinces, written decisions that denied first-attempt divorce petitions were generally about 20–30% shorter than those that granted first-attempt divorces. Not only were adjudicated denials shorter in length, but, as we will see in Chapter 7, they also contained a considerable amount of recycled and rehashed boilerplate text justifying marital preservation in therapeutic, moral, and ideological terms. Although subsequent-attempt divorce decisions were longer than first-attempt divorce decisions, much of the text they contained was redundant. Because many of the facts do not change – or require only minor revisions – between attempts, judges simply copy and paste portions of their first-attempt decisions into their subsequent-attempt decisions.

The Divorce Twofer Disproportionately Impacted Women

When judges sacrificed divorce cases in response to their heavy caseloads, they did not do so in a gender-neutral manner. Women were disproportionate casualties of the divorce twofer. Table 6.8 shows female representation among plaintiffs and defendants by case type. Among plaintiffs, women were concentrated in divorce cases. In Henan, divorce cases accounted for 17% of all plaintiffs but for 29% of all female plaintiffs in first-instance civil adjudications. In Zhejiang, divorce cases accounted for 5% of all plaintiffs but for 11% of all female plaintiffs in first-instance civil adjudications. Women also went to court in large numbers as both plaintiffs and defendants in “other marriage and family cases,” which consisted primarily of disputes related to child custody, inheritance, eldercare, adoption, and relationship breakups among unmarried couples (e.g., betrothal gifts among couples who called off their marriage engagement as well as property division and child custody among cohabiting couples who split up). Commensurate with their overrepresentation among plaintiffs filing for divorce, women were underrepresented among plaintiffs filing for relief in contract, tort, and administrative disputes. These patterns provide further evidence that judges attached greater importance to and took more seriously cases filed by men than cases filed by women. Drawing on and reinforcing cultural stereotypes, judges could more easily justify the divorce twofer by characterizing the divorce petitions of women as “frivolous” and “impulsive,” and therefore without legal merit.

Moreover, the divorce twofer disproportionately impacted rural women. Not only were divorce cases concentrated in rural courts, but women’s representation among plaintiffs filing for divorce was greater in rural areas than in urban areas (Chapter 4). Because contract disputes were overrepresented in urban courts, judges in rural courts dealt with women seeking divorce on a more regular basis than their urban counterparts did. As we will see throughout this book, women’s vastly poorer outcomes in rural courts likely stemmed at least in part from the greater durability of patriarchal cultural beliefs in rural areas.