Introduction

A balance between crime and punishment is a fundamental guarantee of fairness and justice. Such a balance requires a systematic ladder of punishment to determine how criminals should be punished. Cesare Beccaria, who developed ladder theory, made the following comment: “If mathematical calculation could be applied to the obscure and infinite combinations of human actions, there might be a corresponding scale of punishments, descending from the greatest to the least” (Beccaria 1995). Accordingly, punishment ladders must be used to allocate and apply penalties of varying levels of severity to crimes of different levels of seriousness, thus establishing a balance between crime and punishment. In response to criminal activity, the ladder of punishments should start with less severe punishments (e.g., fines) and gradually progress to more severe punishments (e.g., probation or incarceration) for defendants who commit more or more severe crimes (Lappi-Seppälä 2016; Hinkkanen and Lappi-Seppälä 2011).

Hence, developing and implementing a punishment ladder is an important way of establishing a balance between crime and punishment. Researchers have engaged in a series of discussions concerning the punishment ladder. Several papers have examined the effectiveness of a particular rung (e.g., prison sentences) on the ladder (e.g., Kling 2006; Green and Winik 2010; Aizer and Doyle 2015; Eren and Mocan 2017; Bhuller et al. 2020). Other scholars have investigated the impacts of incarceration on defendant outcomes at both the extensive and intensive margins (Owens 2009; Green and Winik 2010; Kuziemko 2013; Barbarino and Mastrobuoni 2014; Aizer and Doyle 2015; Harding et al. 2018; Dobbie et al. 2018), and these researchers have reported a variety of useful conclusions. In China, scholars have focused mainly on the legislative level and analysed issues pertaining to the establishment of such a ladder. Some researchers have discussed the issue of defining crime gradients on the basis of crime assessment and explored how to construct a sequence of penalties ranging from mild to severe (Zhou 2007; Chen 2019). In addition, other scholars have focused on the judicial level, including by exploring the correspondence between the extent of the crime and the severity of the penalty (Lin 2014; Chen 2020).

In summary, although scholars have engaged in a series of beneficial explorations of the punishment ladder, most such researchers have focused on only one rung of the ladder. Judicial research in China has focused only on the correspondence between circumstances and penalties, thereby overlooking the fundamental issue of whether the punishment ladder is reasonable. In fact, the balance between crime and punishment is not merely a principle of criminal legislation but also a guideline for criminal justice. Only through effective judicial activities can a balance between crime and punishment truly be established. Therefore, the tasks of identifying and addressing issues pertaining to the punishment ladder in judicial practice are highly important.

This paper explores the rationality of the punishment ladder used in this context on the basis of the judicial implementation of this punishment ladder, specifically with the goal of developing a more reasonable version of the punishment ladder. In the legislation that has been implemented in China, the design of statutory penalties reflects the punishment ladder. Criminal legislation stipulates different types and ranges of penalties for various crimes in accordance with their severity. With respect to the numerous crimes included in China’s criminal law, we select the crime of infringing upon citizens’ personal information for this study in light of its representativeness, which is reflected in two main aspects.

The first such aspect is the clarity of the punishment ladder. The Ninth Amendment to the Criminal Law stipulates clear statutory penalty levels for the crime of infringing upon citizens’ personal information, which is divided into two penalty levels: for serious circumstances, the offender shall be sentenced to fixed-term imprisonment or criminal detention of up to 3 years, whereas for particularly serious circumstances, the offender shall be sentenced to fixed-term imprisonment of 3–7 years. On the basis of these two statutory penalty levels, the punishment ladder for the crime of infringing upon citizens’ personal information was officially established.

The second aspect is a clear correspondence between the circumstances of the crime and the punishment ladder. In crimes for which the resulting social harm can be quantified, quantity is the most important condition for conviction. Consider as an example the crime of infringing upon citizens’ personal information; in this context, the correspondence between the circumstances of the crime and the punishment ladder is determined on the basis of the quantity of information involved.Footnote 1 Regarding the amount of information that corresponds to different crime circumstances, the judicial interpretation issued by the Supreme People’s Court and the Supreme People’s Procuratorate in 2017 (hereinafter referred to as “Two High Court Judicial Interpretations”) established clear standards (Table 1).Footnote 2

Table 1 Standards for the quantity of information in different criminal circumstances pertaining to the crime of infringing upon citizens’ personal informationa.
Full size table

Specifically, illegally obtaining, selling, or providing more than 50 pieces of sensitive information, more than 500 pieces of important information, more than 5000 pieces of general information, or a combination thereof totalling more than 5000 pieces of information, are viewed as “serious” circumstances; 10 times the amount of information specified for “serious circumstances” is considered to constitute “particularly serious circumstances”.Footnote 3 According to the Criminal Law, different severities of circumstances correspond to different ranges of legal penalties. This approach establishes a link between the circumstances of the crime and the punishment ladder, thus allowing the severity of the circumstances to be determined on the basis of the amount of information involved and thereby to determine the corresponding punishment ladder.

Therefore, this paper takes the crime of infringing upon citizens’ personal information as an example and examines the rationality of the punishment ladder by reference to a combination of judicial judgement data with empirical research methods such as regression discontinuity design (RDD). This empirical research aims to identify a reasonable relationship between crime and punishment and to develop a more rational punishment ladder. The empirical results reveal that after the judicial interpretation in 2017 defined the amounts of information that corresponded to different circumstances, a significant discontinuity emerged at the threshold of the amount of information for fixed-term imprisonment sentences. This finding indicates that the balance between crime and punishment within different statutory penalty ranges is relatively good but also that the level of connectivity among different punishment ladders is insufficient. Cases near the threshold of the amount of information are not associated with punishments that correspond with the severity of the crime, thus standing in contrast to the principle of a balance between crime and punishment. In addition, the greater the amount of information is, the lower the penalty per unit of information. This finding reveals that the threshold has a strong deterrent effect; however, once this threshold is crossed, this deterrent effect becomes significantly weaker, thus leading to the failure of efforts to establish a progressively increasing deterrent ladder. This situation requires weakening the decisive role of the threshold to be weakened, such that the amounts of information that correspond to different degrees of severity overlap, thereby improving the fairness of punishment. Regarding the issue of the gradually decreasing penalty per unit of information beyond the threshold, the penalties associated with larger quantities of information should be paired with more severe fines to achieve the purpose of punishment.

The marginal contributions of this paper are as follows. First, previous research on the punishment ladder has failed to examine whether the punishment ladder is reasonable, and our research on the rationality of the punishment ladder can thus enrich the framework for research on the balance between crime and punishment. Second, in response to the current situation of judicial research in China, which emphasizes theoretical research, lacks empirical verification, and focuses on normative research, thereby overlooking the need for empirical analysis, we test the rationality of the punishment ladder employed in this context through the use of judicial judgement data and empirical methods such as regression discontinuity design (RDD). This research thus provides empirical support for legislative improvement and fair justice. Third, by using the crime of infringing upon citizens’ personal information as an example in this investigation of the rationality of the punishment ladder and by offering valuable suggestions with respect to the legislative improvement and judicial application of that ladder, our study can improve the fairness of the judiciary in cases involving infringements upon citizens’ personal information.

The structure of this paper is as follows. First, this paper introduces China’s judicial system and legislation pertaining to the crime of infringing upon citizens’ personal information. Second, we present the data sources and research methods used in this study. Third, we present the empirical results of this research and a heterogeneity analysis. Finally, we present a further empirical analysis and discuss recommendations for ways in which the punishment ladder can be implemented rationally.

China’s judicial system and legislation pertaining to the crime of infringing upon citizens’ personal information

China’s judicial system

In China, the people’s courts represent the national judiciary, and the organizational system used in this context includes the Supreme People’s Court, local people’s courts operating at various levels and special people’s courts. The Supreme People’s Court is the highest judicial organ of the state, which exercises the highest judicial power in accordance with the law while simultaneously supervising the work of local people’s courts at all levels and that of specialized people’s courts. According to the People’s Republic of China’s Judges Law, judges are judicial personnel who exercise the judicial authority of the state according to law, including presidents, vice presidents, members of judicial committees, chief judges and associate chief judges of divisions, judges and assistant judges of the Supreme People’s Court, local people’s courts at various levels and special people’s courts such as military courts.

A judge shall be appointed or removed from a post in accordance with the limits of authority pertaining to and procedures involved in appointment or removal as prescribed by the Constitution and relevant laws. The President of the Supreme People’s Court shall be elected or removed by the National People’s Congress. The vice-presidents, members of the judicial committee, chief judges and associate chief judges of divisions and judges shall be appointed or removed by the standing committee of the National People’s Congress upon the recommendation of the President of the Supreme People’s Court. The presidents of local people’s courts at various levels shall be elected or removed by local people’s congresses at various levels. The vice-presidents, members of the judicial committees, chief judges and associate chief judges of divisions and judges shall be appointed or removed by the standing committees of the people’s congresses at the corresponding levels upon the recommendation of the presidents of those courts. The term of office of the presidents of the people’s courts at all levels shall be the same as the term of office of the people’s congresses at that level, in which context each term shall last five years. A people’s court shall establish a commission to examine and assess judges, which shall be responsible for the appraisal of the judges associated with the court in question. The result of this appraisal shall be taken as a basis for awards, punishments, training, and dismissals pertaining to judges as well as for readjustments of judges’ grades or salaries.

Sentencing is a activity pertaining to trials, in which context the court decides whether to impose a sentence on a defendant, what kind of sentence to impose, what degree of sentence to impose, and whether to execute the sentence immediately on the basis of a conviction, according to the severity of the criminal act in question and criminal responsibility for that act (Wang et al. 2022). According to the sentencing guidelines employed in China, sentencing should be based on facts and the law, and a sentence should be determined according to the facts of the crime, the nature of the crime, the circumstances of the crime and the degree of harm caused to society. Sentencing is based first on the basic facts of the crime within the corresponding legal penalty range, which helps relevant actors determine the starting point for sentencing. Second, on the basis of other facts pertaining to the crime, such as the amount associated with the crime, the number of times the crime was committed and the consequences of the crime that affect the composition of the crime, the severity of the punishment is increased on the basis of the starting point, thus determining the baseline sentence. Finally, the baseline sentence is adjusted in accordance with the circumstances of the sentencing and, in light of the overall circumstances of the case, the sentence is determined in accordance with the law. In situations involving mitigating circumstances, such as surrender, guilty pleas and active compensation, the sentence may be set at a level below the statutory minimum after strict approval and authorization are received.

Legislation pertaining to the crime of infringing upon citizens’ personal information

With respect to the definition of citizens’ personal information, Article 253-1 of the Criminal Law clearly stipulates that citizens’ personal information refers to all kinds of information that is recorded by electronic or other means that can, either alone or in combination with other information, identify a specific natural person or reflect the activities of a specific natural person, including name, identity document number, means of communication, address, account number, password, property status, whereabouts and tracking information, and other types of information.

The criminal law protection system used for citizens’ personal information in China has gradually evolved through successive amendments to relevant criminal law. In 1997, when the Criminal Law was first promulgated, the term “citizens’ personal information” did not appear, and it was necessary for infringements of citizens’ personal information to be addressed by reference to other crimes, such as the infringement of commercial secrets or the infringement of freedom of communication. The introduction of the Criminal Law Amendment (V) Law in 2005 brought “citizens’ personal information” to the attention of criminal law, and the crime of stealing, buying or illegally providing credit card information was added as a crime in Article 177 of the Criminal Law; furthermore, credit card information and data pertaining to citizens’ personal information were brought into the scope of special protection associated with the Criminal Law. In 2009, the Criminal Law Amendment (VII) Act was introduced to fill the gap in China’s criminal law pertaining to the crime of infringing upon citizens’ personal information; “citizens’ personal information” was explicitly included in relevant criminal law provisions, and the crime of selling or illegally providing citizens’ personal information and the crime of illegally obtaining citizens’ personal information were added as crimes according to Article 253. The Ninth Amendment to the Criminal Law, which was introduced in 2015, combined the crimes of selling or illegally providing citizens’ personal information and illegally obtaining citizens’ personal information into the crime of infringing upon citizens’ personal information, and the level of sentencing associated with this crime was refined. Article 253 of the Criminal Law stipulates that anyone who sells or provides citizens’ personal information to others in violation of relevant state regulations under serious circumstances shall be sentenced to fixed-term imprisonment of not more than three years or criminal detention and that such an individual shall be fined either in addition to this sentence or in isolation; if the circumstances are particularly serious, such a person shall be sentenced to fixed-term imprisonment of not less than three years and not more than seven years and shall be fined in addition to this sentence.

On this basis, the Supreme People’s Court and the Supreme People’s Procuratorate jointly promulgated the Interpretation on Several Issues Concerning the Application of Law in Handling Criminal Cases of Infringing on Citizens’ Personal Information (hereinafter referred to as the Two High Court Judicial Interpretation) in 2017; this document provides clear and detailed provisions concerning the understanding and application of the crime of infringing upon citizens’ personal information.

First, personal information is categorized on the basis of a three-part method; specifically, such information is divided into sensitive information, important information and general information. In this context, whereabouts information, communication content, credit information, and property information represent sensitive information; accommodation information, communication records, health and physiological information, transaction information, and other personal information of citizens that may affect the safety of persons and property represent important information; and other information represents general information.

Second, different starting points for criminalization are provided for different types and levels of information. According to paragraphs 3–7 of Article 5 of the Two High Court Judicial Interpretations, the following quantities of information are considered to constitute “serious circumstances”: illegally obtaining, selling or providing more than 50 pieces of information concerning citizens’ whereabouts and tracking information, content pertaining to communication, credit information or property information (i.e., sensitive information); illegally obtaining, selling or providing more than 500 pieces of citizens’ personal information (i.e., important information), such as accommodation information, communication records, health and physiological information, and transaction information, which may affect the safety of persons and property; and illegally obtaining, selling or providing more than 5000 pieces of citizens’ personal information (general information) other than those stipulated in the third and fourth subparagraphs. Furthermore, the quantity in question must not meet the standards stipulated in the third to fifth subparagraphs, but it must meet the relevant quantitative standards in accordance with the corresponding proportion. When the quantity reaches more than ten times the standard for “serious circumstances”, the crime is considered to have been committed in “particularly serious circumstances”, i.e., more than 50,000 articles in total according to the corresponding proportion. Accordingly, this paper is weighted according to the type and quantity of such information, and the quantities of information used in this context represent comparable values after weighting.

Third, the criteria for crimes that are committed in the performance of individuals’ duties or the process of providing services are defined. Article 5, paragraph 8, of the Two High Court Judicial Interpretations stipulates that the sale or provision to others of the personal information of citizens that is obtained during the process of performing duties or providing services, specifically in quantities or in amounts that reach a level of more than half of the criteria stipulated in paragraphs 3–7, shall be identified as “particularly serious circumstances”. Accordingly, this article extracts information concerning whether the defendant obtained the citizen’s personal information during the performance of his or her duties or while providing services and, if this is the case, weights the amount of information to ensure that it is comparable to other general cases.

The degree of seriousness is closely related to the number of pieces of personal information involved in the infringement. This point raises a key question: what kind of information can be recognized as a piece of “citizens’ personal information”? In both academia and judicial practice, a general consensus indicates that a set of information that pertains to individual citizens is usually recognized as a piece of personal information. The distinctive feature of citizens’ personal information is that it is recognizable, and each piece of citizens’ personal information has a corresponding information subject. The personal information of citizens involved in the infringement committed by the perpetrator is usually counted according to the number of “individuals” involved (Li and Wang 2019). This view has also been recognized in judicial practice, and it is reasonable in judicial practice to identify a set of information pertaining to an individual citizen as a piece of personal informationFootnote 4.

Data sources and research methods

Data source and processing

The data used in this research were obtained from the China Judgements Online Database (https://wenshu.court.gov.cn), and the samples included judgements pertaining to the crime of infringing upon citizens’ personal information. After we obtained the raw data, we first performed data pre-processing, including through deduplication (i.e., deleting duplicate entries of the same case) and filtering for eligible judgements (i.e., excluding cases in which the sentencing focused on other charges and cases for which no full judgement texts were available). For data analysis, only judgement texts were retained, resulting in 8124 documents available for the analysis (as of February 23, 2021).

Note that for judgements that involved multiple defendants, previous empirical studies have often chosen to exclude these documents from their analyses with the aim of reducing extraction errors. Given that cases that involve multiple defendants account for a significant proportion (approximately 54%Footnote 5) of crimes involving infringements upon citizens’ personal information, to avoid substantial sample loss, we listed multiple defendants separately in the documents, in which context each defendant corresponded to one data entry. The final sample size was determined by reference to the number of defendants, resulting in a total of 15,981 data points in the sample.

Figure 1 illustrates the trend in the number of judgements pertaining to the crime of infringing upon citizens’ personal information and the number of defendants over time. Notably, prior to 2015, the crime of infringing upon citizens’ personal information was divided into two crimes in China, namely, the crime of selling or illegally providing citizens’ personal information and the crime of illegally obtaining citizens’ personal information. Thus, data concerning periods prior to 2015 were retrieved in terms of these two crimes. Given that the number of judgements may be limited by the timing of their upload or disclosure, a discrepancy might occur between the actual number of cases that occurred after 2020 and the statistical data. However, overall, the number of cases involving the crime of infringing upon citizens’ personal information in China is characterized by a clear growth trend.

Fig. 1: Data trends.
figure 1

Trends of judgements and defendants pertaining to the crime of infringing upon citizens’ personal informationFootnote

Data source: China Judgements Online (https://wenshu.court.gov.cn).

.

Full size image

We process the data in further. Given that the core explanatory variable of this paper is the number of defendants who violated citizens’ personal information, some documents that did not mention the number of pieces of information were excluded, and extreme values (i.e., cases in which the number of pieces of information is greater than 10 million information and the amount of profit is greater than 10 million yuan) were excluded, resulting in a final sample consisting of 9037 defendants.

Moreover, the accuracy of variable extraction was tested. Random samples of documents, which accounted for 1% of the total, were checked manually to calculate the accuracy rate, which was no less than 92%Footnote 7. For large-scale text information extraction, this result is satisfactory.

Variable selection and analysis

This paper aims to investigate whether the penalty ladder currently in use is reasonable, specifically with respect to the possibility of a break point in the penalty amount at the threshold of information quantity as a result of different degrees of severity in terms of the circumstances. The dependent variable is the penalty assigned for the crime of infringing upon citizens’ personal information, including the duration of the fixed-term imprisonment and the amount of the fine associated with this crime.

The core explanatory variables are the quantity of citizens’ personal information of citizens involved in the infringement and whether this quantity exceeds 50,000. The threshold between serious and particularly serious circumstances is 50,000 pieces of information, on which basis we can determine whether the penalty amount changes smoothly at the threshold, thus verifying the establishment of a balance between crimes and penalties.

Given the difference in the standards for quantity corresponding to various information types, we calculate the number of pieces of information according to the Two High Court Judicial Interpretations.Footnote 8 The calculation process is as follows:

$$\begin{array}{ll}{\rm{Quantity}}\;{\rm{of}}\;{\rm{information}}={\rm{Quantity}}\;{\rm{of}}\;{\rm{sensitive}}\;{\rm{information}}* 100\\\qquad\qquad\qquad\qquad\qquad\qquad+{\rm{Quantity}}\;{\rm{of}}\;{\rm{important}}\;{\rm{information}}* 10\\\qquad\qquad\qquad\qquad\qquad\qquad+{\rm{Quantity}}\;{\rm{of}}\;{\rm{general}}\;{\rm{information}}\end{array}$$

In addition, the Two High Court Judicial Interpretations specify quantity standards for obtaining citizens’ personal information in the process of performing duties or providing services, and we treat such cases accordingly.Footnote 9

The control variables included in this research are circumstances that may affect sentencing, which mainly include the following:

  1. (1)

    The amount of profit (logarithm), tribunal size, principal criminal, accessary criminal, first offender, surrender, admittance of guilt and acceptance of punishment, and presence of a defender. With the exception of the amount of profit, all these variables are dummy variables ranging from 0 to 1, where 1 indicates the occurrence of the event.

  2. (2)

    Since the sample investigated in this paper is based on defendants, it controls for the demographic characteristics of those defendants, including gender, educational attainment (in years)Footnote 10, and unemployed status (in which context unemployed defendants take a value of 1).

  3. (3)

    The number of pieces of information involved in this crime and the amount of profit are also affected by the level of economic development and the level of technological development in the region in question. We include the logarithm of per capita gross domestic product (GDP) and the logarithm of research and development (R&D) expenditure to control for the impacts of these macro factors; the corresponding data are drawn from the China Statistical Yearbook. Table 2 presents descriptive statistics concerning the variables.

    Table 2 Descriptive statistics of the study variables.
    Full size table

Establishment of the empirical model

This paper aims to investigate whether the punishment ladder currently used for the crime of infringing upon citizens’ personal information leads to discontinuity at the threshold of information quantity, thereby testing the rationality of this use of the punishment ladder. We rely on the discontinuity rule that categorizes an amount that is greater than 50,000 pieces of information as “particularly serious” and employ regression discontinuity design (RDD) to test this issue. RDD, a quasinatural empirical method that closely resembles a natural experiment, can control for unobservable confounding factors surrounding the threshold by restricting the bandwidth, thereby effectively mitigating the endogeneity problem resulting from omitted variables. In addition, the functional form of the forcing variable can be adjusted with the aim of eliminating potential sources of bias, thereby ensuring effective causal inference (Chen et al. 2013).

The estimation methods associated with RDD generally include both nonparametric and parametric estimations. Nonparametric estimation exhibits a low level of dependence on the form of functions but requires many observable values at the break point (Imbens and Lemieux 2007). Parametric estimation is relatively easy to perform but is susceptible to fitting bias. In light of the number of observations and the availability of data, this paper uses a semiparametric method for estimation (Ito and Zhang 2020), thus restricting the sample within a band surrounding the information quantity threshold. This method not only accounts for the loss of sample size but also restricts the fitting interval, thereby producing more consistent regression results.

With respect to the issue under investigation, the following polynomial RDD model is established:

$${{Above}}_{i}=\left\{\begin{array}{c}1,{{Num}}_{i}\ge 50000\\ 0,{{Num}}_{i} < 50000\end{array}\right.$$
(1)
$${{Punish}}_{i}=\alpha +{\beta }_{1}{{Above}}_{i}+{\beta }_{2}f\left({{Num}}_{i}\right)+\Gamma X+{\rho }_{p}+{\delta }_{t}+{\varepsilon }_{i}$$
(2)

where \({{Punish}}_{i}\) is the penalty, including fixed-term imprisonment and fine amounts, that is imposed on defendant \({i;}\) \({{Above}}_{i}\) indicates whether the number of pieces of information involved in the infringement committed by defendant \(i\) exceeds the threshold, which is the critical point of 50,000 pieces of information for “serious” and “particularly serious” circumstances; \(f\left({{Num}}_{i}\right)\) is a polynomial function, in which context the explanatory variable is the number of pieces of information \({{Num}}_{i}\) involved in the infringement committed by defendant \(i\), which is also known as the forcing variable; \(X\) includes control variables that affect sentencing, demographic characteristics, and macro factors; \({\rho }_{p}\) represents province fixed effects; \({\delta }_{t}\) represents year fixed effects; and \({\varepsilon }_{i}\) is the random error term.

From Eq. (2), the paper focuses primarily on the coefficient \({\beta }_{1}\) of \({{Above}}_{i}\), which captures the local average treatment effect (LATE) for defendants whose information quantities exceed 50,000 pieces of information relative to those whose information quantities are less than 50,000 pieces of information at the threshold. According to the principle of RDD, if the observations near the break point are random, then the break point variable should affect only the dependent variable, namely, without impacting other explanatory variables. Thus, when the assumption of randomness holds, the addition or subtraction of control variables should not significantly affect the results of the RDD. However, to overcome the endogeneity problem resulting from omitted variables to the greatest extent possible, we include additional control variables in the model. In addition, since the RDD results are highly sensitive to the functional form of the model, different polynomial degrees are used in the baseline regression.

Empirical findings and heterogeneity analysis

Baseline regression results: fixed-term imprisonment sentences exhibit a discontinuity at the threshold

In RDD, different polynomial degrees are first used to estimate the baseline model pertaining to fixed-term imprisonment sentences and fine amounts. Plotting can facilitate an intuitive analysis of the differences on both sides of the break point. In Fig. 2, the points on the left and right sides of the break point represent the fixed-term imprisonment sentences for defendants whose information quantities are below and above the threshold, respectively, thus indicating that defendants whose infringements involved more than 50,000 pieces of personal information receive longer fixed-term imprisonment sentences than do defendants whose infringements involved fewer than 50,000 pieces of information.

Fig. 2: Scatter plot and RD plots.
figure 2

Scatter plot and RD plots of the amount of information versus the duration of the imprisonment sentence under linear fitting, quadratic fitting and cubic fitting.

Full size image

Table 3 presents the estimation results of the RDD analysis pertaining to Eq. (2), which were obtained on the basis of a semiparametric estimation method. Different bandwidths correspond to different RDD results. In Table 3, the bandwidth is including a range of plus or minus 10,000 pieces of information from the threshold. Subsequent analyses use other bandwidths for robustness checks. According to the Two High Court Judicial Interpretations, the threshold for the quantity of citizens’ personal information was implemented beginning on June 1, 2017; therefore, the sample interval for the baseline regression is restricted to the period after June 1, 2017.

Table 3 Number of pieces of information and term of imprisonment/amount of fine: Polynomials.
Full size table

Columns (1)–(3) of Table 3 present the RDD results on the basis of linear, quadratic, and cubic polynomial fits, respectively. According to the estimates presented in the first three columns, the treatment effect of infringements involving the personal information of more than 50,000 citizens on fixed-term imprisonment sentences ranges between 8.374 and 8.425; all the effects are significant at the 5% level. This finding indicates a significant discontinuity at the threshold of 50,000 pieces of information, thus suggesting that defendants whose infringements involved more than 50,000 pieces of personal information face significantly longer fixed-term imprisonment sentences than do those whose infringements involved fewer than 50,000 pieces of information.

The amount of the fine does not constitute a significant break point at the threshold

This paper also investigates whether the fine sentence amount is characterized by a break point at the information quantity cut-off value, in which context we use Eq. (2) to estimate the break point of the fine sentence amount; the results are presented in Columns (4)–(6) of Table 3.Footnote 11 Regarding the amount of the corresponding fine, the RDD analysis does not reveal a significant discontinuity at the threshold of information quantity. The treatment effect on fine amounts ranges between −11.990 and −10.080 across different polynomial fits, but these effects are not statistically significant, even after we control for provincial and yearly fixed effects. This finding indicates that the fine amounts do not exhibit a significant discontinuity at the 50,000-information threshold, thus suggesting a smoother transition in fine amounts across different quantities of information subject to infringement. We also perform robustness tests (see Appendix). The results reveal that consistent conclusions can still be drawn.

Empirical findings: a better crime-penalty balance within the statutory penalty interval but insufficient consistency between different penalty ladders

The empirical results of this research reveal that after the implementation of the Two High Court Judicial Interpretations in 2017, a significant increase in the term of imprisonment sentences occurred at the quantitative cut-off value of 50,000, which resulted in a better balance the crime and the penalty within the different statutory penalty intervals but insufficient consistency among different penalty ladders. This finding suggests that for cases near the cut-off value, the appropriateness of the penalty to the circumstances of the crime is not ensured, and the principle of balanced penalties is not followed in this context. This approach leads to a situation in which the marginal amount is decisive, and the amount of information near the cut-off value may be divided into different legal penalty bands with a difference of only ten pieces (or even one piece) of information.

The first principle of an ideal crime‒penalty relationship is the equalization of crimes and penalties. The basic meaning underlying the balance between crime and punishment is that the degree of social harm caused by the crime in question is an important basis for the determination of the severity of the penalty, and the duration of the associated sentence corresponds to the severity of the crime, resulting in heavy punishments for serious crimes and light punishments for light crimes. The nature of the penalty should be appropriate with respect to the nature of the crime. If the relationship between crime and punishment is displayed in a two-dimensional coordinate system, it should be presented as a monotonically increasing function without segments. According to the principle of a balance between crime and punishment, with respect to the crime of infringements involving citizens’ personal information, the punishment, including the term of imprisonment and the amount of the corresponding fine, should be appropriate to the circumstances of the crime. The quantity of citizens’ personal information involved in the infringement, as the most important circumstance surrounding the crime, should ensure that the punishment and the quantity of information are appropriate.

Feasibility analysis of regression discontinuity: continuity test for regression discontinuity

Use of the RDD approach requires that two prerequisite assumptions must be fulfilled: first, with the exception of the selected treatment variable, all other related factors must change smoothly at the threshold to avoid capturing discontinuity effects resulting from other factors, and second, for the situation does not involve the potential manipulation of the forcing variable.

Regarding the first assumption, Fig. 3 illustrates the smoothness of the control variables at the threshold. With the exceptions of surrender and education attainment, most variables exhibit continuous changes at the threshold. With regard to surrender and education attainment, we use the RDD model to investigate whether significant discontinuities occur at the threshold of the quantity of information. The results indicate no significant discontinuities, thus suggesting that these variables also change continuously; these findings confirm the validity of the first assumption.

Fig. 3: Continuity test.
figure 3

The continuity test of each control variable shown above each panel.

Full size image

Concerning the second assumption, the absence of potential manipulation of the forcing variable also holds true in the context of this paper. The forcing variable in this context is the quantity of citizens’ personal information involved in the infringement, and the threshold for “particularly serious” circumstances is set at 50,000 pieces of information. A review of a certain number of verdicts reveals that when defendants commit this crime, they often engage in multiple acts of personal information infringement of various types and that different types of information feature different criteria for determination. It is challenging for defendants to calculate the total number of pieces of information involved in the infringement accurately. Therefore, we maintain that the possibility to manipulate the quantity of information, as the forcing variable, is low.

Feasibility analysis of regression discontinuity: bandwidth sensitivity

The accuracy of the RDD results can easily be affected by the choice of bandwidth. A smaller bandwidth ensures that the influencing factors on either side of the threshold are more similar to one another, thus reducing the impacts of omitted variables on the dependent variable; in turn, this approach facilitates a more accurate estimation of the treatment effect at the threshold and can help mitigate the estimation bias resulting from endogeneity. However, an excessively small bandwidth can also lead to an excessive loss of sample size, thereby potentially biasing the results. In this section, on the basis of the baseline regression, the bandwidth is set to a range of plus or minus 12,000 pieces of information (Table 4, Panel A) or 8000 pieces of information (Table 4, Panel B) from the threshold in the context of testing bandwidth sensitivity, including control variables and fixed effects pertaining to year and province. The regression results indicate that the causal relationships between the quantity of information and both the duration of fixed-term imprisonment and the amount of the fine assigned as a penalty does not change significantly at different bandwidths or polynomial selections.

Table 4 Bandwidth sensitivity test.
Full size table

Heterogeneity analysis

To investigate the possibility of differences in the baseline regression results across different regions, namely, whether discontinuities in fixed-term imprisonment sentences and fine penalties differ across regions, we divide the samples into Eastern and Central/Western regions of China for the sake of analysis. Table 5 presents the empirical results of the heterogeneity analysis, in which context Panels A and B present the estimation results for Eq. (2) in the Eastern and Central/Western regions, respectively, including control variables and fixed effects pertaining to the province and year.

Table 5 Heterogeneity analysis.
Full size table

With respect to the Eastern region of the country, the estimated coefficients presented in the first three columns indicated that exceeding 50,000 pieces of information entails a treatment effect on fixed-term imprisonment sentences that ranges from 8.777 to 8.885; all these values are significant at the 5% level. Furthermore, the estimated coefficients shown in the last three columns indicated that exceeding 50,000 pieces of information entails a treatment effect on the number of fines levied as a penalty ranging from 13.17 to 15.15; all of these values are significant at the 10% level. This finding indicates that for defendants whose crime of infringement involved more than 50,000 pieces of personal information, both the duration of fixed-term imprisonment and the number of fines levied as a penalty increase significantly in comparison with the corresponding figures for defendants whose crime of infringement involved fewer than 50,000 pieces of information. With regard to the Central/Western regions, the treatment effects on both fixed-term imprisonment sentences and fine penalties are not significant, thus indicating no significant increase in sentencing at the threshold. In summary, significant discontinuities exist at the threshold with regard to both fixed-term imprisonment sentences and fine penalties in the Eastern region, and measures should be taken to mitigate the decisive role played by marginal quantities in this context. In contrast, in the Central/Western regions of the country, the corresponding sentencing is relatively smooth at the threshold, as indicated by the absence of any significant discontinuities. We believe that a likely reason for this finding pertains to the relatively high level of economic development in Eastern China and the relatively low level of economic development in Central China and Western China. For defendants whose crimes led to the same consequences, such as through infringement involving the same number of pieces of personal information or the generation of the same amount of illicit profit, the resulting social harm may be more serious in the Central and Western regions, and the sentencing may be somewhat more severe in this context. This situation leads to the absence of any significant break between the different statutory sentencing ranges in the Central and Western regions of the country.

Further discussion

No significant discontinuity before the implementation of the Two High Court Judicial Interpretations

The baseline regression conducted with respect to the judicial judgements following the implementation of the judicial interpretation indicates a discontinuity at the threshold of the quantity of information pertaining to fixed-term imprisonment sentences. Was this discontinuity present before the implementation of the judicial interpretation? Was this pattern already formed in judicial practice? This paper uses data points prior to June 1, 2017, to estimate Eq. (2) with the aim of addressing these questions.

Table 6 presents the estimation results. Columns (1)–(3) report the RDD results by reference to linear, quadratic, and cubic polynomial fits. The estimated values presented in the table are not significant, thus indicating that before the implementation of the judicial interpretation, no significant discontinuity occurred at the threshold of 50,000 pieces of information with respect to fixed-term imprisonment sentences.

Table 6 Regression discontinuity design on the basis of the sample before the implementation of the Two High Court Judicial Interpretations.
Full size table

The absence of a gradually increasing penalty deterrence ladder after the threshold (i.e., more than 50,000 pieces of information)

The RDD at the threshold of the quantity of information indicates that the marginal penalty volume is maximized within adjacent intervals at the threshold. Does the penalty volume differ depending on the quantity of information before and after the threshold? This section analyses the intervals before and after the threshold with the goal of identifying any differences in deterrence in this context.

Table 7 presents the ordinary least squares (OLS) regression results after the sample is segmented according to the threshold. Columns (1) and (2) present observations associated with information quantities less than 50,000; Columns (3) and (4) present observations associated with information quantities above 50,000; and Columns (5) and (6) present the squared quantity of information in the regression. The dependent variables in Panels A and B are fixed-term imprisonment sentences and fine penalties, respectively, and even columns control for provincial and yearly fixed effects. The results presented in Column (2) indicate that for every 10,000 less than 50,000, the fixed-term imprisonment sentence increases by 0.460 months. Column (4) shows that for every 10,000 above 50,000, the fixed-term imprisonment sentence increases by 0.009 months. Column (6) indicates an inverted U-shaped relationship between the quantity of information and fixed-term imprisonment sentences, in which context the turning point occurs at 501,000, thus suggesting a trend that is characterized by an initial increase followed by a decrease. Since nearly 99% of the observations are on the left side of the turning point, we can assume that for the vast majority of cases, the marginal fixed-term imprisonment sentence gradually decreases alongside an increase in the quantity of information. The results presented in Panel B indicate that the regression outcomes pertaining to fine penalties are not significant.

Table 7 Analysis of deterrence before and after cut-off values.
Full size table

Recommendations for the establishment of a rational penalty ladder

Mitigating the decisive role of the threshold

To mitigate the decisive role played by the threshold in this context, the sentencing guidelines used for theft in the UK introduce overlapping sentencing ranges for the critical threshold (see Table 8). The guidelines state the following: “Sentencers should identify whether any combination of these (referring to sentencing circumstances), or other relevant factors, should result in an upward or downward adjustment from the starting point. In some cases, it may be appropriate to move outside the identified category range when reaching a provisional sentence”.Footnote 12 Overlapping sentencing ranges surrounding the critical point offer judges more discretion, thereby weakening the decisive role played by the threshold in this context.

Table 8 Establishment of a starting point for sentencing and a corresponding range for theft according to the UK sentencing guidelinesa.
Full size table

Specifically, in the context of the sentencing issues faced in China with regard to the crime of infringing upon citizens’ personal information, it is difficult to adjust the sentencing range directly as a result of the clear delineation of statutory penalties. However, it is possible to focus the interpretive space on the quantities that correspond to “serious” and “particularly serious” circumstances, thus allowing for overlaps in the quantities of information across different severity levels. For instance, on the basis of the judicial interpretation, “serious” and “particularly serious” are currently set to ranges of 5000–50,000 and higher than 50,000, respectively. To allow for overlapping ranges, the possibility of revising these ranges to 5000–60,000 and higher than 50,000 can be explored. This suggestion offers a new approach to sentencing with respect to this crime; the specific size of the overlap continues to require further discussion to establish a balance between judicial discretion and fairness. However, as long as the crime of infringing upon citizens’ personal information requires that a certain quantity of information is reached, the complete elimination of the decisive role played by specific amounts in this context remains technically challenging. Therefore, direct amendments to criminal law can mitigate the threshold more easily, thus facilitating more flexible sentencing in judicial practice.

Enhancing deterrence after the threshold

The empirical results of this research indicate that after the threshold is crossed, the greater the quantity of information is, the lower the penalty per unit of information. According to the principle of a balance between crime and punishment, penalties should correspond to the severity of the crime. However, a decreasing penalty per unit of information as the quantity increases does not reflect the principle of equivalent punishment for equivalent crime, nor does it ensure the fairness of the corresponding penalties.

According to deterrence theory, the deterrent effect of penalties relies on their severity to some extent. The state should implement a gradated system of retributive measures that feature different severity levels corresponding to the degree of harm resulting from different actions with the goal of ensuring a rational marginal deterrent effect. Therefore, because fixed-term imprisonment is the most widely used punishment in the context of the crime of infringing upon citizens’ personal information, its duration is directly related to the severity of the punishment, thus reflecting the deterrent effect. Moreover, the penalties must also be embedded in a system of graded penalties that features varying degrees of severity, thus ensuring the reasonableness of the marginal deterrent effect, which requires the amount of information contained in a unit in the crime of infringing upon citizens’ personal information to be basically the same in terms of the number of penalties that are to be imposed.

However, since the maximum sentence for this crime is 7 years and the quantity of infringed information is largely unlimited,Footnote 13 the penalty per unit of information inevitably decreases as the quantity increases. To enhance the deterrent effect after the threshold, higher fines can be paired with larger quantities of information with the aim of ensuring the rationality of the marginal deterrent effect. The empirical results indicate that the fine amounts are not characterized by an increasing penalty per unit of information as the quantity increases, thus leading to a mismatch between the penalty and the nature of the crime. Consequently, given the limitations on the maximum sentence, a gradually increasing deterrent ladder for fixed-term imprisonment cannot be established, nor do fine amounts significantly enhance deterrence. To ensure that the principle of a balance between crime and punishment is followed and achieve fairness in terms of penalties, we can consider improvements in terms of fine amounts, thereby pairing larger quantities of information with higher fines to increase deterrence after the threshold.

Conclusion

Judicial fairness is the most urgent and critical issue in the contemporary judicial field, and developing a rational penalty ladder can help enhance judicial fairness. On the basis of the judicial practice associated with the penalty ladder, this paper explores the rationality of the penalty ladder with the goal of establishing a more rational system. Through empirical research on the crime of infringing upon citizens’ personal information as a representative case, by reference to data concerning judicial judgement and on the basis of methods such as RDD, this paper investigates the rationality of the penalty ladder.

The results of this research reveal that after the 2017 judicial interpretation defined the quantity of information corresponding to different severities of infringement upon citizens’ personal information, a significant discontinuity emerged at the threshold of the quantity of information for fixed-term imprisonment sentences, thus making the marginal penalty volume the greatest within adjacent intervals. This finding indicates good internal consistency within statutory penalty ranges but insufficient connectivity among different penalty ladders. Cases near the threshold do not receive punishments that are commensurate with the crime’s severity, thus contradicting the principle a balance between crime and punishment. In addition, the greater the quantity of information is, the lower the penalty per unit of information. This finding demonstrates that the threshold exhibits strong deterrence but that once it has been crossed, this deterrence decreases significantly without leading to the development of a gradually increasing deterrent ladder.

To address the discontinuity observed at the threshold, the paper suggests that overlapping quantities of information should be allowed for different severities of infringement upon citizens’ personal information, thereby weakening the decisive role played by the threshold in this context. Furthermore, to address the decreasing penalty per unit of information after the threshold, we recommend that larger quantities of information should be paired with more severe fines to achieve the purpose of punishment.