Sentencing disparities in the Czech Republic: Empirical evidence from post-communist Europe

Abstract

Although there is an extensive literature on sentencing disparities in common law countries, there have been only a few empirical studies in continental countries and virtually none in post-communist ones. This article presents findings from the Czech Republic, which show that there are important disparities in post-communist Europe, comparable to those in the USA before the introduction of sentencing guidelines. I employ a multi-level modelling approach to study the sentencing practices in Czech district courts for the three most common offences. The values of intra-cluster correlation are found to be between .066 and .178 for the various models, which is considered high. The specifics of civil law post-communist countries are further discussed in relation to choosing appropriate ways of reducing disparities.

Keywords

Czech Republic post-communist countries sentencing disparities

Introduction

Do judges and courts differ in the way they set sanctions? Many studies have suggested that they do (for reviews, see Pina-Sánchez and Linacre, 2016; Sporer and Goodman-Delahunty, 2009). It is natural that, where there is discretion in decision-making, unwarranted disparity appears. My key question is thus not whether these disparities exist, but how significant they are and how they might be reduced. In this paper, I explore whether there are significant sentencing disparities in a post-communist continental setting, specifically in the Czech Republic.

There are various possible approaches to studying sentencing disparities. Existing research on the extent of sentencing disparities can be clustered into four major groups, which look at disparities from the perspective of (1) differences between courts, (2) differences between judges, (3) legal factors, and (4) the offenders’ extra-legal characteristics.

Studying differences between courts has been one of the most popular approaches (Pina-Sánchez and Linacre, 2013). The main reason for this, albeit a very pragmatic one, is that it is hard to do otherwise: it is often impossible to identify particular judges (so as to be able to study differences between judges) and the information on which the judges based their decisions is frequently not available either. The result is that many studies have resorted to examining inter-court disparities in sentencing, in the USA, Canada, and England and Wales (for example, Fearn, 2005; Johnson, 2006; Pina-Sánchez, 2015; Pina-Sánchez and Linacre, 2013; Reid, 2014; Weinstein, 2006).¹ Some of those studies have also considered the impact of court district characteristics on sentencing.

In a few cases, differences between judges have been studied, notwithstanding the complexity of doing so. In the US context, judges were found to have been responsible for 5 percent of variation in sentencing in 1999–2000 in Pennsylvania (Johnson, 2006), 6 percent in the 2000s in Massachusetts (Scott, 2010) and 17 percent across the United States in the 1980s, subsequently reduced to 11 percent by the introduction of sentencing guidelines (Anderson et al., 1999). Several studies have looked at whether the judge’s personality influences sentencing. Sentencing philosophies (Hogarth, 1971; Sporer and Goodman-Delahunty, 2009) and religious affiliation (Myers, 1988) have been found to play a role. The impacts of the judge’s race (Johnson, 2006; Spohn, 1991; Steffensmeier and Britt, 2001), gender (Spohn, 1991; Steffensmeier and Hebert, 1999) and age (Johnson, 2006; Myers, 1988) are unclear; various studies have arrived at different conclusions.

The third approach investigates whether disparity arises from different assessments of the numerous legal factors that influence the sentence. Examples of such factors are previous convictions (Roberts and Pina-Sánchez, 2015; Roberts and Von Hirsch, 2010), sentencing for multiple offences (Vibla, 2015) and aggravating or mitigating factors (Pina-Sánchez and Linacre, 2014; Roberts, 2014).

Lastly, many studies have examined the impact of offenders’ extra-legal characteristics, which should not influence sentencing. The offender’s race, gender and age (for example, Daly and Tonry, 1997; Koons-Witt, 2002; Steffensmeier and Demuth, 2006; Steffensmeier et al., 1998), and less obvious factors such as the offender’s attractiveness (Sporer and Goodman-Delahunty, 2009), have been found to play an important role.

The vast majority of the studies I have mentioned focused on measuring the level (or extent) of sentencing disparities, as I do in this paper; other studies have focused on why sentencing disparities exist and how judges arrive at their conclusions (for example, Hutton and Tata, 2002; Tata, 2002, 2007). This is not, however, the focus of this paper and thus I do not go into further detail about this aspect of the theory of sentencing disparities.

This paper is structured as follows: I begin by discussing civil law research on sentencing disparities while paying specific attention to post-communist countries. I then examine the Czech legal context, present my methodology and results, and conclude with a discussion on possible methods of reducing sentencing disparities that would be appropriate for the Czech Republic as a post-communist civil law country.

Continental European research on sentencing disparities

Although there has been abundant research published in English on inter-court and inter-judge sentencing disparities in common law systems, very little research has yet been published in English on such disparities in civil law countries. Some empirical research has appeared concerning the Netherlands (Berghuis, 1992; Berghuis and Mak, 2002; Fiselier, 1985), Finland (Lappi-Seppälä and Hinkkanen, 2004) and Germany (Albrecht, 1994; see research mentioned in Heinz, 2013). In other countries, suspicions of inter-court disparities have been raised and contemplated, but there seems to have been little or no empirical research into these; this is the case for Italy (Mannouzzi, 2002), Belgium (Monsieurs et al., 2011) and France (Hodgson and Soubis, 2016). Only a few authors have touched upon the impact that the specifics of civil law systems have on sentencing disparities (Council of Europe, 1992; Killias, 1994; Krajewski, 2016; Lappi-Seppälä, 2001).

There has not yet been any research into inter-judge or inter-court sentencing disparities in the post-communist countries.² Although there are national debates about sentencing (for the Czech Republic see, for example, Scheinost, 2015; Scheinost and Válková, 2015), empirical research in the field of sentencing is scarce and empirical research into sentencing disparities is virtually non-existent. One of the few existing studies looked at foreign nationals in the criminal justice system, and found a very low level of discrimination (Vávra, 2016). Descriptions of sentencing practices in post-communist countries in English are also rare, the latest exceptions being Krajewski (2016) on Poland and Plesničar (2013) on Slovenia.

Before studying the post-communist countries, we must first ask ourselves whether there is any reason to do so. Are these countries substantially different from those that have been studied already? What sets post-communist civil law countries apart from other countries?

As in other civil law countries, judges decide on the accused’s guilt and on the sentence at the same time. This differs from the practice in common law countries and this feature has not been given enough consideration in previous work on sentencing disparities. Furthermore, judges in post-communist countries often interpret the law in a very formal way (Kühn, 2004); this might have an impact on sentencing, which is typically a matter of principles more than of rules. Judges also play a less important role in shaping sentencing in civil law countries because the initiative for changes in sentencing comes from the legislator (Lappi-Seppälä, 2001). In post-communist countries, there is specific distaste for secondary legislation owing to its former abuse by the communist regimes.

The post-communist countries were not included in the Council of Europe’s discussions preceding the Recommendation on Consistency of Sentencing in 1992. Even since that time, many post-communist countries have arguably had more important issues to resolve and have thus virtually ignored the Recommendation. Yet, since the prisoner population index is higher in post-communist European countries than in other civil law European countries (Council of Europe, 2014), sentencing disparities might be a more important issue in the post-communist countries owing to the sheer number of people being sentenced.

Lastly, some post-communist countries (for example, Slovenia and, until recently, the Czech Republic) do not provide researchers with regular access to data on sentencing from which they could calculate disparities. In some countries, such data are not available in an aggregated form, and in others the governments do not make the data available at all. These differences suggest that the situation in post-communist European countries might be different from that revealed by existing research in ‘western’ Europe or the USA. It is therefore desirable to carry out empirical research into disparities in the post-communist countries and I proceed to do so in this paper, focusing on the Czech Republic.

Methodology

Czech legal context

To understand inter-court disparities in the Czech Republic properly, three areas of the country’s legal system must be noted: (1) the legal framework, (2) the organization of the courts, and (3) the sentencing and sanction regimes.

Firstly, the legal system is based on the broader civil law culture. This implies that there is a sole Penal Code (law n. 40/2009 Coll.) in which all offences and sanctions are enumerated, together with the general provisions of substantive criminal law. Criminal law is limited to the most serious offences, while administrative penal law deals with most violations of legal provisions.

Secondly, there are four levels of general criminal courts: 86 district courts, 8 regional courts, 2 high courts, and the Supreme Court. District courts decide in the first instance on all offences for which the minimum sentence is less than five years of imprisonment, while regional courts decide the most serious offences and some specific offences. Regional courts also serve as appellate courts for the district courts, while high courts serve as appellate courts for the regional courts. The Supreme Court can rule only on the legal provisions and not on the merits of the case. Moreover, the Supreme Court has interpreted the Code of Criminal Procedure such that the Supreme Court can neither interfere in setting sanctions nor decide on the principles of sentencing. In addition to these general courts, there is a Constitutional Court, which can quash any court decision submitted to it and often deals with criminal matters.

Thirdly, the Penal Code offers a wide variety of sanctions. The most common sanction is a sentence of suspended imprisonment (65.8 percent),³ followed by non-suspended prison sentences (16.1 percent) and by an array of (mainly alternative) sanctions: community work (9.9 percent), fine (3.2 percent), expulsion (0.6 percent), prohibitions, home detention, seizure of things, seizure of property and the sanction of degradation. The Czech system of sanctions is a dual-track system: the sanctions are divided into punishments and safety measures (Šámal et al., 2014). Safety measures may be imposed only under strict conditions; these apply mainly to insane offenders who are dangerous to society.

When setting sanctions, judges operate with wide discretion. The main sentencing principles prescribed by the law are that sanctions must be proportionate to the committed offences, all sanctions must be individualized to the offender, and harsher punishments cannot be imposed if more lenient ones would suffice. A new Criminal Code adopted in 2009 explicitly left the aims of punishment undefined. The Penal Code prescribes the characteristics that judges must consider very generally and specifies possible aggravating and mitigating circumstances. However, no further detailed guidance is provided either by the legislator or by jurisprudence. The provisions thus offer the judges an almost unrestricted choice of sanctions in less serious cases (compare ss. 60, 62 and 67 of the Criminal Code). Even though the judges are compelled by law to give reasons for their choice of sanction, they often do not provide sufficient reasons for their specific choice other than a general indication of their reasoning, such as ‘appropriate to the offender and the offence committed’, or a generic description of the offender’s criminal history.⁴

The data

To analyse inter-court consistency and disparities, data on real sentencing from court records are the most appropriate (Dhami and Belton, 2015). Data on sentencing in the Czech Republic are gathered by the Czech Ministry of Justice, which collects data on every criminal proceeding in the Czech courts in an anonymous form. The dataset includes information about the offender, the offence, the proceedings, the types of offences and the sanctions. These data were provided to me in full by the Ministry of Justice.

As Roberts and Hough (2015) point out, every database has certain limitations. The main disadvantage of my dataset is that it does not include basic information about the seriousness of an act within the definition of a particular offence. Information about the victim is scarce, as are any details of the offence. Despite this, the information provided can still form a valid basis for research. The implications of the limited scope of the available data are reflected in the design of the study and in the interpretation of the results.

The choice of offences

I chose to focus on sentencing disparities related to three specific offences: Evasion of Alimony Payments (s. 196 of the Penal Code), Repeated Theft (s. 205/2 of the Penal Code; a theft occurring within three years of a sentence or punishment related to another theft) and Frustrating Execution of an Official Decision (s. 337/1a of the Penal Code; this refers in practice to driving without a permit – compare jurisprudence in Šámal et al., 2012). Between 2012 and 2014 244,497 offences received guilty verdicts. Of those, 30,320 sentences were set for Evasion of Alimony payments, 19,633 for Frustrating Execution of an Official Decision and 16,798 for Repeated Theft. The two most common offences, simple theft (36,907 cases) and driving under the influence (32,837 cases) could not be used in this analysis because only a very small proportion of them resulted in non-suspended prison sentences.

These offences were chosen for two reasons. First, since it was not possible to analyse the seriousness of the offences committed, it seemed most reliable to study types of offences that are likely to be very similar across the country, for which we may assume there is less variation in the particulars of the specific offences dealt with by different courts. Second, in order to obtain statistically significant results across 86 district courts, I needed to look at offences with large enough samples. Finding a balance between case variation and sufficient numbers of cases is a common issue in sentencing research (Roberts and Pina-Sánchez, 2015).

The offences I have chosen are some of the most common offences. Choosing these enables me to focus on only certain subsections of these offences while maintaining sufficiently large samples. Analysing cases of a single offence or subsection of an offence limits the variation in seriousness across the cases and thus ensures higher comparability between the analysed cases. To give an example, there are several types of property offences, such as theft and fraud, and their subsections consist of simple and aggravated versions. When these types of offences (or their subsections) are examined separately rather than jointly, the cases are more similar to one another and the case-based differences between the offences dealt with by the different courts are smaller, which renders the research more internally valid.

A further advantage of studying the most common offences is that judges have more opportunities to develop consistent sentencing patterns when dealing with a high number of the same type of offence. These offences also involve the biggest proportion of offenders, so together with the most serious offences they are the most policy-relevant offences. Finally, the most common and least serious offences are less likely to be corrected by appellate courts (Tata, 2002) and thus any disparity in the way they are initially sentenced deserves greater attention.

The cases in my dataset were further filtered so as to provide a sample of cases that were as similar as possible while large enough for meaningful statistical analysis. The following characteristics were chosen to limit all of the examined offences: the offence was committed by an adult (over 18 years old at the time), decided by a district court⁵ and resulted in a guilty verdict⁶ in which the offender was found guilty of only one specific type of offence. This is an advantage compared with some other sentencing research in which only the offence with the highest sentence was considered‚ leading to possible bias when comparing multiple-offence cases with single-offence cases (Merrall et al., 2010). Moreover, I considered only cases in which the offender was sentenced according to the new Penal Code, law no. 40/2009 Coll, in order to enhance similarity (a very small minority of cases − 0–2 percent – in the period I looked at were sentenced according to the previous Penal Code). The cases considered were decided between 2012 and 2014. Preliminary analysis showed that 97 percent of cases sentenced in 2007 had been entered into the database by the end of the following year; criminal procedure in the Czech Republic has become speedier in recent years (Dušek, 2015), and datasets were available for the years 1996–2015. I wished to examine the most recent data, and thus it seemed most reasonable to examine cases decided in 2012–14. The most significant limitation of this approach is that overall consistency across the system is not discernible: I can identify disparity related only to the specifically chosen offences. Less common and more varied offences are not considered in this research.

That said, there are reasons to suspect that the offences that I do not analyse here may be handled with greater inconsistency than those I do examine. For instance, if a judge decides 50 cases of theft per year, after a few years they develop their own sentencing patterns specific for thefts and their decisions become (at least in theory) more consistent. On the other hand, if the same judge decides only three cases of large-scale drug possession per year, it is harder for them to develop a consistent sentencing policy. Thus, if the simplest and most common offences are sentenced inconsistently, there will likely be even more inconsistency in the sentencing of more complicated and less common offences.

The variables

Since incarceration is the harshest possible punishment, from a policy perspective it is most important to study disparities in the use of prison sentences (Tonry, 2001). My three measures are therefore all linked to prison sentences and how they are set in the Czech Republic: (a) a binary variable indicating whether a non-suspended sentence was given or not in each individual case; (b) a continuous variable measuring the length of the non-suspended sentence; (c) a continuous variable measuring the length of the suspended sentence.

The decision whether or not to immediately incarcerate an offender is the most important decision that the judge makes when sentencing; the second most important decision is the length of the imprisonment. Since suspended prison sentences are the most common form of sanction in the Czech Republic and the most common for the offences analysed, I also consider their length. Suspended and non-suspended prison sentences need to be analysed separately, because suspended imprisonment is viewed by both professionals and the public as a special type of punishment and as different from a non-suspended prison sentence (Novotná, 2009; Sotolář et al., 2000).

Since the residuals of the continuous outcome variables were not normally distributed; they were log-transformed to improve their distribution towards a normal distribution.

As independent variables I chose legal factors, defined as those that could and should have an impact on the sentence according to a strict interpretation of criminal law in the Czech Republic. These include the following:

Factors related to recidivism: whether the offender is viewed legally as a recidivist and how many previous convictions they have. The number of previous convictions is considered in categorical form since it has been suggested that the impact of previous convictions is not linear (Roberts and Pina-Sánchez, 2015): the categories are: 0; 1–2; 3–6; 7–10; >10 convictions. The year of the decision is also considered because an amnesty in 2013 influenced the legal view of recidivism and the execution of some non-custodial punishments, which should have influenced the sentencing process.

Factors related to the specification of the main punishment: the type of prison to which the offender is sent; whether a conditional prison sentence is with or without supervision; the length of probation.

Factors related indirectly to the seriousness of an offence: the type of indictment (full indictment or simplified indictment called ‘motion to punish’) and the number of main hearings. For the offences of Evasion of Alimony Payments and Frustrating Execution of an Official Decision, a legal amendment passed in 2012 that should have influenced sentencing is taken into consideration. For Evasion of Alimony Payments, different subsections of the offence that indicate different levels of seriousness are also considered.

All of the variables are described in Table 4 in the Appendix.

Analytical approach

The preferred quantitative approach to studying inter-court disparities is currently a multi-level modelling approach (Merrall et al., 2010; Pina-Sánchez and Linacre, 2016). This has recently been used in several studies measuring inter-judge and inter-court disparities (Fearn, 2005; Johnson, 2006; Merrall et al., 2010; Pina-Sánchez and Linacre, 2013, 2014; Roberts and Pina-Sánchez, 2015). Although this method has many advantages, its one disadvantage is that it does not enable direct comparisons across jurisdictions (Merrall et al., 2010; Pina-Sánchez and Linacre, 2016).

Multi-level modelling is necessary because the data on sentencing are hierarchically organized. A simple regression would not work well given that actors at different levels can have various impacts on sentencing, and this cannot be explained by a basic regression. Multi-level modelling tells us what effect the different levels have and how much of the inconsistency is a result of the higher levels. In this case, individuals are nested within courts. It would be interesting to establish another two levels – judges and appellate courts. Unfortunately, when this research was carried out, no reliable data were available to link particular judges and cases. Meanwhile, the number of appellate courts is too small, so it was not possible to carry out more than a two-level model of inter-court disparities.

My analysis employs a basic type of multi-level model called a random intercept model. This consists of a regression with an intercept that is allowed to vary in order to reflect differences at court level that cannot be explained by the models (Pina-Sánchez and Linacre, 2016).

I report the results of this method in two forms: first as regression tables showing the independent variables, and second as the intra-cluster correlation (ICC), which is a measure of the variation in the outcome variable that occurs between groups as a proportion of the total variation present (Finch et al., 2014). Simply put, the ICC tells us what proportion of the overall variation is due to differences between the courts (Pina-Sánchez and Linacre, 2013). In an ideal world without disparities, the ICC should be zero.

The ICC is measured slightly differently for linear and logit models. Let V_A be between-court variability (area-level variance) and V_I within-court variability (individual-level variance). I calculate the ICC for linear models using the following formula:

Formula 1: ICC for linear models

I C C = V_{A} / (V_{A} + V_{I}) .

To calculate the ICC for logit models, the linear threshold model method is used (for further discussion, see Merlo et al., 2006; Snijders and Bosker, 1999). V_I is equal to π²/3 in this case because the unobserved individual variable has a logistic distribution (Merlo et al., 2006). However, because there is ‘an inherent difficulty of distinguishing the individual level and the area level variance in the logistic case’ (Merlo et al., 2006: 8), the outcomes of logistic multi-level regressions may not be entirely reliable. The formula used is:

Formula 2: ICC for logit models

I C C = V_{A} / (V_{A} + π^{2} / 3) .

Altogether, nine different models of fixed effects are reported to measure the three different outcome variables for the three different offences. For the purposes of calculating the ICC, two different models are considered for each offence and outcome variable: a null model and a model incorporating legal factors. Furthermore, the log-likelihoods of the null logit models are reported for comparison with the log-likelihoods of the advanced models.

The program used for the calculation was RStudio (R version 3.2.3, RStudio v. 0.99.896), with packages lme4 (v. 1.1-12) and lmerTest (v. 2.0-30). It is best to refer to the theory and practice behind the lme4 package (Bates, 2014; Bates et al., 2015; Finch et al., 2014) for a possible critical evaluation of the methods.

Results

Descriptive statistics

I begin by describing the differences between courts without considering any of the characteristics of the offenders or their offences. Only courts in which a reasonable number of cases were decided are considered, in order to limit possible anomalies and enhance generalizability.⁷ Although these descriptive statistics can offer some insight into the issue, it is important to keep in mind that descriptive statistics of this kind may suggest higher disparity than exists in reality (Roberts and Pina-Sánchez, 2015). The statistics related to the Frustrating Execution of an Official Decision offence (= driving without a licence after a ban) are reported in Figures 1 –4 because this offence is presumed to be the most similar offence across the country. The results for the other two offences are very similar.⁸

Figure 1.

Custody rate by court.

Figure 2.

Average non-suspended prison sentence by court.

Figure 3.

Average suspended prison sentence by court.

Figure 4.

Average number of months of non-suspended sentence per case by court.

In order to test the hypothesis that some courts send fewer people to prison but for longer periods than others (or vice versa), Figure 4 depicts the total length (in months) of non-suspended sentences set by the court divided by the number of cases the court dealt with. It thus incorporates two of the most important measures of severity (percentage of imprisoned offenders and length of non-suspended prison sentences) into one.

Multi-level modelling

When employing multi-level models, the main variable of interest is the ICC, which tells us what proportion of overall variation the courts are responsible for. The lower it is, the higher the consistency. When discussing the results from logit multi-level models, we must bear in mind that their results are slightly less reliable than those of the linear models.⁹

Table 1 provides information about the ICCs, while the regression tables, numbers of cases, R² and the log-likelihoods are reported in Tables 2 and 3. The ICCs range between .066 and .178, meaning that 6.7–17.8 percent of the overall variation in sentencing is due to differences between courts. If we calculate the average ICCs for the three outcome variables, we see that there is no important difference between them (values of 0.123, 0.123 and 0.116). This suggests the inconsistencies are similar when considering different outcome variables.

Table 1.

Intra-cluster correlations.

Outcome variable	Evasion of Alimony Payments (s. 196)		Repeated Theft(s. 205, subs. 2)		Frustrating Execution of an Official Decision(s. 337, subs. 1a)
Outcome variable	Null model	Legal factors	Null model	Legal factors	Null model	Legal factors
Incarceration (yes/no)	.097	.154	.139	.160	.084	.106
Non-suspended prison sentence	.066	.178	.095	.109	.126	.161
Suspended prison sentence	.100	.169	.091	.068	.121	.168

Table 2.

Log-likelihoods and R² of various models.

Outcome variable	Evasion of Alimony Payments (s. 196)		Repeated Theft (s. 205, subs. 2)		Frustrating Execution of an Official Decision (s. 337, subs. 1a)
Outcome variable	Null model	Legal factors	Null model	Legal factors	Null model	Legal factors
Incarceration (log-likelihood)	−6661	−5204	−8233	−7539	−6758	−2063
Non-suspended prison sentence		.599		.379		.270
Suspended prison sentence		.623		.190		.564

Table 3.

Regression tables of various models.

Incarceration (yes/no)	Evasion of Alimony Payments (s. 196)		Repeated Theft (s. 205, subs. 2)		Frustrating Execution of an Official Decision (s. 337, subs. 1A)
	OR	95% CI	OR	95% CI	OR	95 % CI
(Intercept)	0***		0.24***	0.11, 0.52	0.11***	0.07, 0.18
Recidivist by law (no/yes)	2.33***	1.95, 2.80	2.07***	1.82, 2.35	2.10***	1.66, 2.66
No. of prev. convictions: 1–2 (ref: 0 for s. 196 and 205, ref: 1–2 for s. 337)	16.70***	7.97, 35	1.15	0.54, 2.45
No. of prev. convictions: 3–6	66.80***	32.1, 139	4.24***	2, 8.96	7.23***	5.47, 9.57
No. of prev. convictions: 7–10	135.00***	64.8, 283	6.26***	2.96, 13.30	15.40***	11.44, 20.73
No. of prev. convictions: >10	177.0***	83.3, 380	8.06***	3.8, 17.10	29.11***	21.02, 40.32
Year of conviction: 2013 (ref: 2012)	0.33***	0.28, 0.40	0.68***	0.62, 0.75	0.42***	0.31, 0.57
Year of conviction: 2014	0.86.	0.72, 1.02	0.90*	0.82, 0.99	0.73*	0.54, 0.99
Trial beginning (Indictment – Motion to punish)	0.49***	0.43, 0.56	0.51***	0.44, 0.59	0.22***	0.15, 0.33
Amendment no. 390/2012 Coll. (no/yes)	0.86.	0.73, 1.01			1.34*	1.05, 1.72
Subsection: 2 (ref.: 1)	1.90***	1.45, 2.49
Subsection: 3	2.82***	2.48, 3.19
Length of non-suspended sentence	E(Coeff)	SE	E(Coeff)	SE	E(Coeff)	SE
(Intercept)	6.539***	1.140	6.016***	1.140	4.418***	1.152
Recidivist by law (no/yes)	1.063*	1.027	0.959*	1.018	0.935	1.052
No. of prev. convictions: 1–2 (ref: 0)	1.084	1.139	1.272.	1.138
No. of prev. convictions: 3–6	1.164	1.138	1.340*	1.137	1.008	1.072
No. of prev. convictions: 7–10	1.208	1.139	1.388*	1.138	1.092.	1.079
No. of prev. convictions: >10	1.150	1.143	1.410**	1.138	1.105**	1.084
Year of conviction: 2013 (ref: 2012)	0.876***	1.031	1.000	1.015	1.005	1.065
Year of conviction: 2014	0.855***	1.029	1.019	1.014	1.065	1.066
Prison level: lightest (ref. light)	0.967	1.048	0.914	1.067
Prison level: hard	1.056**	1.021	1.227***	1.015	1.373**	1.106
Prison level: hardest			1.807*	1.311	1.843***	1.111
No. of main hearings	0.997	1.007	1.023***	1.006	1.011	1.016
Trial beginning (Indictment – Motion to punish)	0.941**	1.021	0.888***	1.023	0.966	1.082
Amendment no. 390/2012 Coll. (no/yes)	0.931**	1.025			0.989	1.053
Subsection: 2 (ref.: 1)	1.445***	1.044
Subsection: 3	1.930***	1.022
Length of suspended prison sentence	E(Coeff)	SE	E(Coeff)	SE	E(Coeff)	SE
(Intercept)	3.398***	1.017	5.792***	1.060	2.661***	1.057
Recidivist by law (no/yes)	1.024*	1.011	1.005	1.017	1.038	1.026
No. of prev. convictions: 1–2 (ref: 0)	1.114***	1.006	1.018	1.055	1.160***	1.016
No. of prev. convictions: 3–6	1.194***	1.006	1.049	1.055	1.246***	1.018
No. of prev. convictions: 7–10	1.245***	1.009	1.102.	1.056	1.344***	1.024
No. of prev. convictions: >10	1.214***	1.016	1.064	1.056	1.372***	1.033
Year of conviction: 2013 (ref: 2012)	0.927***	1.006	1.013	1.012	0.991	1.021
Year of conviction: 2014	0.935***	1.007	1.017	1.012	1.006	1.023
The length of probation (months)	1.024***	1.000	1.015***	1.000	1.033***	1.001
With supervision (no/yes)	1.111***	1.007	1.055***	1.011	1.094***	1.025
Additional prohibition (no/yes)	0.989.	1.006	1.041	1.025	1.003	1.090
No. of main hearings	1.000	1.002	0.999	1.007	0.983	1.016
Trial beginning (Indictment – Motion to punish)	0.950***	1.006	0.919***	1.019	1.071	1.048
Decided via penal order (yes/no)	0.986.	1.007	1.062***	1.018	1.066*	1.029
Amendment no. 390/2012 Coll. (no/yes)	0.991	1.006			0.983	1.017
Subsection: 2 (ref.: 1)	1.306***	1.011
Subsection: 3	1.453***	1.007

Notes: OR = Odds Ratio, CI = Confidence Interval, E(Coeff) = Exp(Coefficient of correlation), SE = standard error, *** = correlation is significant at 0.001 level, ** = p < .01 level, * = p < .05 level, . = p < .1 level.

When we calculate average ICCs for each offence, we find certain differences between them (.127, .110 and .128). Unexpectedly, in eight out of nine cases the null models produced lower ICCs than the models incorporating individual-level variables. We can speculate that this is because courts assess these variables (especially previous convictions) in different ways since there is very little guidance on how they should influence the sentence.

The R² and log-likelihoods as shown in Table 2 reveal why it is important to study different offences separately: the same characteristics are taken into consideration differently for different offences. Whereas independent variables explain more than 60 percent of the length of non-suspended prison sentences for Evasion of Alimony Payments offences and 56.4 percent for offences of Frustrating Execution of an Official Decision, they explain only 19 percent in the case of Repeated Theft. Similarly, whereas the characteristics of the case and offender explain 59.8 percent of the length of suspended prison sentences for Evasion of Alimony Payment offences, the same characteristics explain only 37.8 percent of the length of suspended prison sentences for Repeated Theft and just 27 percent for Frustrating Execution of an Official Decision . The extent of the explained length of sentences is thus high for some and lower, yet still substantial, for other offences.

The regression tables as reported in Table 3 reveal a similar story. Different characteristics play different roles for various sanctions and offences. A clear example is how the number of previous convictions influences sentencing. Previous convictions play a very important role in judges’ decisions on whether to incarcerate an offender or not. However, when it comes to deciding the length of a non-suspended prison sentence, the number of previous convictions is influential only in cases dealing with Repeated Theft, and not the other two offences. On the other hand, the number of previous convictions is influential in deciding the length of suspended prison sentences for the other two offences (Evasion of Alimony Payments and Frustrating Execution of an Official Decision), but not for Repeated Theft.

The year of conviction is seen to have significantly influenced judges’ decisions on whether or not to incarcerate offenders but to have played little or no role in determining the length of their sentences. It is also unclear what role a statutory change in the sentencing range played (or should have played). It had a different influence on sentencing for each of the two offences whose sentencing ranges were changed: Evasion of Alimony Payments and Frustrating Execution of an Official Decision. All of these results suggest that there are certain mechanisms in play that we do not yet fully understand and that these mechanisms differ for different types of sentence and offence. However, (unfortunately) they are not the aim of this article and thus are not debated.

Limitations

This study has high external validity since it considers all sentences imposed in the Czech Republic during the chosen period for the offences studied. On the other hand, it examines only cases decided at district courts and only three types of offence, none of which are very serious or violent. For these reasons, it may not be possible to draw conclusions from this paper about overall inter-court disparity in the Czech Republic.

The study’s internal validity is threatened primarily by the fact that the official datasets do not include more detailed data on the seriousness of the offences. Although I have partially indirectly controlled for seriousness, as explained above, this will have an important impact on the interpretation of the results. Several other minor limitations have also been discussed above.

Another matter of importance for my results is that of diversions, which the data do not take into consideration. These are cases that were dealt with by the prosecution alone. In 2013 this affected 5.6 percent of cases (Supreme Prosecutor’s Office, 2014). Since these diversions are not very common, variations in their usage should not substantially influence the internal validity of the study, though they may have a slight impact, in particular in relation to the proportion of incarcerated offenders.

Discussion

Although there is disparity in every sentencing system, my results strongly suggest that there are substantial and significant disparities in the Czech Republic. The empirical research I have carried out is unique in post-communist Europe to date and rare in countries with a civil law system. The descriptive statistics reveal large differences: rates of incarceration are up to 20 times higher in the most punitive courts than in the least punitive courts. It is difficult to imagine that this situation could be caused solely by the offenders’ characteristics or those of the offences they committed. However, since there has been very little empirical research on sentencing in the Czech Republic, more research is needed to establish why judges choose to send offenders to prison. These reasons might be different in post-communist civil law systems than in their ‘western’ counterparts, and even more different in common law countries, especially the USA. In particular, further research is needed to find out how strongly Czech judges’ personalities influence their sentencing practices and how substantial the inter-judge disparities are.

On the other hand, my descriptive statistics suggest that disparities in the lengths of prison sentences are lower than the disparities in rates of incarceration. This does not mean they are not problematic: the most severe courts still sentence offenders for up to twice as long as the most lenient do. However, the lengths of sentences given vary less than the rates of incarceration.

The results of my multi-level models also suggest that there are substantial sentencing inconsistencies in the Czech Republic. The ICCs are between .066 and .178. In England and Wales, a study of sentencing for assault offences obtained an ICC of .018, which was deemed low (Pina-Sánchez and Linacre, 2013). In the USA the ICCs for inter-judge disparity were around .17 before the introduction of sentencing guidelines (Anderson et al., 1999), and .05–.11 afterward (Anderson et al., 1999; Johnson, 2006; Scott, 2010) and these were considered significant enough to prompt changes in sentencing structures. From a comparative point of view, it thus seems reasonable to assert that there is relatively high inconsistency in sentencing in the Czech Republic.

The Czech disparities are high both from a layperson’s view – about 12 percent of a sentence depends on the court that hands it down – and from a comparative perspective. My results indicate that the disparity is substantial enough to merit measures being brought in to reduce it. Put simply: something needs to be done, especially to better define the custody threshold.

What is it that needs to be done? Post-communist civil law countries are specific, as I explained above, so any measures taken to reduce sentencing disparities would need to be country-specific. The first step towards finding an appropriate solution is to understand the mechanisms of sentencing (Hawkins, 2003); the legislator then can choose measures that reflect the specific country’s constitutional principles and legal traditions (Council of Europe, 1993) and are realizable in the current political climate (Ashworth, 2009). Properly structuring sentencing and reducing unwarranted disparities is one of the greatest challenges for the legislature (Roberts, 2009). Although full discretion means widespread disparity, rules that are too narrow result in treating unlike cases alike (Roberts, 2009; Tonry, 1996: 180). At some point, it might be more appropriate to use instruments other than rules, since less coercive methods might be more effective (Hawkins, 2003).

Certain measures should be employed in states where sentencing is highly pressured by the political and social context, such as in England and Wales (Wasik, 2004); other measures will be employed in countries where such pressures are not common, as in the Czech Republic. There is no uniform strategy that could be applied to any state. On the contrary, there is a diverse mixture of mechanisms that can be applied in various ways (Tonry, 1996: 190).

The following are some measures that I suggest as appropriate for the Czech Republic, beginning with the least intrusive: enable the Supreme Court to decide on sentencing principles (which it currently does not); clarify statutes regarding the aims of punishments (Ashworth, 2009), the custody threshold (Padfield, 2011) and aggravating and mitigating circumstances (Roberts, 2014); require judges to give reasons for their sentencing decisions (see Lappi-Seppälä, 2016, and Weigend, 2016), and provide this reasoning to the appellate courts; publish data on sentencing and develop a sentencing information system (see Miller, 2014).

Furthermore, the Czech Supreme Court is empowered to edict a unifying opinion in order to ensure the uniformity of court decisions. These unifying opinions are not connected to any particular case, and do not bind other judges but simply recommend certain approaches. This is a legacy of the socialist legal system and is unknown in Western countries (Smejkalova, 2009). Although those opinions have received strong theoretical criticism (Balák, 2007; Baráková, 2016; Kühn, 2012), they are one of ideal instruments for unifying jurisprudence where sentencing is concerned. However, the Czech Supreme Court rarely uses it for this purpose.

Finally, the most intrusive measure acceptable in civil law post-communist countries would be to introduce prosecutorial guidelines. In this, the Czech Republic might take inspiration from the Netherlands (see, for example, Tak, 2001, or Van Wingerden, 2014). Since in civil law systems, including the Czech Republic, prosecutors usually suggest specific sanctions (Hodgson and Soubis, 2016; Krajewski, 2016; Lappi-Seppälä, 2016; Plesničar, 2013; Scheirs et al., 2016; Weigend, 2016), sentencing disparities might be reduced if those prosecutorial suggestions were unified. I expect such prosecutorial guidelines to decrease sentencing disparities because the Czech prosecution service is much more hierarchically organized than the court one and the Czech Supreme Prosecutor has the authority to edict guidelines binding all other prosecutors. Yet it is unclear to what level disparities would be lowered by adopting this measure.

Although various measures such as these might be employed to enhance consistency in sentencing, care should be taken to ensure this does not lead to more severe sanctions, and unusual hardship could be avoided by taking into account the offender’s characteristics (Council of Europe, 1993). As Gertner notes, a ‘uniformity-focused, criminal-record emphasis, incarceration-obsessed criminal justice policy’ should never be the goal of any sentencing reform (Gertner, 2014: 315) and such reforms would contradict the humanistic civil law traditions of imposing the least harmful possible sanction.

Footnotes

Appendix

Table 4.

Descriptive statistics for the independent variables.

	Evasion of Alimony Payments (s. 196)	Repeated Theft (s. 205, subs. 2)	Frustrating Execution of an Official Decision (s. 337, subs. 1A)
All cases
Proportion of recidivists by law	0.064	0.197	0.096
No prev. conv.	5228	51	2887
1–2 prev. conv.	9750	2859	5568
3–6 prev. conv.	9892	5125	5524
7–10 prev. conv.	2722	2718	2418
>10 prev. conv.	706	2016	1252
Year of decision: 2012	8823	4362	5454
Year of decision: 2013	10,386	3704	6654
Year of decision: 2014	9089	4703	5541
Proportion of trials beginning with motion to punish	0.425	0.077	0.027
Proportion of cases decided after amendment no. 390/2012 Coll.	0.421		0.272
Subsection 1	21,729
Subsection 2	1399
Subsection 3	5170
Number of cases	28,298	12,769	17,649
Non-suspended prison sentences
Proportion of recidivists by law	0.185	0.262	0.220
No prev. conv.	7	9	0
1–2 prev. conv.	258	461	150
3–6 prev. conv.	875	1791	744
7–10 prev. conv.	384	1157	565
> 10 prev. conv.	116	955	399
Year of decision: 2012	825	1543	711
Year of decision: 2013	257	1240	466
Year of decision: 2014	558	1590	681
A (lightest)	65	41	47
B (lighter)	1072	1583	874
C (harsh)	503	2747	937
D (harshest)	0	2	0
Average no. of main hearings	1.777	1.461	1.639
Proportion of trials beginning with motion to punish	0.543	0.077	0.050
Proportion of cases decided after amendment no. 390/2012 Coll.	0.348		0.280
Subsection 1	789
Subsection 2	83
Subsection 3	768
Number of cases	1640	4373	1858
Suspended prison sentences
Proportion of recidivist by law	0.047	0.133	0.052
No prev. conv.	5059	23	2300
1–2 prev. conv.	8072	916	3218
3–6 prev. conv.	7004	1089	2419
7–10 prev. conv.	1710	542	845
> 10 prev. conv.	407	362	393
Year of decision: 2012	6252	871	2557
Year of decision: 2013	9131	1093	3995
Year of decision: 2014	6869	968	2623
Average length of supervision in months	23.560	29.447	21.203
Proportion of susp. sent. w/ supervision	0.112	0.319	0.073
Proportion of additional prohibitions	0.251	0.044	0.003
Average no. of main hearings	0.351	0.343	0.232
Proportion of trials beginning with motion to punish	0.407
Proportion of decisions via penal order	0.407	0.067	0.020
Proportion of cases decided after amendment no. 390/2012 Coll.	0.430		0.263
Subsection 1	18,136
Subsection 2	1094
Subsection 3	3022
Number of cases	22,252	2932	9175

Acknowledgements

I would like to thank Katherine Auty, who was my supervisor in Cambridge, for all her advice, and Julian Roberts, José Pina-Sánchez, Tony Bottoms, Ingrid Obsuth, Lukáš Drápal, Libor Dušek, Antje du Bois-Pedain, Wolfgang Heinz, Susanne Karstedt, Mojča Plesničar, Tapio Lappi-Seppälä and the members of the Institute of Criminology in Ljubljana for their recommendations. I am grateful to Annie Bartoň, Catherine Byfield and Gabrielle Linelle for helping perfect my English.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by the Ministry of Education, Youth and Sports of the Czech Republic – Institutional Support for Long-term Development of Research Organizations – Charles University, Law Faculty [PRVOUK 06 and PROGRES Q02].

Notes

ORCID iD

Jakub Drápal

References

Albrecht

H-J

. (1994) Sentencing and disparity – a comparative study. European Journal on Criminal Policy and Research 2(2): 98–104.

Anderson

Kling

Stith

(1999) Measuring inter-judge sentencing disparity before and after the federal sentencing guidelines. Journal of Service Research 42(1): 271–307.

Ashworth

(2009) Techniques for reducing sentence disparity. In Von Hirsch

Ashworth

Roberts

(eds) Principled Sentencing: Readings on Theory and Policy. Oxford and Portland, OR: Hart Publishing, 243–257.

Balák

(2007) Ke sjednocování soudní praxe Nejvyšším soudem. Soudce 92(4): 2–8.

Baráková

(2016) Stanoviska nejvyšších soudů: efektivní prostředek sjednocování judikatury nebo přežitek socialistické justice? Časopis pro právní vědu a praxi 2: 245–258.

Bates

(2014) Computational methods for mixed models. URL (accessed 16 April 2018): https://cran.r-project.org/web/packages/lme4/vignettes/Theory.pdf

Bates

, et al. (2015) Fitting linear mixed-effects models using lme4. Journal of Statistical Software 67(1).

Berghuis

(1992) De harde en zachte hand. Een statistische analyse van verschillen in sanctiebeleid. Trema 15(3): 84–93.

Berghuis

Mak

(2002) De straftoemeting in Nederland van 1995 tot en met 2001. The Hague: Ministerie van Justitie.

10.

Bolker

(2015a) Better convergence tests. URL (accessed 16 April 2018): https://github.com/lme4/lme4/issues/120.

11.

Bolker

(2015b) Lme4 convergence warnings: troubleshooting. URL (accessed 16 April 2018): https://rstudio-pubs-static.s3.amazonaws.com/33653_57fc7b8e5d484c909b615d8633c01d51.html

12.

Council of Europe (1992) Recommendation No. R (92) 17 of the Committee of Ministers to Member States Concerning Consistency in Sentencing.

13.

Council of Europe (1993) Consistency in sentencing: Recommendation to member states and explanatory memorandum. Criminal Law Forum 4(2): 355–392.

14.

Council of Europe (2014) Prison population rates for 2014. URL (accessed 16 April 2018): http://wp.unil.ch/space/space-i/prison-stock-on-1st-january/prison-stock-2014-2015/

15.

CrossValidated (2014) How scared should we be about convergence warnings in lme4. URL (accessed 16 April 2018): http://stats.stackexchange.com/questions/110004/how-scared-should-we-be-about-convergence-warnings-in-lme4

16.

Daly

Tonry

(1997) Gender, race and sentencing. Crime and Justice 22(1): 201–252.

17.

Dhami

Belton

(2015) Using court records for sentencing research: Pitfalls and possibilities. In Roberts

(ed.) Exploring Sentencing Practice in England and Wales. Basingstoke: Palgrave Macmillan, 18–34.

18.

Dušek

(2015) The effects of a simpler criminal procedure on criminal case outcomes: Evidence from Czech district-level data. Charles University, Center for Economic Research and Graduate Education. URL (accessed 16 April 2018): http://www.cerge-ei.cz/pdf/wp/Wp528.pdf

19.

Fearn

(2005) A multilevel analysis of community effects on criminal sentencing. Justice Quarterly 22(4): 452–487.

20.

Finch

Bolin

Kelley

(2014) Multilevel Modelling Using R. Boca Raton, FL: CRC Press.

21.

Fiselier

JPS

(1985) Regionale verscheidenheid in de strafrechtspleging II. Delikt en Delinkwent 15(4): 283–300.

22.

Gertner

(2014) How to talk about sentencing policy—and not disparity. Loyola University Chicago Law Journal 46(2): 313–326.

23.

Hawkins

(2003) Order, rationality and silence: Some reflections on criminal justice decision-making. In: Gelsthorpe

Padfield

(eds) Exercising Discretion: Decision-making in the Criminal Justice System and Beyond. Cullompton: Willan Publishing, 186–219.

24.

Heinz

(2013) Kriminalität und Kriminalitätskontrolle in Deutschland. Version 1/2016. URL (accessed 16 April 2018): http://www.uni-konstanz.de/rtf/kis/Kriminalitaet_und_Kriminalitaetskontrolle_in_Deutschland_Stand_2013.pdf

25.

Hodgson

Soubis

(2016) Understanding the sentencing process in France. Crime and Justice 45(1): 221–265.

26.

Hogarth

(1971) Sentencing as a Human Process. Toronto and Buffalo: University of Toronto Press.

27.

Hutton

Tata

(2002) The judicial role in the ‘balance’ between two visions of justice in sentencing. In: Tata

Hutton

(eds) Sentencing and Society: International Perspectives. Aldershot: Ashgate, 307–322.

28.

Johnson

(2006) The multilevel context of criminal sentencing: Integrating judge- and country-level influences. Criminology 44(2): 259–298.

29.

Karabec

, et al (2011) Criminal Justice System in the Czech Republic. Prague: IKSP. URL (accessed 16 April 2018): http://www.ok.cz/iksp/docs/386.pdf

30.

Killias

(1994) Sentencing reform – from rhetorics to reducing sentencing disparity. European Journal on Criminal Policy and Research 2(1): 19–28.

31.

Koons-Witt

(2002) The effect of gender on the decision to incarcerate before and after the introduction of sentencing guidelines. Criminology 40(2): 297–328.

32.

Krajewski

(2016) Sentencing in Poland: Failed attempts to reduce punitiveness. Crime and Justice 45(1): 175–218.

33.

Kühn

(2004) Worlds apart: Western and Central European judicial culture at the onset of the European enlargement. American Journal of Comparative Law 52(3): 531–567.

34.

Kühn

(2012) Role judikatury v českém právu. In: Bobek

Kühn

(eds) Judikatura a právní argumentace. Prague: Auditorium.

35.

Kuklík

(2015) Czech Law in Historical Contexts. Prague: Karolinum.

36.

Lappi-Seppälä

(2001) Sentencing and punishment in Finland. In: Tonry

Frase

(eds) Sentencing and Sanctions in Western Countries. Oxford: Oxford University Press, 92–150.

37.

Lappi-Seppälä

(2016) Nordic sentencing. Crime and Justice 45(1): 17–82.

38.

Lappi-Seppälä

Hinkkanen

(2004) Selvityksiä raiskausrangaistuksista ja rangaistuskäytännön yhtenäisyydestä. Helsinki: National Research Institute of Legal Policy. URL (accessed 16 April 2018): https://helda.helsinki.fi/bitstream/handle/10138/152557/TTA61_LappiSeppala_Hinkkanen_2004.pdf?sequence=1

39.

Mannouzzi

(2002) Are guided sentencing and sentence bargaining incompatible? Perspectives of reform in the Italian legal system. In: Tata

Hutton

(eds) Sentencing and Society: International Perspectives. Aldershot: Ashgate, 110–132.

40.

Merlo

, et al. (2006) A brief conceptual tutorial of multilevel analysis in social epidemiology: Using measures of clustering in multilevel logistic regression to investigate contextual phenomena. Journal of Epidemiology & Community Health 60(4): 290–297.

41.

Merrall

ELC

Dhami

Bird

(2010) Exploring methods to investigate sentencing decisions. Evaluation Review 34(3): 185–219.

42.

Miller

(2014) Sentencing reform ‘reform’ through sentencing information systems. In: Tonry

(ed.) The Future of Imprisonment. New York: Oxford University Press, 121–156.

43.

Monsieurs

Vanderhallen

Rozie

(2011) Towards greater consistency in sentencing: Findings from a survey of Belgian magistrates. European Journal of Criminology 8(1): 6–16.

44.

Myers

(1988) Social background and the sentencing behavior of the judges. Criminology 26(4): 649–676.

45.

Novotná

(2009) Podmíněné odsouzení v československém a českém trestním právu. Law Faculty, Masaryk University. URL (accessed 16 April 2018): https://is.muni.cz/th/143843/pravf_m/diplomova_prace.pdf

46.

Padfield

(2011) Time to bury the custody ‘threshold’? Criminal Law Review 8: 593–612.

47.

Pina-Sánchez

(2015) Defining and measuring consistency in sentencing. In: Roberts

(ed.) Exploring Sentencing Practice in England and Wales. Basingstoke: Palgrave Macmillan, 76–92.

48.

Pina-Sánchez

Linacre

(2013) Sentence consistency in England and Wales: Evidence from the Crown Court Sentencing Survey. British Journal of Criminology 53(6): 1118–1138.

49.

Pina-Sánchez

Linacre

(2014) Enhancing consistency in sentencing: Exploring the effects of guidelines in England and Wales. Journal of Quantitative Criminology 30(4): 731–748.

50.

Pina-Sánchez

Linacre

(2016) Refining the measurement of consistency in sentencing: A methodological review. International Journal of Law, Crime and Justice 44: 68–87.

51.

Plesničar

(2013) The individualization of punishment: Sentencing in Slovenia. European Journal of Criminology 14(4): 462–478.

52.

Reid

(2014) Revisiting leniency: An intra-provincial study of sentencing variation in British Columbia. Canadian Criminal Law Review 18(1): 39–56.

53.

Roberts

(2009) Structuring sentencing discretion. In Von Hirsch

Ashworth

Roberts

(eds) Principled Sentencing: Readings on Theory and Policy. Oxford and Portland, OR: Hart Publishing, 229–236.

54.

Roberts

(ed.) (2014) Mitigation and Aggravation at Sentencing. Cambridge: Cambridge University Press.

55.

Roberts

Von Hirsch

(eds) (2010) Previous Convictions at Sentencing. Oxford: Hart Publishing.

56.

Roberts

Hough

(2015) Empirical sentencing research: Options and opportunities. In: Roberts

(ed.) Exploring Sentencing Practice in England and Wales. Basingstoke: Palgrave Macmillan, 1–17.

57.

Roberts

Pina-Sánchez

(2015) Paying for the past: The role of previous convictions at sentencing in the Crown Court. In: Roberts

(ed.) Exploring Sentencing Practice in England and Wales. Basingstoke: Palgrave Macmillan, 154–172.

58.

Šámal

, et al. (2012) Trestní zákoník. Komentář, 2nd edn. Prague: C.H. Beck.

59.

Šámal

, et al. (eds) (2014) Trestní právo hmotné, 7th edn. Prague: Wolters Kluwer.

60.

Scheinost

(ed.) (2015) Trestní sankce – jejich uplatňování, vliv na recidivu a mediální obraz v televizním zpravodajství. Prague: IKSP. URL (accessed 16 April 2018): http://www.ok.cz/iksp/docs/423.pdf

61.

Scheinost

Válková

(eds) (2015) Sankční politika a její uplatňování, Prague: IKSP. URL (accessed 16 April 2018): http://www.ok.cz/iksp/docs/429.pdf

62.

Scheirs

Beyens

Snacken

(2016) Belgian sentencing as a bifurcated practice? Crime and Justice 45(1): 267–306.

63.

Scott

(2010) Inter-judge sentencing disparity after ‘Booker’: A first look. Stanford Law Review 63(1): 1–66.

64.

Smejkalova

(2009) The nature of unifying opinions of Czech Supreme Courts. SSRN Electronic Journal. URL (accessed 16 April 2018): http://www.ssrn.com/abstract=1695925

65.

Snijders

TAB

Bosker

(1999) Multilevel Analysis: An Introduction to Basic and Advanced Multilevel Modelling. Thousand Oaks, CA: SAGE.

66.

Sotolář

Púry

Šámal

(2000) Alternativní řešení trestních věcí v praxi. Prague: C.H. Beck.

67.

Spohn

(1991) Decision making in sexual assault cases: Do black and female judges make a difference? Women and Criminal Justice 2(1): 83–105.

68.

Sporer

Goodman-Delahunty

(2009) Disparities in sentencing. In: Oswald

Bieneck

Hupfeld-Hienemann

(eds) The Social Psychology of Punishment of Crime. Chichester: John Wiley, 379–401.

69.

Steffensmeier

Britt

(2001) Judges’ race and judicial decision making: Do black judges sentence differently? Social Science Quarterly 82(4): 749–764.

70.

Steffensmeier

Demuth

(2006) Does gender modify the effects of race–ethnicity on criminal sanctioning? Sentences for male and female white, black, and Hispanic defendants. Journal of Quantitative Criminology 22(3): 241–261.

71.

Steffensmeier

Hebert

(1999) Women and men policymakers: Does the judge’s gender affect the sentencing of criminal defendants? Social Forces 77(3): 1163–1196.

72.

Steffensmeier

Ulmer

Kramer

(1998) The interaction of race, gender and age in criminal sentencing: The punishment cost of being young, black and male. Criminology 36(4): 763–798.

73.

Supreme Prosecutor’s Office (2014) Zpráva o činnosti státního zastupitelství za rok 2013. URL (accessed 16 April 2018): http://www.nsz.cz/images/stories/PDF/Zpravy_o_cinnosti/2013/zoc_2013.pdf

74.

Tak

(2001) Sentencing and punishment in The Netherlands. In: Tonry

Frase

(eds) Sentencing and Sanctions in Western Countries. Oxford: Oxford University Press, 151–187.

75.

Tata

(2002) Accountability for the sentencing decision process – Towards a new understanding. In: Tata

Hutton

(eds) Sentencing and Society: International Perspectives. Aldershot: Ashgate, 399–428.

76.

Tata

(2007) Sentencing as craftwork and the binary epistemologies of the discretionary decision process. Socio-Legal Studies 16(3): 425–447.

77.

Tomášek

(ed.) (2010) Czech Law between Europeanization and Globalization. Prague: Karolinum.

78.

Tonry

(1996) Sentencing Matters. New York: Oxford University Press.

79.

Tonry

(2001) Punishment policies and patterns in western countries. In: Tonry

Frase

(eds) Sentencing and Sanctions in Western Countries. Oxford: Oxford University Press, 3–28.

80.

Van Wingerden

SGC

(2014) Sentencing in the Netherlands: Taking Risk-Related Offender Characteristics into Account. The Hague: Eleven International Publishing.

81.

Vávra

(2016) Estimating disparities in the treatment of foreign nationals in the criminal justice process in the Czech Republic. Master’s thesis, University of Economics, Prague. URL (accessed 16 April 2018): https://www.vse.cz/vskp/52350_estimating_disparities_in_the_treatment_of_foreign_nationals_in_the_criminal_justice_process_in_the_czech_republic

82.

Vibla

(2015) More than one crime: Sentencing the multiple conviction offender. In: Roberts

(ed.) Exploring Sentencing Practice in England and Wales. Basingstoke: Palgrave Macmillan, 221–234.

83.

Wasik

(2004) Going around in circles? Reflections on fifty years of change in sentencing. Criminal Law Review April: 253–265.

84.

Weigend

(2016) No news is good news: Criminal sentencing in Germany since 2000. Crime and Justice 45(1): 83–106.

85.

Weinstein

(2006) The historical roots of regional sentencing variation. Roger Williams University Law Review 11(2): 495–509.