How do defendants choose their trial court? Evidence for a heuristic processing account

Abstract

In jurisdictions with two or more tiers of criminal courts, some defendants can choose the type of trial court to be tried in. This may involve a trade-off between the probability of acquittal/conviction and the estimated severity of sentence if convicted. For instance, in England and Wales, the lower courts have a higher conviction rate but limited sentencing powers, whereas the higher courts have a higher acquittal rate but greater sentencing powers. We examined 255 offenders’ choice of trial court type using a hypothetical scenario where innocence and guilt was manipulated. Participants’ choices were better predicted by a lexicographic than utility maximization model. A greater proportion of “guilty” participants chose the lower court compared to their “innocent” counterparts, and estimated sentence length was more important to the former than latter group. The present findings provide further support for heuristic decision-making in the criminal justice domain, and have implications for legal policy-making.

1 Introduction

In several countries including Australia, Canada, and England and Wales there are two or more tiers of courts dealing with criminal cases (Association of Commonwealth Lawyers, 2012a-c). Typically, higher tier courts deal with more serious (indictable) offences whereas lower tier courts deal with less serious (summary) ones. However, there may be offences in-between that can be tried in lower or higher tier courts. For instance, in the English criminal justice system, defendants charged with what are called “triable either-way” offences,¹ such as actual bodily harm, burglary, supply of drugs and dangerous driving, can be tried either by a bench of three lay judges (or one professional judge) at the lower tier (magistrates’) court or by a jury of twelve members of the public in the presence of a professional judge at the higher tier (Crown) court.² The box labeled “Mode?” in Figure 1 shows the point in the court proceedings where the defendant’s choice occurs (as described in Section 20 of the Magistrates’ Courts Act, 1980). An important question that remains to be systematically researched is how defendants choose their trial court type.

Figure 1: Choice of trial court in the English criminal system

Note: The decisions made by the defendant during court proceedings are in bold. The defendant’s choice of trial court happens at an open hearing in the magistrates’ court, and the procedure is regulated by section 20 of the Magistrates’ Court Act 1980. The defendant decides which court will hear his/her case if, and when, he/she pleads not guilty. At this point, if the (lay) judges in the magistrates’ court decide they can try the case, the defendant can give his/her consent to do so, or the defendant can elect to have the case referred to the Crown court for trial. Alternatively, the magistrates’ court can direct the case to the Crown court and, if so, the defendant has no choice in the matter. Until recently, if a case had been tried and convicted in the magistrates’ court, it could still have been sent by the magistrates’ court to the Crown court for sentencing.

When making their choice, defendants might be influenced by common knowledge that the lower tier courts have a higher conviction rate, but limited sentencing powers which restrict the severity of sentence given. By contrast, the higher tier courts have a higher acquittal rate, but greater sentencing powers. In fact, statistics show that the acquittal rate is approximately 40% in the Crown court and 25% in the magistrates’ court (Home Office, 1998). Statistics also show that in 2009, the average custodial sentence given in the Crown court was 24.3 months compared to 2.7 months in the magistrates’ court (Ministry of Justice, 2010). For offences of comparable severity, the difference was slightly more marked, i.e., 25 months versus 2.6 months, respectively. The difference in sentence severity is largely attributable to limits on judges’ sentencing powers in the magistrates’ court that do not apply in the Crown court.³ Thus, the choice of trial court type involves a trade-off between probability of acquittal (favoring the Crown court) and estimated severity of sentence (favoring the magistrates’ court).

The question of how defendants choose their trial court type is of interest from both legal and psychological perspectives, and, although both perspectives might agree on the normative model that is applicable to this choice domain, they differ in their predictions of how defendants might make their choice. Next, we describe the normative model before introducing the different predictions informed by the legal and psychological perspectives.

In normative terms, rational choice under risk or uncertainty should involve the maximization of subjective expected utility (SEU; see Anand, 1995; Edwards, 1992; Fishburn, 1981). Thus, from a set of options, a decision-maker should choose one that maximizes SEU, where SEU refers to an additive integration of the subjective magnitudes of possible positive and negative outcomes weighted by their subjective probabilities of occurrence. With regard to the present study, the choice is between the magistrates’ court and Crown court, and the variables are probability of acquittal/conviction and estimated sentence severity if convicted. The SEU model for this choice task is as follows:

E(U⁻) = P(conviction) × E(months in prison if convicted)

where P and E refer to probability and expected value, respectively. Note that maximization in this context refers to choosing the option that minimizes E(U⁻) because it is defined in terms of expected disutility.

2 Perspectives on defendant’s court choice

2.1 Legal

As mentioned above, the legal and psychological perspectives differ in their prediction of how defendants might choose their trial court type. The prediction informed by the legal perspective is based on the assumption that people can and do maximize SEU. Indeed, some legal reformists have portrayed defendants as tactical, arguing that they may “play the system” in order to obtain the best outcomes for themselves (Home Office, 1998, p. 3). There is concern that defendants who are guilty may choose the higher tier court because of its greater acquittal rate or may choose the lower tier court because of its more lenient sentences. This feeds public perception that defendants are “getting away with it” or “getting off lightly”, respectively, and has led legal reformists to question whether defendants should have the right to choose their trial court (Auld, 2001; Home Office, 1998).

However, to-date, legal and criminological scholars have not rigorously tested the assumption that defendants maximize utility when choosing their trial court type, although they have demonstrated that defendants are aware of the trade-off between probability of acquittal/conviction and estimated sentence severity if convicted, and would wish to exploit the benefits of each type of court. Bottoms and McClean (1976) found that, among a sample of nearly 300 male defendants appearing at court, those who chose to be tried in the magistrates’ court said they wanted to “get a lighter sentence.” By contrast, those who chose to be tried in the Crown court said they wanted a “better chance of acquittal” (see also Gregory, 1976). Hedderman and Moxon (1992) interviewed approximately 280 defendants and 100 of their solicitors (legal representatives) in Crown courts. When given a list of reasons for preferring each court type, the majority of both samples rated “better chance of acquittal” as a “very” or “fairly important” reason for favoring the Crown court (see also Riley and Vennard’s 1988 study of defense representatives). Defendants’ preferences for trial in the magistrates’ court were mostly based on the expectation of a lighter sentence.

2.2 Psychological

The prediction informed by the psychological perspective on how defendants may choose their trial court type is based on empirical findings demonstrating that, rather than use all relevant and available information when making their choice, people often employ heuristic strategies that rely on a small subset of information (e.g., Gigerenzer, Hertwig, & Pachur, 2011; Gigerenzer, Todd, & the ABC Group, 1999; Gilovich, Griffin, & Kahneman, 2002; Kahneman, Slovic, & Tversky, 1982; Nisbett & Ross, 1980). Heuristic processing can be explained by the fact that human cognition is limited in terms of attention, memory and processing capacity, and that tasks may place further constraints on people’s cognitive functioning by, for example, inducing cognitive stress (e.g., Hammond, 2000; Payne, Bettman, & Johnson, 1993; Simon, 1990).

The lexicographic (LEX) model is a prominent example of a heuristic model (see Fishburn, 1974). This describes a non-compensatory choice process, i.e., where a high value attached to one piece of information does not compensate for a low value attached to other information. In a choice task, LEX involves assessing alternatives in rank order of the importance of their attributes. If there is a dominant alternative (i.e., one with a positive feature on the most important attribute), then that is chosen. If not, then the decision-maker assesses the next most important attribute, and so on, until a dominant alternative is found. In the present case, there are two alternatives (magistrates’ court and Crown court) and two attributes (acquittal/conviction probability and sentence length if convicted).

Research comparing the predictive validity of non-compensatory and compensatory models has found support for the former. For instance, Rieskamp and Hoffrage (1999) reported that LEX was the best predictor of students’ choices of which companies had the highest annual profit, when deciding under high time pressure. Similarly, Bröder (2000) found that, in an artificial choice task, students were more likely to switch to a strategy akin to LEX, called Take-the-Best, when investment costs were associated with obtaining information. Finally, of particular relevance, recent research shows that offenders can be better described by non-compensatory than compensatory models when deciding about burglary (Garcia-Retamero & Dhami, 2009; Snook, Dhami, & Kavanagh, 2011), shoplifting, forgery, drugs, and drinking and driving (Dhami & Mandel, 2012).

3 The present research

Thus, in the present study we examined whether defendants choose their trial court type in a manner that maximizes SEU or according to a heuristic model, namely LEX. Following past psychological research on human choice, we would expect a significantly greater proportion of participants’ court choices to be better modelled by LEX than by the SEU model.

In addition, the claim that defendants may behave “tactically” in choosing a type of trial court implies that they might condition their choice on particulars of their case. For instance, the idea that offenders are “getting away with it” suggests that compared to their innocent counterparts, guilty defendants might be more motivated to maximize their chances of acquittal. Accordingly, they may also be more likely to favor the Crown court, which offers a better chance of acquittal than the magistrates’ court. By contrast, the idea that offenders are “getting off lightly” suggests that guilty defendants may be more motivated than innocent defendants to minimize the severity of their sentence (if convicted) and thus be more likely to favor the magistrates’ court, which has limited sentencing powers. We also tested these related hypotheses.

Importantly, the methodology adopted in the present study differs from past studies on this issue so as to more rigorously examine defendants’ choice of trial court type. In the past, researchers have interviewed defendants (or their legal representatives) about their reasons for choosing a particular court type. We did not specifically ask participants for their reasons because post-hoc self-reports of reasons cannot be used to estimate participants’ SEUs for each court, and people’s self-reported reasons for their choices are often inaccurate indicators of true causes (Nisbett & Wilson, 1977). Rather, we used a hypothetical, but realistic, case in which offender guilt was experimentally manipulated, and where offenders’ quantitative estimates of the probability of acquittal/conviction and sentence length for the two types of court were obtained before their statements as to which court type they would prefer to be tried in, and whether probability of acquittal/conviction or sentence length (if convicted) was the more important factor in their choice. This enabled us to assess the extent to which defendants choose in light of their own subjective assessments of acquittal/conviction rates and sentence severity for the two types of court using both the SEU and LEX models. This also allowed us to examine the effect of truth (i.e., guilt vs. innocence) on choices.

4 Method

4.1 Participants

Participants were 255 (152 female and 103 male) offenders accommodated in three English prisons. They were chosen because they are representative of the population of people who may face the choice of trial court type (Hedderman & Moxon, 1992). Eighty-three percent of the sample described their race as White. The mean age of the sample was 33.32 years (SD = 9.53). The majority (64.2%) had no more than a secondary-school education. The sample comprised prisoners who had been charged with crimes against the person (35.8%), property crimes (22.6%), drugs crimes (16.0%), and other crimes (25.6%).

4.2 Procedure and materials

An experienced probation officer unconnected to the prisons collected the data. Prisoners were individually introduced to the study and invited to participate. They were given no incentives and were instructed that they would not face any negative consequences for declining to participate. They were also told that they could withdraw from the study at any stage. Confidentiality and anonymity of responses was assured. Booklets containing the experimental material were distributed and collected during afternoon “lock-up” hours. These were self-administered by prisoners in their cells (all of whom were in single cells), and took approximately 15 minutes to complete.

In the booklet, participants were asked to imagine they had been charged with committing actual bodily harm and had no prior arrest (see Appendix). Truth was manipulated by randomly assigning half of the sample to a guilt condition in which they were asked to imagine that they were in fact guilty of the charge, and the other half to an innocence condition in which they were asked to imagine that they were innocent. The hypothetical case contained the sort of legal representative’s information that is representative of that found to be given in the court (Gregory, 1976; Hedderman & Moxon, 1992). That is, in the case the representative correctly informs the defendant that there is a greater chance of conviction in the magistrates’ court than the Crown court, but that if convicted the maximum sentence in the Crown court is 5 years (or 60 months) in prison, whereas it is 6 months in the magistrates’ court.

After reading the case, participants responded to items imagining that they were the defendant. Participants first rated the probabilities of acquittal and conviction separately for each court type on a 101-point scale from 0% to 100% with numeric anchors at 5% intervals. Given that some studies (e.g., Macchi, Osherson, & Krantz, 1999; Mandel, 2005, 2008) have shown that people’s probability assessments of complementary events, such as acquittal and conviction, are not additive (i.e., they do not sum to a probability of 1 or 100%), we asked participants to rate both acquittal and conviction so that we could ascertain whether their assessments were coherent. If they were coherent, their estimates should sum to 100% on the relevant scale. After providing their probability assessments, for each court, participants estimated the length of the prison sentence they would likely receive if convicted on a 12-point scale from 0-3 months to 54-60 months (which is the maximum penalty for this offence). Next, participants were asked to indicate the court type in which they would choose to be tried and possibly sentenced. Finally, participants indicated which attribute—probability of acquittal or likely length of sentence—was more important in their choice of court. The order of both probability rating (acquittal, conviction) and court type (magistrates’, Crown) were counterbalanced across participants.

5 Results

5.1 Probability of acquittal and conviction

Although the emphasis of this research is on choice, it is useful to begin with an analysis of probability, given that the latter is central to the assessment of the former. The additivity property requires that the probabilities assigned to conviction and acquittal outcomes sum to unity (100% on the relevant scale). To examine if judgments were coherent in this regard, the sum of the probabilities was tested separately for each court type with a one-sample t-test (using a comparison value of 100). Summed probability judgments for the magistrates’ court did not significantly differ from additivity in either direction, M = 99.40, SD = 25.47, t(254) = −0.38, p = .71. However, the summed probabilities for the Crown court were significantly subadditive, M = 103.44, SD = 24.71, t(254) = 2.23, p < .05.

For the purposes of subsequent regression analyses, all probabilities were first adjusted such that values of 0 were changed to 2.5 (out of 100). Then, the probability of conviction in each court was “coherentized” by dividing the probability of conviction estimate by the sum of probability of conviction and the probability of acquittal.⁴ Table 1 presents the mean probability judgments of acquittal and conviction in each court (including the coherentized conviction probabilities) and the mean estimated sentence length in each court. As predicted, participants assessed the probability of acquittal as being more favorable in the Crown court than in the magistrates’ court, t(253) = 4.93, p < .001, Cohen’s d = 0.36. Conversely, participants assessed the probability of conviction as being less severe in the Crown court than in the magistrates’ court, t(253) = 2.20, p < .05, Cohen’s d = 0.17. Likewise, the coherentized conviction probabilities significantly differed between the two court types, t(253) = 4.19, p < .001, Cohen’s d = 0.32. Thus, there was a small, yet statistically significant, effect in the predicted direction.

Table 1: Mean subjective probability of acquittal and conviction and mean sentence length by court type

Coherentized conviction probabilities also differed by truth condition (see Table 2). These probabilities were significantly greater in the guilt condition than in the innocence condition, both for assessments of the magistrates’ court (t[252] = 3.00, p < .005, Cohen’s d = 0.38) and the Crown court (t[252] = 4.55, p < .001, Cohen’s d = 0.57).

Table 2: Mean coherentized conviction probability by court type and truth condition

5.2 Estimated sentence length

To examine estimated sentence length, the midpoint of the range indicated for each response option was taken as the estimated number of months for the relevant court. For instance, if the magistrates’ court was assigned an expected sentence length of 0-3 months, then 1.5 months was entered as the expected value. In line with the UK government statistics noted earlier, participants assigned a significantly longer expected sentence length to the Crown court than the magistrates’ court, t(253) = 16.36, p < .001, Cohen’s d = 1.31 (see the last row of Table 1 presents for the relevant descriptive statistics). In contrast with the finding that truth influenced probability assessments, there was no significant effect of truth on estimated sentence length for either court, ps > .25.

5.3 Attribute importance

When indicating the most important consideration in making their court choice (“importance”), 69.3% of participants were primarily concerned with receiving a shorter sentence, while the remaining 30.7% were primarily concerned with being acquitted. Importance was related to truth: whereas 62.7% of participants in the innocence condition were mainly concerned with getting a shorter sentence, as many as 76.7% had similarly prioritized sentence length as their main concern in the guilt condition, χ²(1, N = 254] = 5.82, p < .025, φ = .15.

5.4 Court choice

Overall, and consistent with the UK national trend(Auld, 2001), a significant majority (69.0%, 99% CI[61.2, 75.9]) of the sample indicated they would choose to be tried and possibly sentenced in the magistrates’ court. Court choice was independent of gender (χ²[1, N = 255] = 1.84, p = .18), age (r = −.07, p = .29, where the magistrates’ court is dummy coded with the higher value), whether the participant described themselves as White or another race (χ²[1, N = 247] = 0.12, p = .73), and whether or not participants had post-secondary education (χ²[1, N = 243] = 0.00, p = .96).

We assessed the validity of the SEU model predicting court choice using binary logistic regression. The dependent variable (choice) was dummy coded 0 for the magistrates’ court and 1 for the Crown court. Two log ratios were entered as predictors:

Plog = Ln(P(C_M)/P(C_C)

Slog = Ln (L_M/L_C)

where C stands for conviction, L stands for estimated sentence length, and the subscripts M and C refer to magistrates’ court and Crown court, respectively. Both predictors were significant: for Plog, B = 1.59, SE = 0.35, Wald = 20.99, p < .001; for Slog, B = 0.58, SE = 0.18, Wald = 10.95, p < .005.

Although participants were sensitive to both probability of conviction and estimated sentence length, there was still a significant preference toward choosing the magistrates’ court that was left unaccounted for by the predictors, as indicated by a significant intercept in a logistic regression of choice on Slog and Plog: B = −0.55, SE = 0.23, Wald = 5.74, p < .025. In addition, the Slog/Plog ratio of regression weights indicates that the utility model is optimized by a power function on sentence length equal to .37, consistent with a concave utility function implied by diminished sensitivity to increasing sentence length.

To probe the preference in choice further, the regression analysis was rerun with importance included in the model, along with Plog and Slog. Importance was dummy coded −1 for “sentence length” and 1 for “conviction”. By coding importance in this manner, the analysis addresses the counterfactual question, what would the constant have been like had there been an equal proportion of participants who indicated each attribute as their most important? Given that participants were 2¼ times more likely to say that reduced sentence length was of primary importance, and given that participants assigned a significantly lower sentence length to the magistrates court, one might expect this control procedure to eliminate the significance of the intercept. Importance was, in fact, a significant predictor, B = 1.22, SE = 0.17, Wald = 49.39, p < .001. More importantly, the intercept was no longer significant, B = −0.27, SE = 0.26, Wald = 1.08, p = .30.

Whereas 78.3% of participants in the guilt condition chose the magistrates’ court, 60.4% did so in the innocence condition, χ²(1, N = 254) = 9.45, p < .005, φ = .19. Given this contingency, a final regression model that included truth (coded −1 for guilt and 1 for innocent), along with Plog, Slog, and importance, showed an almost-significant effect of guilt, B = 0.31, SE = 0.18, Wald = 3.08, p = . 080. Thus, although choice was related to truth, the predictive effect of truth was weaker than the other predictors.

The preceding analyses clearly reveal a preference toward choosing the magistrates’ court, and a preference towards treating reduced sentence length as the most important attribute in arriving at a choice of court. The analyses also reveal that the SEU model cannot fully account for these preferences. Taken together, the preferences indicate that participants might be using LEX to arrive at their choice. That is, they tend to choose the magistrates’ court, which, on average, is judged to be more favorable than the Crown court in terms of sentence length. Thus, for most participants, LEX would predict that choice would be based on a comparison of expected sentence length, with consideration of probability of conviction only in the case of ties or for those who regard that as the primary attribute. Participants were therefore categorized in terms of whether or not their choice conformed to the prediction of the SEU model and/or LEX. These models were calculated using the untransformed probabilities of conviction and untransformed utilities.

Out of the 254 participants, 5 showed no preference by either model. That is, they assigned equal probability of conviction and equal sentence length to the two types of court. Nevertheless, four out of these five participants chose the magistrates’ court. Even with such a small number of cases, one can be 80% confident that the proportion choosing the magistrates’ court is greater than fifty-fifty; that is, the 80% CI = [51.4%, 93.8%]. Of the remaining 249 participants, LEX correctly classified 211 participants’ choices (84.7%), whereas the SEU model correctly classified 189 (75.9%).⁵ The difference in proportions (8.8%) is significant, 95% CI = [1.8%, 15.5%]. In terms of effect size estimation, φ = .634 for LEX and φ = .379 for the SEU model. Williams’ (1959, Steiger, 1980) test of dependent correlations shows that these two estimates differ significantly, t(245) = 6.00, p < .000001. The association between LEX and the SEU model used in the former calculation is φ = .64. In percentage terms, there was agreement between LEX and the SEU model in 216 (86.7%) cases, of which 84.7% correctly predicted choice. Among the remaining 33 cases where there were disagreements, LEX correctly predicted 28 (84.8%) and the SEU model the remaining 15.2%. Given the success rate of LEX in the agreement-with-SEU and disagreement-with-SEU subsets was virtually identical, it seems quite likely that participants in the former subset ought to be classified as following LEX rather than the SEU model.

6 Discussion

The question of how defendants choose their trial court type is important because the public and legal reformists are concerned that defendants may “play the system” in order to obtain the best outcomes for themselves (Home Office, 1998, p. 3). In order to prevent “manipulation of the justice system by [guilty] defendants”, the British Government has made repeated, but as yet unsuccessful, attempts to remove the defendant’s right to elect for jury trial in the Crown court (Auld, 2001; Home Office, 1998, p. 4, words in brackets added), particularly when the magistrates’ court is willing to try the case (see Figure 1). This effort has been particularly fueled by the belief that offenders are “getting away with it”; they are choosing the Crown court because its acquittal rates are higher than the magistrates’ court. Although previous research involving interviews of defendants and solicitors in court has demonstrated the importance of acquittal/conviction rates as well as sentence severity (Bottoms & McClean, 1976; Gregory, 1976; Hedderman & Moxon, 1992; Riley & Vennard, 1988), it has not been able to rigorously measure the relative importance of these two variables. Neither has past research been able to demonstrate how these variables inform the choice of trial court type; namely how defendants deal with the trade-off between them.

By contrast to the legal reformists’ assumption that defendants can and do maximize SEU when choosing their trial court type, the present study revealed that offenders’ choice of trial court type was better predicted by a LEX than a SEU model. According to LEX, most participants’ chose between the two types of court on the basis of estimated sentence length. Since the magistrates’ court has limited sentencing powers and thus less severe sentences, participants had a preference for this lower tier court, rather than the Crown court which has unlimited sentencing powers. Our findings thus extend past research on this issue by showing that likely severity of sentence is, in fact, more important than probability of acquittal/conviction in defendants’ choice of trial court type.

As a heuristic model, for most participants who placed greater importance on sentence length, LEX ignored information on probabilities of conviction in the two types of court. The finding that offenders do not maximize SEU in a task that is representative of one they might face and which involves only two options may be surprising. However, the present finding is compatible with psychological research showing that people, including offenders, use non-compensatory heuristics to make choices (e.g., Broder, 2000; Dhami & Mandel, 2012; Garcia-Retamero & Dhami, 2009; Rieskamp & Hoffrage, 1999). Nevertheless, future research could study the issue using methods where decision-makers face real consequences (see Kassin & Kiechel, 1996) and those that incentivise participants (see Ortmann & Hertwig, 2001). Here, one might predict participants would be more likely to follow a model that weights and integrates information on conviction rates and sentence severity.

In an effort to further extend previous research on choice of trial court type, the present study also examined whether offenders are tactical in their choice of trial court type in a case-contingent manner. We found that whether the defendant in the hypothetical scenario was in fact guilty or innocent influenced offenders’ choice of trial court type. Specifically, “guilty” offenders were significantly more likely to choose the magistrates’ court than their “innocent” counterparts. In addition, “guilty” participants judged the probability of conviction to be significantly greater in both the magistrates’ court and Crown court compared to their “innocent” counterparts. And, although there was no significant effect of truth condition on participants’ estimates of sentence length if convicted in both types of court, “guilty” participants were significantly more likely than “innocent” ones to state that sentence severity was the most important attribute.

The findings pertaining to truth condition, therefore, provide stronger support for the view that guilty defendants may be “getting off lightly” than “getting away with it”. This is compatible with research showing that defendants with previous convictions (a possible proxy for guilt) are more likely to report “better chance of lighter sentence” as a reason for their choice when choosing the magistrates’ court (Bottoms & McClean, 1976; Gregory, 1976).

Although not a main focus of the present study, our findings also contribute to the literature on additivity in probability judgment. Based on some prior research (e.g., Macchi et al., 1999; Mandel, 2005), one might have expected to observe superadditivity of probability judgments—namely, judgments that sum to less than one. However, participants’ judgments of the magistrates’ court, on average, were additive, whereas their judgments of the Crown court were significantly, albeit weakly, subadditive. Overall, then, there was not much evidence for coherence violation in participants’ probability judgments. This may be due, in part, to the fact that the complementary judgments were elicited in immediate succession—a condition that Mandel (2005) has shown attenuates coherence violations in probability judgment.

The findings of the present study also shed some light on other aspects of the English criminal justice system. First, the fact participants in the guilt condition perceived a greater likelihood of conviction in both court types compared to their innocent counterparts suggests that offenders have some faith in the ability of the court system (lay and professional judges as well as juries) to make accurate decisions at trial. This is compatible with research showing that the general public believes that jury decision-making is accurate (e.g., Roberts & Hough, 2009). It may also explain why “guilty” participants, in the present study, were more likely to choose the magistrates’ court; trying to get off lightly.

Second, the present findings suggest that the actual differential acquittal rate between the magistrates’ court and Crown court may not be because juries are more lenient than judges (see Hans & Vidmar, 2003 and Kalven & Zeisel, 1966). Rather, the difference may be a function of innocent defendants being more likely than guilty ones to choose the Crown court whom juries consequently acquit.

Finally, the finding of a concave utility function for estimated sentence length in our test of the SEU model, suggests that imprisonment (or at least the perception of it) might be subject to the principle of diminishing sensitivity, regardless of any adaptation effects. This may affect the effectiveness of deterrence-based sentencing policies that rely on perceptions of punishment as much as experience of punishment (see Paternoster, 2010). In terms of choice of trial court type, it may be useful to compare the choices of those who have been exposed to prison and those who have not in order to better understand how estimated sentence length might have affected choice, since research has shown differences in the experiences of first-time and recurrent inmates (e.g., Souza & Dhami, 2010).

6.1 Legal reform?

Legal reformists argue that the courts, and not defendants, are best qualified to determine where the case should be tried (Home Office, 1998). However, the courts may use extra-legal factors in their decision-making and thus make poor decisions on the behalf of defendants (Cammiss & Stride, 2008). We suggest that the idea that defendants may be tactical in choosing their trial court type is not a sound basis for legal reform. It seems entirely reasonable that, given a choice between two tiers of courts, defendants would choose the one they thought was better for them. Indeed, psychologists offer prescriptive theories and tools in an effort to improve individual judgment and decision-making (see Baron & Brown, 1991). Similarly, procedural justice in law highlights the importance of giving defendants (who should be presumed to be innocent until proven guilty) the right to choose trial by jury.

Perhaps a better argument for legal reform rests with the unfairness resulting from discrepancies between the two tiers of courts where differences in outcomes appear to be arbitrary. Efforts should be made to reduce these discrepancies, partly by ensuring that decision-makers follow comparable processes (e.g., sentencing guidelines), and also by reducing the differences in sentencing powers between the two courts (at least for triable-either way offences).

In conclusion, being given the option of trial court type is considered to be important by defendants (Gregory, 1976), and the issue of who is in the best position to make this choice is a contentious one. To-date, groups supporting defendants’ rights and representatives of the legal professions have strongly and successfully opposed any proposed changes to remove a defendant’s right to choose trial court type in triable either-way offences. However, their arguments have often highlighted the dearth of research on how individual defendants make this choice. The present study sheds some light on this issue.