Appendix: Detailed comments

Extra / Saturday, October 7th, 2023

In her article “Singapore’s death penalty for drug trafficking: What the research says and doesn’t”, Mai Sato reviewed six studies conducted by or for the Ministry of Home affairs about the use of the death penalty for managing drug trafficking. This Appendix provides her additional comments.

Study 1

Survey on Singapore Residents’ Attitudes Towards the Death Penalty 2021

In Study 1, the Ministry carried out a survey during March-May 2021, with 2,000 respondents aged 15 and above, and used stratified random sampling based on age, race, gender and citizenship status.

Study 2

Public Perception towards Singapore’s Anti-drug Policies (pages 52-54) 

Study 2 was a survey carried out during July-October 2018 with 2,000 respondents using random probability sampling.

The Ministry of Home Affairs asked respondents the extent to which respondents agreed with the statement “Death penalty is an appropriate punishment for drug traffickers who traffic a large amount of drugs” (emphasis added). 70% of respondents “strongly agree[d]” or “agree[d]” with the statement, 18 of respondents were “neutral”, and 12% “strongly disagree[d]” or “disagree[d]” with the statement (figures rounded up). 

Study 2 also surveyed minors aged 13 and above. While the study does not provide a breakdown of support for the death penalty by age group, it compares attitudinal differences between those aged 13-30 and those above 30. Support for the use of the death penalty for trafficking a “large amount” of [illicit] drugs was lower by over 20 percentage points in younger respondents (53%) than in older respondents (75%). 

Study 3

IPS Study on Attitudes towards the Use of Capital Punishment

In this study, a survey was carried out during October 2019-January 2020, using stratified random sampling, resulting in a sample of 2,000 respondents aged 18 and above.

 Strongly agreeAgreeNeutralDisagreeStrongly disagreeRefuse to answer
‘I believe the death penalty should be mandatory for a person convicted of intentionally trafficking a substantial amount of drugs’19%43%21%11%5%3%
Table: Support for the death penalty for drug trafficking using a 5-point Likert scale. Notes: Figures are rounded and do not add up to 100%.  [Source: Study 3] 

Study 4

Perception of Residents in Regional Cities on Singapore’s Crime Situation, Law and Safety

In Study 4, telephone interviews were carried out in 2018 and 2121, using a larger sample size of 7,221 respondents than in studies 1, 2 and 3. However, this sample is not a probability sample and is unlikely to be representative of the regions surveyed, with youth respondents (aged 20-29) overrepresented. 

Study 5

The Impact of Deterrence on the Decision-Making Process of Drug Traffickers (pages 65-80)

Phase 1 of the study was qualitative in nature, with interviews (n=22) with individuals convicted of drug trafficking in the last two years and focus group discussions carried out with 28 offenders convicted of various offences not limited to drug related offences. 

Phase 2 of the study was a survey of 298 male prisoners, with 163 convicted of drug trafficking and the remaining 134 of other offences. Phase 1 of the study showed that there are plenty of convicted drug traffickers who did not think rationally following the decision-making model outlined above (see quotes on page 70 of the study). The study also argued that “non-drug traffickers” did not engage in drug trafficking because they were risk aware. However, it is unclear from the study if these “non-traffickers” were tempted to traffic drugs in the first place, and what questions were posed to these individuals. Phase 2 of the study is also difficult to interpret because it is unclear how the independent variables in the regression model were coded. The questionnaire used in Phase 2 is also not made available in the study. 

Study 6

Deterrent Effect of Historical Amendments to Singapore’s Sanction (pages 56-64)

Study 6 used a technique called difference-in-differences. It is a quasi-experimental design that enables causal inferences to be drawn when randomisation is not possible. The technique compares the before-after difference in the experiment group’s outcomes (first difference) with the before-after difference in the control group (second difference). The impact of the intervention is calculated by subtracting the second difference from the first. In Study 6, the experiment group was cannabis trafficking cases. The study failed to use opium trafficking cases as an additional experiment group due to insufficient number of cases. Heroin trafficking cases were used as the control group. In sum, the study compares offence rates for cannabis trafficking, for which the mandatory death penalty was introduced in 1990, with offence rates for heroin trafficking, for which the mandatory death penalty was introduced in 1975.The before-after period was a 4-year window before the 1990 amendment and 4-year window after the amendment. 

A preliminary analysis comparing before-after changes for cannabis trafficking (i.e., focusing only on the experimental group) showed that the number of individuals convicted for trafficking cannabis remained the same (N=101). The proportion of individuals who trafficked cannabis above the capital threshold (500g) decreased by 1 percentage point, from 25% before the 1990 amendment to 24% after the amendment. The average net weight of cannabis trafficked decreased by 456g from 1,559g to 1,103g; however, this change was not statistically significant (Table 1). 

Similarly, a preliminary analysis for opium trafficking (again only looking at the experimental group) showed that the number of individuals who trafficked cannabis increased from 25 individuals during pre-amendment to 28 during post-amendment (Table 2). The proportion of individuals convicted of opium trafficking above the capital threshold (1,200g) decreased by 2 percentage points (80% during pre-amendment; 78% during post-amendment). The reduction in 27,000g of average net weight was statistically significant: the amount reduced from 40,700g pre-amendment to 13,700 post-amendment.  

 4-year window before the 1990 amendment4-year window afterthe 1990 amendmentDifference between the two periods
Number of individuals who trafficked cannabis1011010
Proportion of individuals who trafficked more than 500g of cannabis25%24%-1%
Average net weight trafficked 1,559g1,103g-456g
Differences between the two time periods were not statistically significant.
Table 1: Cannabis trafficking  [Source: Study 6]
 4-year window before the 1990 amendment4-year window afterthe 1990 amendmentDifference between the two periods
Number of individuals who trafficked opium2528+3
Proportion of individuals who trafficked net weights above 1,200g of opium80%78%-2%
Average net weight trafficked 40,700g13,700g-27,000g**
**Differences statistically significant at .05 level.
Table 2: Opium trafficking [Source: Study 6]

As noted above, the difference-in-difference analysis was not carried out for opium. The difference-in-difference regression modelling for cannabis trafficking, using heroin trafficking as a control group “suggested that the introduction of MDP [mandatory death penalty] for cannabis might have reduced the probability that cannabis traffickers would choose to traffic above the capital threshold for cannabis” (emphases added). The study also concluded that individuals who had prior convictions for drug offences were less likely to traffic cannabis in amounts above the capital threshold. 

Study 6 acknowledges that preliminary analyses concerning the changes in trafficking behaviour for cannabis and opium, respectively, do not demonstrate the deterrent effect of the legislative amendments. Indeed, the number of convicted cannabis traffickers remained the same, but the number of convicted opium traffickers increased after the introduction of the mandatory death penalty, on which the study is silent in the main body of the article. The only statistically significant difference was observed in opium trafficking, in which the average net weight of opium trafficked reduced (Table 2). However, there are issues associated with this analysis. 

First, assuming that a t-test was done to reach the conclusion that the difference in weight pre- and post-amendment was statistically significant (something not made clear in the article), the study does not provide the standard deviation of the trafficked weight of opium. The study should include at least the standard deviation and perhaps other measures of dispersion to allow the readers to assess if the statistically significant difference is a meaningful difference in the context of the arguments made. 

Second, despite the reduction, the average weight during post-amendment (13,700g) remains well above the capital threshold (1,200g). Third, the study could have looked at the number of cases that trafficked just below the capital threshold in post-amendment cases. Such analysis would have hinted at the possibility that traffickers were well aware of the new capital threshold for mandatory death penalty. 

Finally, broadly speaking, such preliminary analysis ignores the fact that many other factors, beyond the legislative amendment, could have caused the change in the average weight of trafficked opium. For example, one could argue that the demand for opium simply declined during the period in question, resulting in the decline in the average net weight of trafficked opium recorded. Equally, there may have been changes in police and prosecution practice that resulted in changes in numbers of convictions.

Next, it is worth examining the difference-in-differences analysis. It is unclear why the study chose to have the heroin trafficking cases as the control group. One of the key assumptions for using the difference-in-differences analysis is that trend in the control group approximates what would have happened in the experiment group, in the absence of the intervention — referred to as the parallel trend assumption. For example, the demand or supply in cannabis (experiment) and heroin (control) are assumed to behave in the same way — other than the effects of the legislative amendment–but this assumption is never mentioned or examined. Given the experiment and the control groups deal with different illicit drug markets, it is entirely possible that their demand or supply behave separately (no correlation) or behave according to a third variable (partial correlation) or even in opposite directions (inverse correlation). 

Assuming that heroin trafficking is appropriate as a control, we move to the main findings of this study. It argues that: 

The findings from the ‘difference-in-differences’ analyses …suggested that the introduction of MDP [the mandatory death penalty] for cannabis might have reduced the probability that cannabis traffickers would choose to traffic above the capital threshold for cannabis in the four years immediately following the change in the sanction regime by around 15 to 19 percentage points…Our analyses (results not shown in Table…) also suggested that traffickers who had prior convictions for drug offences were less likely to traffic cannabis in amounts above the capital threshold. 
 Model 1 (without trafficker characteristics)Model 2 (with trafficker characteristics)
Reduction in probability of trafficking cannabis with net weights above 500g-15 percentage points-19 percentage points
Statistical significance  10% level5% level
Table 3: Difference-in-Differences estimates for cannabis trafficking (using heroin trafficking as a control group) – Source: Study 6

The quote and the information presented in Table 3 fail to provide information that would assist the reader to appreciate the conclusion reached. The study states that the legislative amendment likely reduced the probability of cannabis being trafficked above the capital threshold by 15 to 19 percentage points. However, it is unclear what the probability was for both time points (pre- and post-amendment) for cannabis (experiment) and heroin (control) in calculating the overall impact. 

In addition, it is unusual that the 15 percentage points reduction in trafficking above the capital threshold for cannabis was reported as statistically significant at 10% level. Scientists typically conclude that an effect is significant at 5% level (p< 0.05) or less. While the significance level value should be set within the research context, no justification is given to the departure from the norm. In addition, the claim that “traffickers with prior convictions for drug offences were less likely to traffic cannabis in amounts above the capital threshold’ is not backed up by data (regression model): the study even acknowledges that the “results [are] not shown”. 

The above findings, if trusted, come with the following assumptions and limitations of the study. 

• The legislative amendment is used as the intervention that deters people from trafficking instead of executions, given Singapore does not publicise executions. It is assumed that the legislative changes were well known to would-be traffickers because such changes were, according to the study, “typically published through multiple mediums by the Singapore Government”. The study does not go into detail about the efforts made by the government back in 1990 to inform the public of the change, and it is unclear what efforts were made to disseminate the information outside of Singapore when internet was not widely available, given trafficking is not committed solely by those within Singapore. 

• The impact of the legislative change measured is a 4-year period, meaning that the findings, if taken for granted, can only speak to the deterrent effect during this period. The study refers to the model being tested for a 5-year period “yield[ing] similar findings”, but such results are omitted from the study. This means that the amendment may have had a moderate deterrent effect during 1990-1995, according to the study. The impact of the amendment beyond that period—or now in 2023—remains untested. 

• The impact of the legislative amendment, using a quasi-experimental design, is limited to cannabis trafficking. This means that the findings cannot be extended to other illicit drugs: morphine, diamorphine, opium, cannabis resin, cannabis mixture, cocaine, or methamphetamine. 

• The impact of the legislative amendment on cannabis trafficking is limited to using heroin trafficking as a control. We do not know the impact, if any, of the legislative amendment with other illicit drugs selected as controls.    

• Final limitation of the study is that data on drug trafficking relate only to cases that led to convictions, not to drug trafficking carried out.

In sum, this study, written by staff at the Ministry of Home Affairs and published by the government’s Home Team Journal: By Practitioners, for Practitioners, shows very little, if any, evidence of a deterrent effect of the mandatory death penalty for drug trafficking.