Contact Crime Excess

Contact Crimes and COVID NPIs

Monthly South African crime statistics are released by SAPS on a quarterly basis. The latest quarterly statistics includes data for crimes reported up until end March 2021. This dataset includes complete monthly crime reporting data from January 2016 and includes data by crime reports by crime categories, province, district and police station. When looking at the aggregated reported crimes per month, it is clear that the COVID pandemic and mitigation interventions has had an impact on crime patterns.

Contact crimes

The reduction of contact crimes during the lockdown of March and April is significant, and the various categories showed similar reductions.

Estimating the impact of COVID on contact crimes reported.

The count of contact crimes reported by month during the period 2016-2019 shows definite seasonality, and it is therefore possible to create a model that takes into account that seasonality and fits the observed data fairly closely.

Prediction

From this fitted model a forecast/prediction for the period January 2020 to March 2021 can be built.

Calculating “Excess” contact crimes

The differences between the actual crimes reported for 2020 and 2021 and the mean prediction can be set as an “excess” of reports (i.e., how many more or fewer crimes were reported than predicted)

South Africa Contact Crimes: Excess versus Predicted values Jan 2020 - Mar 2021
2020 Jan	53,884	54,127	-243	1,304	-1,789
2020 Feb	55,436	53,006	2,430	3,994	867
2020 Mar	53,439	58,841	-5,402	-3,612	-7,191
2020 Apr	23,514	53,409	-29,895	-28,223	-31,566
2020 May	29,721	52,823	-23,102	-21,403	-24,800
2020 Jun	44,244	50,599	-6,355	-4,686	-8,023
2020 Jul	39,669	53,019	-13,350	-11,558	-15,141
2020 Aug	48,195	56,174	-7,979	-6,035	-9,922
2020 Sep	59,064	57,727	1,337	3,379	-704
2020 Oct	80,272	59,750	20,522	22,681	18,364
2020 Nov	60,696	60,141	555	2,773	-1,662
2020 Dec	65,469	68,955	-3,486	-893	-6,078
2021 Jan	37,888	54,127	-16,239	-14,163	-18,314
2021 Feb	52,065	53,006	-941	1,129	-3,010
2021 Mar	59,494	58,841	653	2,992	-1,685
Source: SAPS Quarterly crime statistics https://www.saps.gov.za/services/crimestats.php
Prediction based on ETS(M,N,M) model fit of crime reports 2016 - 2019

Creating metrics for COVID restrictions and observed impacts.

Various lockdown stages and associated restrictions were applied across the entire year, and in order to examine their relation to crime reporting, it is necessary to normalise the time period and to standardise their severity ((i.e. to apply the different measures that were applied on portions of months to a mean for the complete month in order to compare different periods)

Alcohol restrictions

In order to do this, alcohol restrictions have been divided into two basic types: a. Off-sales restrictions, and, b. On-premise restrictions.

Off-sales restrictions relate to sales through retail outlets, and the severity of the restrictions is measured by the permitted hours of sales divided by the potential hours of sales should there be no (additional to normal licensing laws) restrictions. The assumption is made that off-sales hours in a normally licensed period would be a maximum of 12 hours per day, for six days per week (i.e. Monday to Saturday 08h00 ~ 20h00). This is likely erring on the upside, and may understate the impact of COVID alcohol restrictions on contact crimes reported.

On-premise restrictions relate to sales through outlets where the alcohol is consumed on-site (e.g. bars, restaurants, etc.) The assumption is made that the normally licensed period would be a maximum of 16 hours per day, for seven days per week (i.e. Monday to Sunday, 10h00 ~ 02h00). Again, this is likely erring on the upside, as the majority of outlets would have significantly shorter hours/days of trade. Again, this may understate the impact of COVID alcohol restrictions.

In addition, since the crime data are monthly, it is not possible to measure any impact of the COVID restrictions by day or week. In particular, it is not possible to measure any kind of “weekend-effect”.

An overall restriction index can be built by combining and weighting the on-premise and off-sale indices by their estimated proportion of the market (using 2018 market data by Euromonitor) ¹

Govt mandated COVID restrictions. Monthly stringency indices.
2020 Jan	0.0000000	0.0000000	0.0000000	0.0000000
2020 Feb	0.0000000	0.0000000	0.0000000	0.0000000
2020 Mar	0.1935484	0.1935484	0.1935484	0.1935484
2020 Apr	1.0000000	1.0000000	1.0000000	1.0000000
2020 May	1.0000000	1.0000000	1.0000000	0.3750000
2020 Jun	0.7696923	1.0000000	0.5384615	0.0000000
2020 Jul	0.9344205	1.0000000	0.8685783	0.1612903
2020 Aug	0.7301487	0.6612903	0.7992832	0.2500000
2020 Sep	0.3598269	0.2041667	0.5161111	0.2194444
2020 Oct	0.2929327	0.1250000	0.4615385	0.1666667
2020 Nov	0.2801378	0.1250000	0.4358974	0.1666667
2020 Dec	0.3607537	0.2137097	0.5083871	0.1760753
2021 Jan	1.0000000	1.0000000	1.0000000	0.2083333
2021 Feb	0.3470278	0.2500000	0.4444444	0.2083333
2021 Mar	0.3408673	0.2500000	0.4320988	0.2083333
0 - No restriction, 1 - Full ban.

Mobility

The use of lockdowns and curfews to reduce mobility allows the creation of a curfew severity index. This is simply calculated by the proportion of restricted hours per day, averaged over the month.

An alternate method of measuring mobility is to use Google’s mobility data. Using data from mobile phones, Google supplies mobility data for many regions in the world. These are in the form of percentage change from a baseline in early 2020 (prior to COVID restrictions). These are supplied on a daily basis, and it is necessary to normalise them by month. This is done by simply taking the mean for the month and converting the percentage to a simple index between -1 and 1)

Correlation between restrictions and contact crime “excess”

[^1]: Passport: Alcoholic Drinks in South Africa. Euromonitor International, July 2019. Page 8, Table 10 ↩︎