NZ Police Dogs Codebook
This document describes how data was analysed to produce the charts and figures for the article We need to talk about police dogs.
This article relied on four sources of data. I used my Charter data analysis library to load the data where applicable, to process it, and to produce all the charts in this document.
TOR data records how NZ Police used force, with certain exceptions:
- Presentation of tasers or firearms by the Armed Offenders Squad (AOS) or Special Tactics Group (STG)
- Prior to 2018, events in which someone died
For information regarding how police officers record their use of force in the TOR database, please see the document Use of Force Overview - Reporting (PDF Link). This document was extracted from the NZ Police Intranet on 23 August 2019, and released by NZ Police under the OIA on 26 August 2019.
This guideline contains some important caveats regarding when police officers do and don't record certain types of force. In particular, uses of "open hand" don't include pushing a person unless that person falls, and uses of handcuffs are only recorded if it is used with "pain compliance" or if it is used alongside another reportable tactical option:
Employees who use force must report
Shows of force
- a TASER (i.e. presentation, laser painting or arcing)
- A firearm.
To be reported on: Standard TOR
Note: Excluding Armed Offenders Squad or Special Tactics Group shows of a TASER or a firearm and shows of force during training.
Uses of force
- communication, but only when used with one or more of the tactical options below, and only the first time it is used
- metal or plastic handcuffs, a waist restraint belt, and a leg restraint — vehicle:
- with pain compliance
- without pain compliance, but only when handcuffs, a waist restraint belt, and a leg restraint — vehicle are used with another reportable tactical option
- a spitting hood
- a restraint chair
- empty hand techniques excluding touching, guiding, escorting, lifting, and pushing where a person does not fall to the ground
- O C spray (spraying)
- a baton (striking)
- a weapon of opportunity, e.g. a Police torch
- a Police dog when the subject is bitten and/or otherwise injured
- a TASER by discharge and/or contact stun
- a sponge round
- a firearm (discharge).
To be reported on: Standard TOR
Notes:Use of Force Overview - Reporting (PDF Link) | NZ Police
Excluding any use of force during training.
Unintentional or unauthorised TASER or firearms discharges must be reported in an Unintentional/Unauthorised Discharge form.
However, despite the rules around when the use of handcuffs should be reported, NZ Police's annual TOR summary report covering 2019 explained that some non-reportable uses of handcuffs in 2018 had been recorded as TOR events:
The 2018 Annual TOR Report included 74 TOR events where the only tactical option used was handcuffs without pain compliance. Because these events had no reportable use of force, including them in the analyses inflated the total number of TOR events. The data has now been updated and comparisons to 2018 data use the updated values. The updated values differ from the 2018 report, but improve data accuracy and consistency with other years’ reports.Annual TOR summary report 2019 (PDF link) | NZ Police
Each individual TOR report records one officer using force against one person. If two officers use force against one person, this will be recorded as two separate TOR events.
The TOR data uses NZ Police's "Perceived Cumulative Assessment" (PCA) model of risk assessment to classify subject's levels of resistance. It is part of NZ Police's Tactical Options Framework (TOF), and provides a set of categories for police officers to use when making a risk assessment as part of NZ Police's "TENR" (Threat, Exposure, Necessity, Response) risk assessment model, also part of the TOF.
Perceived cumulative assessment
Your subjective assessment and continuous reassessment of an incident, using the TENR model, based on information known about the situation and the subject’s behaviour. The PCA may escalate and/or de-escalate more than once during an incident.
There are five categories in the PCA, which are represented in the TOF – cooperative, passive resistance, active resistance, assaultive, GBH or death.Tactical Options Framework (PDF Link) | NZ Police
The Tactical Options Reporting (TOR) data covering 2018 was released by NZ Police under the Official Information Act (OIA).
On 1 October 2019, I requested the following information from NZ Police under the OIA:
TOR data for the 1 January – 31 December 2018 reporting period, in the same format as was released for the July–December 2017 reporting period (ref IR-01-18-5190).OIA request: Tactical Options Reporting Data January-December 2018 | Mark Hanna
On 16 October, NZ Police extended the deadline of the request by 15 working days, to 20 November 2019, because:
the request necessitates a search through large amounts of information suchthat [sic] a proper response to the request cannot reasonably be made within the original time limit.NZ Police
On 9 December 2019, 13 working days after the extended legal due date, NZ Police refused my request unless I first paid them a charge of $836.00:
Given the amount of time it will take to provide this data to you, I have decided to implement a charge for your request as per the Ministry of Justice Guidelines (https://www.justice.govt.nz/assets/Documents/Publications/1982-Official-Information-Act-charging-guidelines.pdf). These state that time spent by staff searching for relevant material, abstracting and collating, copying, transcribing and supervising access, and undertaking necessary review where the total time is in excess of one hour, may incur a charge. Using the Ministry of Justice guidelines (Justice Ministry OIAS charging guidelines) at $76 per hour (with the third hour free) this would equate to $2432.00. However, based on a previous response to you, I would reduce this charge to $836.00 as previously advised.OIA response 9 December 2019 | NZ Police
This $836 figure was arrived at because it was the charge imposed when they refused my earlier request for TOR data covering January–June 2018 a year earlier. I had not paid this charge, so NZ Police had not released this information:
Based on the amount of time it has taken Police to prepare your two previous requests for the same information from an earlier time period, the likely time taken to respond to your current request is estimated at between 10-12 hours work. Using the Ministry of Justice guidelines (Justice Ministry OIAS charging guidelines) @ $76 per hour (with the first hour free) this would equate to a cost between $684 and $836.OIA response 20 December 2018 | NZ Police
Complaint to the Ombudsman
I believe this data should be released each year as a matter of public interest. Given I intend to request this data annually until NZ Police begins to release it proactively, paying hundreds or thousands of dollars each time is neither affordable nor sustainable. On 17 February 2020 I lodged a substantial formal complaint with the Office of the Ombudsman, asking that they review NZ Police's decision on my request and recommend that the charge be waived in the public interest:
I have requested that NZ Police release their "Tactical Options Reporting" data covering 2018. NZ Police has imposed a charge of $836 on the release of this data. I do not believe this charge they have calculated is reasonable, but more importantly I believe the charge they have imposed should be waived in its entirety due to the public interest in this information being released.OIA complaint - Police charge for 2018 TOR data | Mark Hanna
I've made my complaint to the Office of the Ombudsman publicly available, for anyone who would like to read it in its entirety: OIA complaint - Police charge for 2018 TOR data (PDF link)
I'd like to thank Andrew Ecclestone for his valuable assistance in putting it together.
Interim attempt — coordinated requests
On 19 June 2020, five NGOs who work in the justice sector each requested that NZ Police release the TOR data covering 2018 to them. The requests were worded to make it clear that they were asking for the same data NZ Police had agreed to release to me if I paid them $836. They also each made the case that releasing this data to them would be in the public interest.
This is a request under the Official Information Act 1982. Please provide a copy of the Tactical Options Reporting (TOR) data for the period 1 January - 31 December 2018.
For the purposes of section 16 of the OIA, our preference is to receive the data in a machine-readable format such as CSV or XSLX. The same XSLX format that was used to release the July - December 2017 data (your ref IR-01-18-5190) would be ideal. For the avoidance of doubt, we would like the same fields of data that were disclosed in response to that request to be included in your response to this request. If the Police also choose to publish this data on its website, please provide us with the link to the relevant web page.
On three previous occasions (your references 16/11336, 16/7758/11, and IR-01-18-5190), the Police has released TOR data under the Official Information Act for 6 month periods between 1 July 2016 and 31 December 2017. We understand that in each instance it did so without making a charge for this data.
The NGOs who requested this data from Police were:
- Amnesty International Aotearoa
- New Zealand Council for Civil Liberties (NZCCL)
- People Against Prisons Aotearoa (PAPA)
On 30 July 2020, 9 working days after the legal due date, NZ Police sent almost identical responses to each of these NGOs. In its response, NZ Police refused all five requests with the assertion that the requested data did not exist:
Tactical Options Reporting (TOR) data for the period 1 January - 31 December 2018 is a large and complex data set. It contains 4324 TOR reports (rows) and over 300 columns capturing information about each TOR event. (A TOR ‘event’ is the reportable use of one or more tactical options by one officer, against one individual.) It is not immediately apparent, nor straightforward how to analyse the data to ensure results are accurate and meaningful, (it is easy to generate an analysis that does not provide the intended information.) Given the complexity of the data, the subject matter experts in the Response and Operations; Research and Evaluation (RORE) team peer review complex analyses to ensure accuracy; wholesale release of the data would negate this quality checking mechanism. Data provided in the format requested is of little or no value to the public as there is a high risk of misinterpretation by people who are not familiar with the fields captured, and without any way to quality-check the information, it risks misleading the public about NZ Police use of force and misinformation becoming part of the public record.
Therefore your request is refused pursuant to section 18 (g) of the Official Information Act 1982 as the information requested is not held. That is, this information is not quantitatively recorded in police databases. Thus your request necessitates analysis of qualitative data to generate new data in the form that you request, and this data may not be complete.NZ Police
Chief Ombudsman's final opinion
On 6 November 2020, I received Chief Ombudsman Peter Boshier's final opinion on my complaint from February:
I am now able to advise you of my final opinion that Police’s decision to fix a charge for this request was unreasonable.
My opinion is that due to the nature of the information requested, Police should waive the charge in the public interest. As I expressed my opinion to Police:Chief Ombudsman Peter Boshier
Police use of force is a matter of perennial public significance and there is a substantial public interest in the availability of information which promotes accountability for the way Police exercise those powers. This necessarily requires a great degree of transparency in matters relating to the tactical options employed by Police and the way in which Police respond to incidents that do or may require force to be used.
… it is one of the Act’s purposes to ‘increase progressively’ the availability of official information. In light of this, and given that the issue of Police use of force is of great public significance, I do not consider that there should be obstacles to the release of information which promotes transparency in this area.
On 4 December 2020, NZ Police released their Tactical Options Reporting data covering 2018 to me:
I am writing to you with regards to the response you received dated 9 December 2019 (IR01-19-27258), for which you sought a review from the Office of the Ombudsman. After consultation and review, we have revised our response, and are providing you with the following Tactical Options Reporting (TOR) data spreadsheet.OIA response - Tactical Options Reporting Data January–December 2018 | NZ Police
Alongside the data, NZ Police included an explanation of some caveats, limitations, and metadata:
The records provided in the attached spreadsheet are derived from the Tactical Options Reporting database, which records the reportable use of force by NZ Police. Each row is one ‘TOR event’, which is the reportable use of one or more tactical options, by one officer, against one individual. Multiple TOR events can occur at one incident. The definition of reportable uses of force can be viewed in the Annual Tactical Options Research Reports available on the NZ Police website: https://www.police.govt.nz/about-us/publication/tactical-options-research-reports.
The data attached is for the 2018 calendar year. The data includes TORs which had completed the review process as of 16 May 2019, when data was extracted for the 2018 Annual Tactical Options Research Report. Efforts have been made to clean and correct any identified errors; however, inaccuracies may still exist. As such, the data may not be consistent with future reports.
Some columns have been removed from this dataset, in consultation with the Office of the Ombudsman. Please note that for some of the removed columns, the dataset retains an equivalent coded column that captures and categorises the original content without the risk of identification. The removed columns include:
OIA response 4 December 2020 (PDF link) | NZ Police
- columns with information that either directly identified individuals or contained free text fields which may have directly identified individuals (section 9(2)(a) of the Official Information Act 1982, to protect the privacy of natural persons);
- columns that were not identifying on their own, but in combination with data from other columns, increased the risk that the individuals involved may be identified. Removing these columns reduces the risk that individual persons can be identified, but this risk is not completely eliminated; NZ Police requests that you consider the risk to the privacy of individuals who have had force used against them when deciding how you will use and distribute the attached data (section 9(2)(a) of the Official Information Act 1982, to protect the privacy of natural persons);
- columns that include information that poses a risk to operational integrity and staff safety if it becomes public (section 6(2) of the Official Information Act 1982There is no section 6(2) of the Official Information Act 1982. NZ Police has clarified to me that they had intended to cite section 6(c) of the Official Information Act here.);
- columns which are blank or duplicate information from other columns.
NZ Police also asked that people either rely solely on the annual TOR summary reports they publish, or contact NZ Police directly with questions rather than attempting to analyse the TOR data themselves.
I will echo NZ Police's statement that the TOR data is complex, and that some additional information is necessary to ensure that it is interpreted correctly. I have done my best to provide as much of that additional information in this document as I can, but there are some parts of the TOR data that I cannot attempt to analyse because I do not fully understand them.
If you would like to validate any analysis you have done of this data that you are unsure about, I would encourage you to contact NZ Police to ask about it. Unfortunately, based on my experience, I would not say that you could expect a timely response, but you should still be able to rely on NZ Police to provide you with an accurate one. Depending on your circumstances, you may find it useful to use explore this raw data and do your preliminary analysis using it, and then follow up with NZ Police to confirm your interpretation.
I would also like to echo NZ Police's request that nobody make attempts to de-anonymise any of this data. Privacy is important, especially for vulnerable people and for young people and children. NZ Police have gone to significant, and reasonable, lengths to ensure the privacy of those people whom this data is about is protected.
Loading the data
NZ Police released the TOR data covering 1 January – 31 December 2018 in the following spreadsheet:
I converted this spreadsheet to a CSV using Google Sheets:
This data was loaded into Charter using this code:
For convenience, I added an ID column to each row. In previous datasets, NZ Police had provided an ID column, but there wasn't one in this data.
To make it easy to use this ID value to look up a row in the spreadsheet, I made it equal to the row's index plus 2 (1 to shift it from 0-based to 1-based, and 1 to account for the header row).
Recoding boolean columns
Some columns were coded with each cell being either blank or a
1. Others used the strings
"No", or in one case just
"Y". I recoded these columns to use
For some columns, the value was not relevant to each row. In these cases, I recoded it to
Cleaning aggregate columns
PCA_ALL columns were not coded in a way that made them suitable for processing. Each cell contains one or more values, but there is sometimes no delimiter between these values. One example, from the first row of the
LOCATION_TYPES column, is:
In this case, there are two consecutive values of
"Street/Highway/Motorway", but no clear way to separate them.
Some other values did contain a newline character as a delimiter. However, this was not consistent even within individual cell values. For example:
Outdoor public areaOutdoor public area\nStreet/Highway/Motorway
FIREARM_ALL_DEPLOYMENTS columns used a delimiting character that was also present within several values. One example is
Laser Painting Discharge
In this case, there are two values,
"Laser Painting" and
"Discharge", because the
"Laser Painting" value contains the character used to separate the two values, it requires a bit more processing to convert these cells' values appropriately into arrays.
To recode these columns, I use the enums collected from all the other columns of the same type, spread throughout the dataset, which each contained a single value.
Cleaning incorrect tactic entry
One row incorrectly recorded a use of the
"Communication" tactic as
"Communication " (there was an extra space), so I fixed it to make it consistent with the rest of the data.
This is what the data set looks like now that it had been cleaned.
Cleaning I haven't done
There are still some other data cleaning tasks that may be necessary for working with certain parts of this dataset. Here is a list of data quality issues I have noticed, which may need cleaning or might be indicative of mistakes in data entry:
- Several taser columns (e.g. those for the number of discharges with probes and contact stuns) have values of both
"-". NZ Police have confirmed that cells with either of these values both represent the same thing.
- One cell in the
FIREARM_1_CAPABLE_OF_THREAThas the value
"Immediate and sufficient", which NZ Police have confirmed should be in the "Firearm 1: Tactic Effect" column.
I created a utility
getTacticUseRows function for filtering in a way that looks for taser and firearm discharges but ignores their presentation.
I excluded the "Communication" tactic because it is not a use of force, and I only included uses of the "Taser" and "Firearm" tactics if they were discharged.
I sorted the list of tactics by the number of events at which they were used.
I created the graph by counting the number of events at which each tactic was used, excluding recordings of firearms or tasers being shown but not used.
I calculated the proportion of TOR events at which each tactical option was used, using the
getTacticUseRows function which collects rows representing events at which a tactical option was used.
This function includes uses of tasers and firearms when they were discharged, but ignores their presentation.
The percentages add up to more than 100% because more than one tactical option can be used at a single TOR event.
I created a function to collate how many subject injuries were recorded, grouped by cause and severity, based on the subject injury columns.
Because each TOR event records the use of force by one police officer against one person, it is possible that some of these injuries have been counted more than once if multiple police officers were using force against the same person.
I also used the injury severity categories that NZ Police described on page 10 of the 2018 TOR summary report:
Minor, moderate, and serious are proxy indicators of severity. Minor injuries are those that required no treatment or self treatment only; moderate injuries required medical treatment but not hospitalisation, and serious injuries required hospitalisation.2018 TOR summary report | NZ Police
This paragraph didn't specify how injuries with the
"At Scene" treatment required were coded, but based on the numbers reported it seems all but certain that injuries requiring this level of treatment were coded as moderate.
These are the numbers of subject injuries of each severity level inflicted by police in 2018:
Using the same sequence of labels as for the tactics used graph, so they could be compared more directly, I created a graph of moderate and severe injuries inflicted by police.
I also calculated the proportion of moderate and severe subject injuries inflicted by police attack dogs in 2018:
I looked at the rate at which each tactical option caused injuries of each severity by dividing the total number of injuries they caused at each level by the number of TOR events at which they were used.
In the 2019 TOR report, NZ Police said that "dog deployment is only reported as a tactical option if the dog bites or injures someone". This is not true.
For each use of attack dogs recorded in the TOR data, there is a corresponding "subject bitten" column. Though it is rare, incidents in which attack dogs were used but the subject was neither bitten nor injured are still recorded.
I looked at the rate at which the use of dogs caused injuries of each severity by dividing the total number of injuries they caused at each level by the number of TOR events at which they were used.
See Chart and Figure: Injuries for the code collating injuries by cause and severity.
I also looked at the proportion of injuries of each level of severity, out of all injuries inflicted by attack dogs.
I looked at the rate at which the use of the "empty hand" tactical option caused injuries of each severity by dividing the total number of injuries they caused at each level by the number of TOR events at which they were used.
See Chart and Figure: Injuries for the code collating injuries by cause and severity.
I also looked at the proportion of injuries of each level of severity, out of all injuries inflicted by the "empty hand" tactical option.
I looked at how often each relevant factor recorded in the TOR data was flagged for TOR events at which attack dogs were used.
Those relevant factors are:
- History of carrying weapons
- History of violence against police
- History of violence against non-police
- 1M (This is the code for a "mental health incident" incident type)
- Distressed not 1M
- Excited Delirium
Officers record one of five categories in the "Subject armed" field of a TOR report. This allows for differentiation of not just whether or not a subject was armed, but also if an officer believed them to be armed and whether or not the subject used a weapon.
The categories recorded in this field are:
- No Weapon
- No Weapon but Believed
- Had Weapon and Believed
- Had Weapon but Not Used
- Had Weapon and Used
It's not clear to me why an officer might record "Had Weapon and Believed", instead of recording whether or not that weapon was used.
These are the values recorded in this field for incidents at which police attack dogs were used:
Here's how those values compare with all TOR reports recorded in 2018:
Each time a police officer records using force against someone, they record a category of behaviour in NZ Police's Perceived Cumulative Assessment (PCA) risk assessment framework. This records their assessment of the subject's behaviour at the time they used a particular tactical option against them.
The categories in the PCA framework are, from least resistance to most:
- Passive Resistant
- Active Resistant
During a TOR event, an officer may record multiple different PCAs as the subject's behaviour changes and they become more or less resistant. I have looked at the highest PCA recorded when an attack dog was set on a subject, but in some cases the subject was more resistant at another point in time during the incident.
Here are the same figures for the entire TOR event.
NZ Police recorded subject ethnicity for each TOR event. TOR data released in previous years has been more complex, including entries that encompass more than one ethnicity. For example:
"Maori / European"
However, the TOR data for 2018 appears to have had ethnicity data recoded. NZ Police has clarified to me how this recoding was done:
Ethnicity classifications are based on the Statistics New Zealand Statistical Standard for ethnicity (ETHNIC05 v2) with the priority order Māori, Pacific peoples, Asian, MELAA, European, Other/Unknown. Each subject's ethnicity is coded to only one category based on this priority order.OIA response 25 February 2021 (PDF link) | NZ Police
These are the values recorded in the ethnicity column:
The "MELAA" value is a broad category that encompasses "Middle Eastern", "Latin American", and "African".
This is the distribution of ethnicities of the people who had attack dogs set on them in 2018:
As well as most uses of attack dogs in 2018 being against Māori, most of the moderate and serious injuries caused by those attack dogs were inflicted on Māori:
NZ Police's taser policy prohibits police officers from ever discharging a taser on a person below the "assaultive" threshold in their "Perceived Cumulative Assessment" (PCA) risk assessment model.
Always use a TASER in a manner consistent with the "Tactical Options Framework" and never 'use' in situations below the assaultive range, e.g. active or passive resistant.TASER (Electronic Control Devices) (PDF link) | NZ Police
However, there were some instances in 2018 at which police officers recorded themselves using a taser on someone below the "assaultive" threshold.
Unlike data released for previous years, NZ Police recoded the age column for 2018 into a set of "age group" categories.
The New Zealand Police National Youth Policy Plan 2005 – 2006 (PDF link) defines (p12) a child as being under 14 years old, and a young person being "aged between 14 and 16 years of age".
These definitions are also used the Children, Young Persons, and Their Families Act 1989, whereas the Oranga Tamariki Act 1989 extends the definition of a "young person" to include someone who is 17 years old.
Three of the age groups used by NZ Police in this data set are consistent with the definitions of a child being < 14 years old and a young person being 14–16 years old:
- < 10
In responding to follow-up questions, Police clarified what the various charge columns in the TOR data mean:
|3000||Drugs & Antisocial|
|A-W Traffic||Traffic offences|
These categories are further broken down on page 6 of this Police document: Understanding Recent Movements in Crime Statistics (PDF Link)
I looked at how often each category of charge was laid after TOR events, broken down by whether or not attack dogs were used:
I also looked at a breakdown of how often each type of charge was laid following TOR events where attack dogs were used:
And I looked at the same breakdown for TOR events where attack dogs were not used:
The charge of "disorder" (code 3500) was significantly more common than any others, so I calculated the percent of TOR events for which this charge was laid, broken down by whether or not attack dogs were used:
In table 1 of the 2016 use of force summary report (p2), NZ Police published a table of the number of TOR events at which each tactical option was used. Dogs were recorded as having been used at 323 TOR events.
In table 18 of the report (p8), NZ Police published a table of the number of subject injuries inflicted by each tactical option. Dogs were recorded as having caused 288 injuries.
In table 1 of the 2017 use of force summary report (p2), NZ Police published a table of the number of TOR events at which each tactical option was used. Dogs were recorded as having been used at 290 TOR events.
In table 18 of the report (p8), NZ Police published a table of the number of subject injuries inflicted by each tactical option. Dogs were recorded as having caused 256 injuries.
In table 4 of the 2019 use of force summary report (p24), NZ Police published a table of the number of TOR events at which each tactical option was used. Dogs were recorded as having been used at 375 TOR events.
In table 5 of the report (p25), NZ Police published a table of the number of subject injuries inflicted by each tactical option. Dogs were recorded as having caused 321 injuries.