Toolkit for Quality Assessment

of Administrative Data
for Official Statistics

December 2022
02 Toolkit for Quality Assessment of Administrative Data for Official Statistics
Eastern and Southern Africa
Toolkit for Quality Assessment
of Administrative Data
for Official Statistics

December 2022
UN WOMEN and UNSD: Toolkit for Quality Assessment of Administrative Data for Official Statistics,
version 01. By Silja Emmel, December 2022

Layout and design: Conrad Mudibo

UN Women East and Southern Africa Regional Office

UN Gigiri Complex, UN Avenue;
Block M, Ground Floor
P.O. Box 30218- 00100 Nairobi, Kenya
Tel: +254 20 762 4778

africa.unwomen.org
Email: [email protected]
unwomenafrica
unwomenafrica
unwomen
unwomenafrica
 Table of Contents

Acknowledgement VII

Glossary VIII

Part 1 1 | Quality Assessment of Administrative Data 1

1 Introduction 2
2 Objective 3
3 Approach 4
4 Administrative data and official statistics 5
5 Administrative data and quality assessment 6
6 Users and administrative data quality 7
7 Administrative data and disaggregation 8
8 Administrative data and bias 9
9 Different types of administrative data 10
10 The statistical value chain and assuring administrative data quality 13
11 Metadata 15
12 Quality dimensions for administrative data 17
References 19

Part 2 | Template for Assessing the Quality of Administrative Data 20

Step 1: The Statistical System, Institutional Framework and Agreements 23

Step 2: Input data quality (Source) 25
Step 3: Quality and data processing 27
Step 4: Ensuring quality of statistical outputs 29
Other useful information 32

Toolkit for Quality Assessment of Administrative Data for Official Statistics V

Eastern and Southern Africa
 Part 3 | Country examples and checklists 35

1 GSBPM and administrative data 36

2 Assessing and improving the statistical system and institutional environment 37
3 Assessing and improving source data quality 38
4 Assessing and improving data quality during processing 43
5 Assessing and improving the output quality 46
6 Prioritising 48

VI Toolkit for Quality Assessment of Administrative Data for Official Statistics

Eastern and Southern Africa
Acknowledgement

The author would like to thank the Charlotte Juul Hansen, NBS Bhutan, NSA
following people (and organisations) for Namibia, Leandre Foster Ngogang Wandji,
their assistance and inputs in preparing the Ximena Clark, Statistics Norway, Matthias
report: Reister and the UN NQAF secretariat, as
well as the representatives from the NSOs
Isabella Schmidt and other UN Women
of Ethiopia, Kenya, Malawi, Zimbabwe and
colleagues, Vibeke Oestreich Nielsen
Lesotho that provided input during the
and the UNSD team, the Collaborative
regional meetings.
on Administrative Data and its members,

Toolkit for Quality Assessment of Administrative Data for Official Statistics VII
Eastern and Southern Africa
 Glossary
Administrative Administrative data are data collected by government institutions
Data as part of their routine operations (in this document we exclude
the private sector admin data). The administrative data have not
originally been collected for production of statistics.
Data holder The data holder is the institution that collects and stores the
administrative data from the source (e.g., the source can be a
citizen, company).
Data Source The data source is in this context either a file or a database that
contains the administrative data collected by the respective public
institution that is being examined as a source for the production of
official statistics.
Disaggregation Disaggregation is the breakdown of observations, usually within a
common branch of a hierarchy, to a more detailed level to that at
which detailed observations are taken. An example are individuals
that are registered with a unique personal identifier.
Metadata Metadata are a description of the data. There are structural and
reference metadata. Structural metadata provides identifiers and
descriptors of the data. Without it, reading the data is difficult.
Reference metadata describes the content and quality of the
statistical data.
National The national statistical system (NSS) is the ensemble of statistical
Statistical organisations and units within a country that jointly collect, process,
System and disseminate official statistics on behalf of national government.
Operational Use In this document operational use refers to the data holder using the
administrative data they collected for their routine operations, as
opposed to the purpose of statistical production.
Official Official statistics are statistics published by government agencies or
Statistics other public bodies such as international organisations.
Quality The assessment of the quality of input data, data processing and
Assessment statistical outputs. During the assessment, potential challenges to
the quality can be identified and an improvement plan developed,
ideally jointly by data holder and NSO.
Quality The longer-term and more circular process, taking the information
Assurance gathered during the different phases of quality assessment,
systematising it with the purpose to ensure the longer-term
assurance of quality.
User The user in this document is the on interested in using the final
statistical product that is produced with the help of administrative
data.

VIII Toolkit for Quality Assessment of Administrative Data for Official Statistics
Eastern and Southern Africa
 Part 1
Quality Assessment
of Administrative Data

Toolkit for Quality Assessment of Administrative Data for Official Statistics 1

Eastern and Southern Africa
 1 Introduction

The interest in, and use of administrative official statistics) needs to invest work in
data as an additional source for the ensuring that the quality of the source data
production of official statistics has been is good. Essentially, how suitable is it for
increasing in recent years. Administrative producing national statistics that match
data is the data produced by public the need of the users. Additionally, there
institutions as part of their operations. will also be work in aligning concepts and
Every official department under a ministry definitions used by the different institutions
maintains certain databases and reports, nationally, and with international standards.
private companies might collate some
Many countries are already working with
information based on regulations, laws, etc.
administrative data to a larger or smaller
Although their main purpose is not statistical
extend. There are a number of good
data production, all this data contains
practices and experiences from these
valuable information that can potentially be
countries that can prove valuable for
used for statistical production.
other countries in a similar situation. The
Especially in situations like a global development of the toolkit was started
pandemic and the high costs of surveys due to a large number of requests from
and censuses in combination with limited member countries of the Collaborative on
state budgets, but also when needing more Administrative Data (CAD). This coincided
frequent and more disaggregated data on with a strong interest in the topic in the
the whole population for policymaking, and Southern and Eastern Africa region from
planning, administrative data have great countries covered under the UN Women
potential. Count project. Efforts could be joined in the
development of the toolkit.
One of the big challenges when using
administrative data for statistics is, that the The focus of this toolkit is very practical.
data has not originally been collected for the This first part provides the reader with
production of statistics. The collection and background information, the second part
processing of administrative data is beyond a structure and explanation of the excel
the control of the NSO. It is the data source template for carrying out the quality
holder who manages these aspects. The assessment of a (new) administrative
same is true for the units and variables an data source. The third part contains
administrative data source contains. These examples from countries. Next to
are defined by administrative rules and may providing a template structure for the
therefore not be identical to those required assessment work, the aim of this document
by an NSO (Daas et al 2009). is to provide inspiration in form of a
questionnaire, a collection of tools and
This means that next to agreements
practical country examples, for inspiration,
with the public institutions that collect
ready to be adjusted to the respective
the information as part of their routine
country’s national context.
operations, the NSO (or other producer of

2 Toolkit for Quality Assessment of Administrative Data for Official Statistics

Eastern and Southern Africa
 2 Objective

The purpose of this guidance document administrative data quality, for example
and template is to serve as a tool for under the leadership of the UN NQAF
carrying out quality assessments of group.
administrative data for the production of
The tool is divided into three parts:
statistics.
Part 1 provides background that explains
It is intended for all countries that do
how the template may be used and what
not have their own quality assessment
needs to be considered when carrying out
system yet and want to start working
quality assessment of administrative data
with quality assessments of administrative
for the production of official statistics.
data sources. The written documents
and questionnaire can also be used as Part 2 is a template for quality assessment
additional inspiration for those countries of administrative data, with suggestions
that already carry out quality assessments. for structure and content of both, as well
The tool is ideally a living document, as further explanations of each section on
amended with the learnings and the process.
development of the users, the feedback
received from various expert groups, Part 3 contains country examples for each
ideally in close cooperation with other part of the process.
activities that focus on the assessment of

Toolkit for Quality Assessment of Administrative Data for Official Statistics 3

Eastern and Southern Africa
 3 Approach

The tool is practically oriented and aims working with quality assessment for
at providing ready to use guidance, both, the NSO as well as the MDAs that
templates and examples for countries hold data and sometimes also produce
that want to assess data quality of statistics.
administrative data sources for the
It has been developed following a review
production of statistics.
of the international and regional standards
To make the tool most relevant for and guidelines for quality assurance of
the countries requesting a tool, there official statistics and specifically official
have been group discussions that were statistics based on administrative data.
attended by several countries as well as
The first time an administrative data
individual calls with the NSOs that joined
source is assessed, it is recommended that
the group sessions. This was to learn
the assessment is done in collaboration
about their current practices, hear about
between the NSO and the administrative
examples and best practice and learn
data holder to establish a common
about the countries’ specific needs. The
understanding of why the data has been
meetings helped identify best practices as
collected, what it is normally being used
well as the countries’ most urgent needs
for and also learn about each other’s
when working with the assessment of
concepts and definitions.
administrative data quality.
The self-assessment tool proposed
A first draft version of the questionnaire
here does not result in a score or
has been tested in country during
recommendation but can serve as a
workshops with NSO and a number of
basis for evaluation by the tool user. It is
data holders participating. The feedback
the tool user’s own interpretation of the
of in-country testing was used to further
answers—and the tool user’s prioritization
adjust and improve the tool.
of which ones are especially important for
The aim is to strongly focus on countries the data at hand—that constitutes the tool
expressed needs. The toolkit includes user’s own assessment of data quality.
practical examples and templates and is
However, in a situation of scarce resources,
intended to be used by NSOs and data
both in staff and budget, that many
holders (Ministries, Departments and
NSOs and also Ministries, Departments
Agencies (MDAs) MDAs), ideally jointly.
and Agencies (MDAs) are faced with,
The practical orientation and examples prioritising the most needed actions might
are meant to provide an easier start when be a necessity.

4 Toolkit for Quality Assessment of Administrative Data for Official Statistics

Eastern and Southern Africa
4 Administrative data
and official statistics

Administrative data is not originally being No matter if for operational use or

collected for the production of statistics. statistical, applying statistical methods to
The institution collecting the data is doing administrative data can help improve the
so as part of its regular operations, for quality of the data, which is also beneficial
operational use – e.g., for the monitoring when used solely for operational purposes.
service delivery, the administration of The National Statistical Office can support
programmes, taxes and more. the data-collecting Ministry, Department
or Agency with data analysis to identify
When wanting to use administrative
systematic errors (e.g., systematic under-
data for official statistics, a number of
registration of certain groups, missing
conditions need to be fulfilled, to be in
values in certain periods of the week/
line with international guidelines and
year). This can then be addressed through
recommendations. This is not necessarily
training, resulting in better data quality.
the case when the administrative data
is being used for operational purposes. This document focusses on assessing
Regarding data quality, the more the data the quality of administrative data for
is being used for operational purposes the production of official statistics. Part
by different institutions, and especially of the framework might be usable for
if there are financial implications like administrative data for operational use,
benefits or levies linked to the data, it will but this is not the focus of this guidance.
most likely be checked more thoroughly.

Toolkit for Quality Assessment of Administrative Data for Official Statistics 5

Eastern and Southern Africa
 5 Administrative data
and quality assessment

Because administrative data is not collected but to pragmatically use what is available
for statistical production, as mentioned and identify the potential problems as
above, you (the statistics producer) are not accurately as possible and describe them
in control of the data collection. However, for the users. I.e., data disaggregated
administrative data is a great data source by gender is often not readily available
for producing official statistics and should and by using administrative data sources
be exploited as much as possible. As with gender relevant data can be leveraged for
other data sources being used to produce statistical production.
official statistics such as surveys, you need
Essentially, you want to know how suitable
to consider the quality of the administrative
the administrative data is for producing
data and how well the data fits the purpose
statistics and whether it is fit for purpose.
you want to use it for.
All in all, fit for purpose is the key in
As with other data sources, perfect does assessing the quality of administrative
not exist. There will most likely always be data for producing official statistics. To be
some problems with the data, e.g., not able to do that, knowing your users and
total coverage, under-reporting, over- what they need is essential. “Purpose” is
reporting (More detailed explanation of defined by the users and their needs.
the terms can be found in part 2). The aim
should not be a perfect set of source data

6 Toolkit for Quality Assessment of Administrative Data for Official Statistics

Eastern and Southern Africa
6 Users and administrative
data quality

Statistics are being produced to be used. Ideally, you will engage with your users
Learning about the “purpose” means regularly instead of just once. This makes
knowing your users and what kind of it possible for the you to learn about
statistical product, frequency, format and changes in user needs, be able to commu-
more they need. As this is a precondition nicate potential challenges in production/
for being able to properly assess the dissemination and respond to user feed-
quality of a statistical product, it all begins back.
with user mapping and engagement.
Quality is subjective depending on the
Engaging with your users can seem user of the statistical product and their
daunting because you do not know needs. For one user the quality might be
what they want and might also meet good while for the other it is not. You will
dissatisfaction and expectations that are need to find a compromise between user
hard to match, especially in the beginning. interests, and what is within the realm of
However, through dialogue, you will learn your available resources..
more about the needs and will also be
able to find compromises in terms of
expectations and what is possible.

Toolkit for Quality Assessment of Administrative Data for Official Statistics 7

Eastern and Southern Africa
 7 Administrative data
and disaggregation

Administrative data has great of inequalities are better brought to the

potential when seeking higher levels of surface and analysed.
disaggregation as well as full or close to
Vulnerabilities can arise through i.e.,
full coverage of the population.
gender, disability, LQBTQIA+ and more.
Following the UN Glossary of classification Statistics should be able to identify
terms, disaggregation is the breakdown who the vulnerable, disadvantaged,
of observations, usually within a marginalized, or socially excluded are;
common branch of a hierarchy, to a more how many they are; and where they are
detailed level to that at which detailed located. This is the minimum amount of
observations are taken. With standard information needed for making policies
hierarchical classifications, statistics and designing effective intervention
for related categories can be split programs for marginalized groups.
(disaggregated) when finer details are
Administrative data is often collected at
required and made possible by the codes
individual level and ideally also accessible
given to the primary observations. An
for the statistics producer at individual
example are individuals that are registered
level (in anonymised form). Next to
with a unique personal identifier.
being able to provide a more accurate
Aggregation loses information in the data, and granulate picture of the country
as aggregates often hide the disparities for policymaking, having access to
that exist between population groups. In disaggregate datasets means being able
contrast, disaggregation involves breaking to carry out more thorough error checking
down data into smaller information units. and other quality controls. For some of
When data are sufficiently disaggregated, the quality checks, disaggregation is a
multidimensionality and intersectionality precondition.

8 Toolkit for Quality Assessment of Administrative Data for Official Statistics

Eastern and Southern Africa
 8 Administrative data and bias

An important aspect you want to consider You can perform analysis regularly to
when using administrative data for the better understand the administrative data
production of official statistics and not only and identify missing values and analyse
operational use, and therefore mentioned whether there is a pattern in them, e.g., a lot
here separately, is bias. The data is being of missing values in a certain region/ sub-
generated through people and businesses region/ age group/ gender/ethnic group.
registering certain information. However,
Unlike with surveys, when you work with
certain groups might for many different
administrative data, there is no control over
reasons not register.
who is reporting their data. Thus, there
Both, over- and under-coverage can lead might parts of the population (especially
to bias in the data, which leads to certain vulnerable ones) that do not at all or to
groups/regions/other being under/over- a lower degree enter their data into the
represented or missing in the statistical system and therewith are not represented
product. or under-represented in the data

Toolkit for Quality Assessment of Administrative Data for Official Statistics 9

Eastern and Southern Africa
 9 Different types
of administrative data

When it comes to quality assessment and https://www.unescap.org/sites/default/

considerations not all administrative data d8files/event-documents/UNECE_
sources are alike. As mentioned, of course Developing_SBR_Stats_Cafe_14Jun2021.
the users define what good quality is. pdf
But the different types of administrative
https://www.inegi.org.mx/eventos/2021/
data can have different issues that affect
wiesbaden/
their quality. Depending on the type of
data there might be a high likelihood of https://cros-legacy.ec.europa.eu/content/
bias, under-representation (for example in use-administrative-and-accounts-data-
sensitive areas like gender-based violence, business-statistics_en
a high number of survivors will not show
up in registration) or over-representation https://statswiki.unece.org/download/
(e.g., when registration is linked with a attachments/185794796/Quality%20
financial gain or loss like losing a pension Indicators%20for%20the%20GSBPM%20-
when registering a dead relative). Administrative data for vital statistics
There are a number of different guidance and censuses
documents available for different thematic There are several publications on the use
areas that can help guide the attention to of administrative data for the production
some of the things that can affect quality of vital statistics and in censuses.
in the specific thematic area. Below some The purpose of the publications by
examples. UNECE are to guide national statistical
Administrative data and Business offices and other responsible agencies
Registers on the use of registers and administrative
(The Wiesbaden Group) data in population and housing censuses,
including operational, practical, technical,
The Wiesbaden Group as well as Eurostat, and legal aspects. Additionally, there
the OECD, UNECE and others have is a publication on the assessment of
collected knowledge and conducted quality of administrative data for the
research on using administrative data for use in censuses. The guide has been
business registers and the production of developed drawing on quality frameworks
business statistics. The Wiesbaden Group and best practices adopted by NSOs
is also currently testing a self-assessment across the world. It aims to lead census
model for Statistical Business Registers practitioners through the practical
that is soon ready for piloting. stages of assessment; from working
with an administrative data supplier to
Some publications on SBR and
understand the strengths and limitations
administrative data can be found here:
of a source, all the way to the receipt and
analysis of the actual data. The Guidelines

10 Toolkit for Quality Assessment of Administrative Data for Official Statistics

Eastern and Southern Africa
cover key quality dimensions on which among subpopulations. This granular data
an assessment is made, using a variety facilitates understanding of who is being
of tools and indicators. The Guidelines left behind or excluded from services and
are based on four Stages: Source, Data, why (Gerardo et al., 2018) and provides
Process, and Output, with the first two the locally relevant and actionable data
Stages being the principal focus of the needed to implement gender-responsive
Guidelines. services and programs at the subnational
level (Demombynes & Sandefur, 2014).”
UN ESCAP has investigated what
countries do in the ESCAP region and a Find the full model here:
publication describes how countries use https://data.unicef.org/resources/the-
administrative data in censuses in the administrative-data-maturity-model-
region. adamm/

Find the full documents here: Administrative data for gender

https://unece.org/DAM/stats/ statistics
publications/2018/ECECESSTAT20184.pdf UN Women/WHO guidance on VAW
https://unece.org/sites/default/ Admin data
files/2021-10/ECECESSTAT20214_WEB.
Administrative data on violence against
pdf
women (VAW) is collected when survivors
https://repository.unescap.org/rest/
and perpetrators of violence interact
bitstreams/aa806dbb-8ec1-45d6-8c42-
with hotlines, police and courts, health
5610c8d314c1/retrieve
systems, shelters, and other services. The
Administrative data for statistics on collection and use of high-quality VAW
children administrative data is crucial to inform the
UNICEF Administrative Data Maturity Model policies and programmes developed by
(ADaMM) governments to prevent and respond to
VAW.
Gender data, and the gender statistics
derived from them, are essential for The VAW administrative data can help to:
assessing how effectively we are achieving
equitable outcomes for boys and girls. Not • Understand which survivors are seeking
only do they help to track progress, but services because of violence;
they also identify gaps – telling us where • Estimate the need for such services and
more work and focus are needed. Yet, their costs;
gaps in gender data, vis-à-vis availability, • Understand the need for training among
granularity, timeliness, and adherence service providers; and
to international standards, compromise • Monitor service delivery and outcomes to
the ability of countries to design gender- improve quality and identify unmet needs.
responsive policies (UNICEF 2022, Inter- VAW administrative data can also provide
Agency working group). insights into who is not accessing services,
“Timely and granular disaggregated data helping to make sure no one is left behind
are also needed to guide planning and as we work to end VAW and to provide
service provision at the local level (Dincu timely information for decision-making and
& Malambo, 2019), with national statistics planning. Improving the collection and use of
often hiding entrenched inequalities administrative data on violence against women
is relevant for diverse sectors and contexts.

Toolkit for Quality Assessment of Administrative Data for Official Statistics 11

Eastern and Southern Africa
 The UN Women/ WHO guidance (link below) Member States to regularly collect,
identifies eight steps for improving the analyse and publish statistical data on
collection and use of VAW administrative violence against women, with a special
data. focus on administrative data on the
gender-related killing of women and girls
Find the full guidance here: (femicide/feminicide)
https://www.unwomen.org/en/
digital-library/publications/2022/05/ The document provides a comprehensive
global-technical-guidance-for- statistical framework for measuring the
collection-and-use-of-administrative-data- gender-related killing of women and
on-violence-against-women girls (femicide/feminicide). Besides the
statistical definition of such killings,
Administrative data for statistics on the framework identifies a typology of
femicide gender-related killings of women and
Closely related to violence against girls (femicide/feminicide) and the list of
women, UNODC and the Global Centre variables that can be used to identify and
of Excellence on Gender Statistics has count the various types of such killings.
developed a statistical framework on The framework also identifies main data
femicide using admin data: https:// that should be collected for providing
www.unodc.org/documents/data-and- information on victims, perpetrators and
analysis/statistics/Statistical_framework_ state response to gender-related killings
femicide_2022.pdf of women and girls (femicide/feminicide).
The general recommendation No. 35
of the Committee on the Elimination of
Discrimination against Women urges

12 Toolkit for Quality Assessment of Administrative Data for Official Statistics

Eastern and Southern Africa
10 The statistical value chain and
assuring administrative data quality

The statistical value chain or Generic country. This will be reflected in the steps
Statistical Business Process Model to take as quality assessment at each step
(GSBPM) aims to ensure that the same will depend on whether quality has been
terminology is used by all when referring assessed at earlier steps already.
to statistical business processes.
When working with administrative data
It is an ideal description of the processes as a source for official statistics, after
and might not in all cases reflect the local identifying the needs through user
realities 100%. Also, it does not need to involvement, it all starts with the mapping
be followed in a linear way. Work can be of potential data sources. What kind of
on-going on several processes at the same administrative data is available, what is it
time, all feeding into the improvement of being collected for and how? This is part
the overall work. of the initial assessment of the source’s
quality: An initial evaluation if it is good
Of course, countries have different
enough to be used for statistics, and if not,
systems and there are also differences
what needs to happen to improve it.
between MDAs owning data within a

Generic Statistical Business Process Model (GSBPM)

Toolkit for Quality Assessment of Administrative Data for Official Statistics 13

Eastern and Southern Africa
 After a source is found to be useable, and rooting these in the organisation’s
agreements for regular transfer to the business processes will help with
statistics producing entity need to be put assessing the quality of new data sources
in place. and in the long run will most likely result
in statistical outputs of better quality over
Also, metadata needs to be identified and
time – also those based on administrative
all treatment of the data documented with
data. For an adapted GSBPM model for
metadata, ideally in a unified system.
the work specifically with administrative
Both metadata and quality management data, see example 3.1 from the Latin
are cross cutting topics in the GSBPM. American region.
Having a structured approach to both

14 Toolkit for Quality Assessment of Administrative Data for Official Statistics

Eastern and Southern Africa
 11 Metadata

Metadata is a description of the data. • Administrative organization and the

There are structural and reference purpose of the collection
metadata. Structural metadata provides
• Concepts, definitions, classifications,
identifiers and descriptors of the data.
and protocols used
Without it, reading the data is difficult.
Reference metadata describes the content • Collection, processing, validation,
and quality of the statistical data. and quality assessment methods and
procedures
In order to understand and interpret
the data, comprehensive and clear • Reporting units and variables; including
metadata and documentation about data dictionaries, file structures, formats
the administrative source need to be and relationships within the data
available. Without this, it is not possible to
understand and assess the administrative The key challenge is to ensure that
source against the intended use. The metadata are captured as early as possible
metadata should include details about the: and stored and transferred from phase

Toolkit for Quality Assessment of Administrative Data for Official Statistics 15

Eastern and Southern Africa
 to phase alongside the data they refer to. process, or across several processes that
To ensure this, communication from early use common components (UNECE Quality
on in the process with the data holder is Indicators).
essential.
In order to store metadata about the
Metadata generated by the different sub- whole process, standardised templates
processes themselves are also of interest and a metadata management strategy and
as an input for quality management. These system(s) are recommended (SQAF)..
evaluations can apply within a specific

16 Toolkit for Quality Assessment of Administrative Data for Official Statistics

Eastern and Southern Africa
12 Quality dimensions
for administrative data

Depending on the different steps (NQAF bination of general international rec-

levels) of the process this guidance sug- ommendations and administrative data
gests applying different sets of quality specific countries’ best practice.
dimensions – tailored to the specific level
Next to the quality dimensions of the
in the process. One additional level, the
UN NQAF, there are three additional
assessment of the quality of the input/
dimensions added for assessing
source data, to the ones in the NQAF is
administrative data quality. These are
suggested.
completeness, uniqueness, and validity.
1. The statistical system and institutional More detailed descriptions of all the
environment quality dimensions can be found in part
2 in the respective sections.
2. The input/source data
3. The data processing and finally In the overview table below, all dimen-
sions are mapped showing how which
4. The statistical outputs
dimensions apply to the four steps of the
The quality of the actual data source is quality assessment process outlined in
especially challenging when it comes part 2.
to administrative data, as it is normally
Like the GSBPM, an assessment of the
outside of the control of the statistics
quality is ideally not a linear process
producer. The investigation of the input/
but happening in a circular way with
source data could be included in levels 1
constant feedback and improvement
and 3. It was chosen to make it a sepa-
between the initial collection point and
rate level in this guidance for the ease of
the other steps in the data value chain.
the users, as many of the countries con-
sulted indicated that this is currently one This guidance being a generic one,
of their main challenges. countries are encouraged to select from
it what is fitting in the national context.
Unlike in the UN NQAF, the dimensions
Also, especially in the beginning working
are not only applied to the output lev-
with a much smaller set of dimensions
el. The reasoning behind it is that the
or only parts of the questions for each
source/input data is the basis for the
chosen dimension can be a good idea.
statistical outputs and when it comes to
administrative data, often not in the con- A more detailed description of the
trol of the statistics producer. dimensions for each step will be
provided in the template in Part 2 of this
The quality dimensions applicable in
guidance.
each of the levels are based on a com-

Toolkit for Quality Assessment of Administrative Data for Official Statistics 17

Eastern and Southern Africa
18 Toolkit for Quality Assessment of Administrative Data for Official Statistics
Eastern and Southern Africa
References

BLUE ETS List of quality groups and indicators ONS Guidelines for measuring statistical
identified for administrative data sources: output quality: http://www.ons.gov.uk/ons/
http://www.pietdaas.nl/beta/pubs/pubs/ guide-method/method-quality/quality/
BLUE-ETS_WP4_Del1.pdf guidelines-for-measuring-statistical-quality/
guidelines-for-measuring-statistical-output-
BLUE ETS Report on methods preferred for
quality.pdf
the quality indicators of administrative data
sources, Piet Daas, Saskia Ossen (CBS) Rwanda quality assessment guidelines:
https://statistics.gov.rw/file/6668/
Daas et. al. 2009 Checklist for quality
download?token=4pqW8o-y
evaluation of administrative data
sources: https://ec.europa.eu/eurostat/ South African Statistical Quality Assessment
documents/64157/4374310/45-Checklist- Framework: https://www.statssa.gov.za/
quality-evaluation-administrative-data- standardisation/SASQAF_Edition_2.pdf
sources-2009.pdf/24ffb3dd-5509-4f7e-9683-
Statistics New Zealand and Statistics Norway:
4477be82ee60
Extending TSE to Administrative Data:
ESS Handbook on quality and metadata A Quality Framework and Case Studies
reports: https://ec.europa.eu/eurostat/ from Stats NZ: https://www.researchgate.
documents/3859598/13925930/KS-GQ-21- net/publication/318084895_Extending_
021-EN-N.pdf/143394de-e5a0-31ac-2c90- TSE_to_Administrative_Data_A_Quality_
2aa9c15803f0?t=1639042312202 Framework_and_Case_Studies_from_Stats_
NZ/fulltext/595c0ef20f7e9bf415b4aa91/
ISTAT Evaluating administrative data quality
Extending-TSE-to-Administrative-Data-A-
as input of the statistical production process:
Quality-Framework-and-Case-Studies-from-
https://www.istat.it/it/files/2014/10/Articolo-7_
Stats-NZ.pdf?origin=publication_detail
Evaluating-adfministrative....pdf
UNECE Quality Indicators for the Generic
ONS Draft guidance on quality of admin
Statistical Business Process Model (GSBPM)
data in statistics: https://best-practice-and-
- For Statistics derived from Surveys and
impact.github.io/admin-data-quality-stats/
Administrative Data Sources: https://statswiki.
departments.html#Introduction
unece.org/download/attachments/185794796/
ONS GSBPM: https://gss.civilservice.gov. Quality%20Indicators%20for%20the%20
uk/wp-content/uploads/2016/01/Generic- GSBPM%20-%20For%20Statistics%20
Statistical-Business-Process-Model.pdf derived%20from%20Surveys%20and%20
Administrative%20Data%20Sources_Final.
ONS quality toolkit: https://osr. pdf?api=v2
statisticsauthority.gov.uk/wp-content/
uploads/2019/02/qualityassurancetoolkit_ UNECE Using administrative and secondary
updated_Feb19_2.pdf sources for official statistics: https://unece.
org/fileadmin/DAM/stats/publications/Using_
Administrative_Sources_Final_for_web.pdf

Toolkit for Quality Assessment of Administrative Data for Official Statistics 19

Eastern and Southern Africa
 UN ESCAP Emerging Trends in Census UNODC, UN Women and CEGS: Statistical
Approaches in Asia and the Pacific: https:// framework for measuring the gender-related
www.unescap.org/events/2022/asia-pacific- killing of women and girls: https://www.unodc.
stats-cafe-series-emerging-trend-use- org/documents/data-and-analysis/statistics/
administrative-data-asia-pacific Statistical_framework_femicide_2022.pdf

UNESCO Quality assessment framework: UNSD Handbook on Registers-Based

http://uis.unesco.org/sites/default/files/ Population and Housing Censuses: https://
documents/training-workshop-manual-data- unstats.un.org/unsd/statcom/53rd-session/
quality-assessment-framework-2017-en_0.pdf documents/BG-3e-Handbook-E.pdf

UNICEF: The Administrative Data Maturity UNSD Handbook on Management and

Model (ADaMM): https://data.unicef.org/ Organization of National Statistical Systems:
resources/the-administrative-data-maturity- https://unstats.un.org/wiki/display/HSO/
model-adamm/ Full+Handbook+v2.3

UNICEF: Improving Data for Women UNSD Interoperability guide: https://

and Children Guidance on strengthening unstats.un.org/wiki/display/InteropGuide/
administrative data systems for gender Chapter+2%3A+Data+and+metadata+models
statistics: https://data.unicef.org/wp-content/
UN Women Advancing Administrative Sourc-
uploads/2022/05/Improving-data-for-women-
es of Data for Monitoring Gender-Specific
and-children_2022.pdf
Sustainable Development Goals in Africa:
UNICEF: Strengthening administrative data https://africa.unwomen.org/sites/default/
systems to close gender data gaps: https:// files/Field%20Office%20Africa/Attachments/
data.unicef.org/resources/strengthening- Publications/2019/Study20ReportAdvanc-
administrative-data-systems-to-close-gender- ing20Administrative20Sources20of20Da-
data-gaps/ ta20for20Monitoring20GenderSpecific20Sus-
taina.pdf

20 Toolkit for Quality Assessment of Administrative Data for Official Statistics

Eastern and Southern Africa
 Part 2
Template for assessing the
quality of administrative data

Toolkit for Quality Assessment of Administrative Data for Official Statistics 21

Eastern and Southern Africa
 This section explains the use of the with the NSO (especially the first time).
excel template as well as the division This is a great opportunity to strengthen
of the template into the different relationships and enhance dialogue and
steps. The process described in the mutual understanding. Also, according to
template is a general one. For different international recommendations, the NSO
countries, different parts might be of is supposed to be the coordinator of the
more importance and for different National Statistical System (NSS) and can
types of administrative data, different provide methodological/statistical expertise
considerations might be needed. Countries where needed.
are encouraged to actively engage with the
Once the relevant administrative data
excel template and tailor a version to the
sources have been identified, information on
national context and resources available.
what the data is originally collect for, how
Some guidance on consideration with the
frequently, at what level of disaggregation
different types of administrative data has
and by whom should be gathered. Also,
been included in part 1.
contact information of the institution that
The questionnaire is designed to be is collecting the data is to be recorded. If
filled in by the NSO together with the the data source is new and has never been
data holder (especially regarding the assessed before, focussing on the questions
input data quality part). In the case of the first steps is recommended, to find
of a decentralised statistical system, out if the source is potentially useable.
where the Ministries, Departments and Questions regarding the statistical outputs
Agencies (MDAs) are producing official (step 4) are often covered by a national
statistics, it is still recommended to quality assurance framework. If they are, this
complete the questionnaire together part of the toolbox is of course less relevant.

22 Toolkit for Quality Assessment of Administrative Data for Official Statistics

Eastern and Southern Africa
Step 1: The Statistical System,
Institutional Framework
and Agreements
a. Legal provisions A description of what is currently being done
when it comes to coordinating regarding the
There should be a short description of the
specific data source under review should be
legal framework/ the legal basis on which
provided in this section.
access to the respective data is ensured.
c. Professional independence,
It is important that there is a legal framework
impartiality, and objectivity
that ensures access to administrative data for
the statistics producer (NSO or the unit in the As stated in the UN NQAF, statistical
MDA that is responsible for production of a agencies should develop, produce, and
certain statistical product). disseminate statistics without any political
or other interference or pressure from other
If there is no legal framework that is ensuring government agencies or policy, regulatory or
access to administrative data for statistical administrative departments and bodies, the
production, section b on Memoranda of private sector or any other persons or entities.
Understanding and data sharing agreements is Such professional independence and freedom
of special importance. from inappropriate influence ensures the
For specific statistics (e.g., children, GBV) credibility of official statistics.
legal frameworks other than the statistical one This should apply to the national statistical
and special confidentiality concerns might be office as well as to other producers of official
important to describe as well. statistics. In the context of administrative data,
it is important too, as the data that is the basis
b. Coordination of the National
of the statistics is being collected by public
Statistical System
institutions. Guaranteed impartiality of the
As stated in the UN NQAF Coordination of the data is important for the users to be able to
work of the members of the NSS is essential trust in the statistics.
for improving and maintaining the quality of A description of the mechanisms and laws
official statistics. Especially when working that guarantee this independence should be
with administrative data, coordination and included in this section.
cooperation are key because in most cases
two and more institutions are involved in the d. Statistical standards and
data collection, transfer and then production procedures
of the statistics (e.g., the data holder(s) and
The UN NQAF states: “Standards refer to a
statistics producer(s)).
comprehensive set of statistical concepts,
The institutions involved need to agree on definitions, classifications and models,
many things ranging from a standardised methods and procedures used to achieve
set of concepts and definitions to who is the uniform treatment of statistical issues
producing which statistics and which numbers within or across processes and across time
are the ones being published. and space. The use of standards promotes
the consistency and efficiency of statistical
systems at all levels.”

Toolkit for Quality Assessment of Administrative Data for Official Statistics 23

Eastern and Southern Africa
 Especially when using administrative data, f. Confidentiality and security
the uniform treatment across institutions or
Confidentiality concerns the protection of data
at least an awareness of the differences is key
and in particular data at individual level, that allows
to the quality of statistical outputs. Ideally, the
identifying specific persons and organisations.
standards should not only be national but in
It is therefore important to know whether the
line with the international standards for that
administrative data source has systems in place
thematic area to ensure the final statistical
which keep the individual data confidential so
products are also comparable internationally.
information about persons or organisations is
e. Agreements with data holders not disclosed without authorisation or misused.
Mechanism to protect confidentiality could include
A Memorandum of Understanding (MoU) is anonymisation procedure and log system to track
always a good idea as a way of formalising who access data as well as staff awareness of the
the details of cooperation and exchange importance of confidentiality and signing of non-
between an NSO and an MDA. This section disclosure forms as part of employment contracts.
should briefly describe the kind of agreement,
duration and other relevant details that could Security relates to data storage and protection
affect data quality. e.g., from hackers or disasters that could either
delete data entirely or temporarily prevent access.
Next to a general commitment to cooperation Storing data on a centralised server increases
the MoU should contain technical details data protection by avoiding random storage of
about the data exchange. The technical details confidential information on e.g., staff’s hard discs
include: or flash drives. A centralized server gives the
• Type of data to transfer (e.g., possibility to define centrally managed institutional
information on data fields, metadata security measures.
structure, variables) Effective data transmission and storage security
• Format of delivery (e.g., API, CSV etc.) provisions seek to ensure that data is stored and
transmitted in a secure way, so data is not leaked
• Means of transmission (e.g., email, or lost. Further, putting in place proper back-up
secured transmission system etc.) and disaster recovery procedures will reduce the
risk of data being exposed to hackers or disasters.
• Frequency of data transfer (monthly,
quarterly, annually etc.) Protecting confidentiality and ensuring security
help increase trust and goodwill from the persons
If the data transfer concerns microdata
and organisations whose data we are dependent
(data at individual level), the MoU can
on. organisations whose data we are dependent on.
also include information on how data is to
be kept confidential and secure and only g. Staffing
going to be used for statistical purposes,
so the administrative data holder is A key quality principle is the independence of the
reassured when transferring the data to statistics production from political interference.
another authority. One way of checking the independence of
statistics is to ask questions on the independence
Details of what to include and templates of the unit at the administrative data holder and
for an MoU can be found here: https:// whether they can work independently from the
unstats.un.org/capacity-development/ often more political institution, which they are part
admin-data/docs/mou-guide-and- of. Furthermore, having sufficient staff to collect
template.pdf the data and produce statistics is important to
ensure the statistical products can be checked for
their completeness, validity etc. in due time.

24 Toolkit for Quality Assessment of Administrative Data for Official Statistics

Eastern and Southern Africa
Step 2: Input data quality (Source)

This step concentrates on the phase of data Systematic errors in all the areas mentioned
entry from the first data reporter to when data can often be identified with the help of
is received by the administrative data holder automated data analysis. Once possible errors
i.e., the data collection itself. The data might in data collection have been identified, it useful
go through several hands before it reaches the to set up measures to mitigate the identified
administrative data holder. risks to increase the quality of data entry in
the future, e.g., trainings, reporting manuals,
Data collection is a crucial step in ensuring
changes to the reporting form.
high quality of administrative data as the
quality of the input data is key for the overall b. Validity
quality of the statistical product.
Validity refers to the extent to which data
a. Accuracy conforms to the expected format, type,
and range e.g. dates are written in a certain
Data accuracy refers to the degree to which
format, an email address must have an ‘@’
the information correctly describes the
symbol. When data is valid it is easier to link
phenomena it seeks to measure. How well
it with other datasets and to run automated
does the data match reality. The methods used
processes.
for collecting data can influence the accuracy.
From a data collection point of view, the more
Errors in the data can already occur from the
clarity the data reporter has on the expected
point of first data entry, if the reporting form/
format, type and range, the more correct data
system, which the data holder has designed,
they enter into the system. Therefore, it is
is e.g., unclear, gives cause to confusion, too
useful to offer the data reporter guidance on
complex in its set-up, too lengthy or has too
which format the data should be reported in.
much overlapping data requests with other
The reporting forms can also be developed
reporting forms.
with data validation features that help the data
Data errors can also occur when the data reporter to report data in the expected format,
reporter enters data into the reporting form/ type, and range.
system e.g. poor internet facilities, lack of
capacity or appreciation of the importance of
c. Unique ID and disaggregation
data reporting, difficulty in capturing all data Uniqueness refers to whether the data can be
needed for reporting. tracked back to an individual population unit.
Uniqueness is important when wanting to link
Data errors can further occur when data is
data across different data sources and also in
transmitted and validated by the district level.
order to have disaggregated information on
These could for example be errors resulting
various vulnerable groups.
from accidentally deleting or changing data.
Finally, errors can occur once data is received To identify a unique population unit, the data
by the administrative data holder and stored system needs to include a unique identifier.
or processed in its repository. An example of a unique identifier is a person

Toolkit for Quality Assessment of Administrative Data for Official Statistics 25

Eastern and Southern Africa
 number issued in some countries, or a social of the event recorded in the register and the
security number. Using a nationally generated census reference date, or the length of time
unique identifier also provides opportunities between the most recent data updating and
to link data from different data sources when the data are then delivered to the NSO).
across government institutions and gives
Sometimes events are registered weeks or
opportunities for more advanced data analysis.
months later than they occurred for reasons
Disaggregation refers to the breakdown that can be practical, institutional, or legal
of observations to a more detailed level (e.g., the law allows families to report births
to that at which the observations were and deaths up to one year after the event
taken. In other words, the data can be split occurred, birth registration in rural areas
up (disaggregated) when finer details are might be hard to access). Data might also
required. This is made possible by the codes be collected/transferred with a different
given to primary observations. frequency than expected which might
influence the quality of the data.
Disaggregation is linked to uniqueness. If data
is available at individual level, the data can Typically, there is also a trade-off between
often be broken down by sex, age, geographic timeliness and accuracy. The more attention
location and more. This allows to compile that is given during data processing to
valuable statistics about vulnerable groups. producing accurate results the longer that
It allows for intersectional analysis and can processing usually takes.
help decision makers better understand
the multiple inequalities faced by specific e. Respondent burden
subpopulations, such as the most marginalized Reporting data is a burden on data reporters
women and children. as they spend time on (collecting and)
reporting data. Reducing the burden imposed
Also without a unique identifier there are ways
on data reporters in connection with the
to link data using various techniques.
production of statistics feeds into better
d. Timeliness quality. It is therefore crucial to investigate
whether systematic efforts to reduce the
Timeliness refers to how well the data reflects
reporting burden have been made. Efforts to
the period they are supposed to represent and
reduce the burden could include reducing the
how up to date the data and its values are.
volume of data collected to the minimum and
Data entry can influence the timeliness of streamlining different reporting forms, saving
data e.g., the time lag between the time of the data in a common system that allows pre-
event and when it is recorded in the system filling forms with already collected information
(e.g., in the context of a register-based census, so that the same type of data does not need
timeliness of the data can also be taken to to be supplied more than once.
refer to the length of time between the date

26 Toolkit for Quality Assessment of Administrative Data for Official Statistics

Eastern and Southern Africa
Step 3: Quality and data processing

This section refers to the processing of • Selectivity i.e., whether the data source
the data once it has been collected. Data only contains information on selected part
processing is an important step in assessing of the population, which could lead to bias
quality as during this phase errors to the data as whole sub-population groups might be
are detected and if possible corrected, and missing
metadata is analysed and produced.
• Redundancy i.e., whether the data source
Next to identifying problems with the source includes duplications of the exact same
data, data processing that is needed for statis- objects
tical analysis can also lead to errors in the final
outputs if not performed correctly. Detailed Conducting controls for completeness of
documentation (methodological metadata) of the data is an essential way to establish an
all steps of data processing is important to be understanding of the quality, especially for
able to back-trace the processes performed users since they need to know how accurate
and the sources of error that result from the the statistics is. Trying to identify and
integration of statistical micro-data. understand the reasons for incomplete data,
e.g., are missing data random or is there a
a. Accuracy and completeness system in certain type of missing data, will
provide an opportunity to correct its causes in
Completeness refers to the degree to which
the future. However, a 100 % complete dataset
a data source includes the data needed to
is unrealistic and might also require enormous
describe what is wanted by the user of the sta-
amounts of resources.
tistical product. Completeness is an important
dimension when we talk about administrative The data also needs to be checked for invalid
data sources because the data is not originally values e.g., outliners (data point which are
intended for statistical production. either way too high or way too low as could
be expected for the type of data) or wrongly
Data is complete when all the data required
entered format (e.g., comma instead of point,
for your purposes are present and available for
date etc.).
use. This does not mean the dataset needs all
100% of fields filled in to be complete. If not Once you know the reasons for incomplete
100% are available, this is ideally explained in and invalid data, it can increase the quality
metadata. However, the essential values and of the data, if where possible adjustments
units should be present. are put in place. Adjustments could
include imputation, correction of invalid
Completeness can be affected by:
data etc. These should be documented in
• Under-coverage i.e., whether objects are methodological metadata.
missing in the data source, which should
Also, it is important to remember that a
be present.
complete dataset may still be inaccurate if it
• Over-coverage i.e., whether the data has values that are not correct.
source contains objects, which should not
be present

Toolkit for Quality Assessment of Administrative Data for Official Statistics 27

Eastern and Southern Africa
 b. Validity the same definitions across different sources
is useful to ensure that the same type of
When assessing the quality of processing of
information is treated in a similar way. Also,
administrative data, validity refers to the extent
when international and national classifications
to which processing rules have been established
are used e.g., the International Classification of
to identify whether data conforms to the
Diseases or International Standard Classification
expected range and format. The same goes for
of Education, we can compare data across
having set up automated processes to identify
countries.
the invalid data. Having clearly defined and
implemented automated processes for checking Lastly, written down operating procedures or
and identifying data that does not conform with guidelines which define how the administrative
the expected range will have appositive effect on data is processed help to further clarity because
the data quality. the processing follows a structure and is
transparent, which also links to the previous
Next to identifying the invalid data it is important
point on validity.
for the data quality to have processes in place
for what to do with the invalid data once it has
d. Coherence and Consistency
been identified. E.g., should they be removed,
adjusted, corrected for invalid values like outliers Consistency refers to the ability to reliably
and wrongly entered formats. Errors can for combine statistics and data sets in different
example occur and impact the final statistical ways and for various uses. As briefly mentioned
product if missing values are just replaced with before, deviations from international and national
a 0. concepts, definitions and classifications will
influence both consistency and comparability.
c. Clarity If deviations exist, they need to be described
and justified to create an understanding of
Clarity during data processing refers to the
the differences and why these deviations are
availability of appropriate documentation
important to keep.
relating to the data, in particular information
about the concepts, definitions and Besides concepts, definitions and classifications,
classifications used. there might be other sources which could result
in a lack of comparability over time and regions/
Concepts, definitions, and classifications relat-
countries e.g., changes made in the geographical
ed to administrative data can differ from the
coverage of the regions or changes made to
ones needed for statistical purposes and, if
which type of administrative data is collected.
they differ, influence the level of consistency
and comparability. As a first step you need to
e. Technical checks
ensure that concepts, definitions, and classifica-
tions used for the data are available, and in line Technical checks are only relevant if the
with international and national standards. This administrative data is shared between
can be documented for the users in conceptual institutions. Ensuring the correct transmission of
metadata. data will reduce the risk of errors and time spent
on understanding the data. Transmission of
Developing a glossary that contains all concepts data should comply with the agreed format as
used and their definitions will provide clarity. described in the Memorandum of Understanding
The knowledge is important in understanding if such an agreement exists or documented in a
what the data describes e.g., if the term data sharing agreement. Furthermore, metadata
“school” is used without any details about (i.e., data that describes the administrative data
the level we might not know whether the such as concepts, definitions, and classifications)
school is at primary or secondary level, or a should be included in the transfer of data.
school providing vocational training. Using

28 Toolkit for Quality Assessment of Administrative Data for Official Statistics

Eastern and Southern Africa
Step 4: Ensuring quality
of statistical outputs

Quality related to statistical outputs can also provide the opportunity to inform them
be described in terms of how well outputs about upcoming changes to the statistical
meet user needs, or whether they are ‘fit output, e.g., in case of changes to international
for purpose’ – the purpose being defined standards and methodologies as well as
in dialogue with the users of the statistical potential challenges they should be aware of.
outputs.
Another way to learn about the needs of users
There are trade-offs between different Another way to learn about the needs of users
output quality components. The dimensions is to investigate how satisfied they are with the
of statistical output quality are not mutually statistical output. This could be done through
exclusive; there are relationships between the an anonymous user satisfaction survey e.g.,
dimensions and there are instances where published on the website or sent directly to
improvements in one dimension would lead to specific users but also through more direct
deterioration in another dimension. user needs assessments or user consultations.
The results of such studies should be analysed
For example, increasing the timeliness of
and discussed with relevant stakeholders, and
the output may mean a deterioration in the
appropriate measures should be implemented
accuracy as less data has been received and
to meet the identified user needs, of course
the producer has been able to make fewer
within the realm of what is possible considering
checks on the raw data. Trade-offs that must
resources available. Ideally, user needs that
be made in these circumstances should
cannot be satisfied should be communicated
be communicated to users, along with the
transparently.
reasons that those decisions were made. Also,
decisions on which dimensions to prioritise b. Accuracy and reliability
should be based on needs the users have
When it comes to user interests, accuracy and
expressed. For example, users may favour
reliability of the statistical product is normally
more timely statistics that come at the price of
very high on the priority list. Therefore, it is
accuracy, other times accuracy can be key and
very important to provide the users with clear
timeliness less important.
and accessible information on the accuracy
a. Relevance and reliability of the statistical product they
are consulting. Also, if there are reasons for
Relevance refers to the degree to which
concern or challenges with accuracy and
the statistical output meets user needs. To
reliability it is good to let the user know.
understand the needs of users, they need to
be consulted and asked. Regular contact with This can be done by publishing a quality
main users e.g., via a user group that meets report together with the statistical product.
twice a year on a regular basis, is a good The quality report summarises what has
way to learn about their needs and uses of been found when checking accuracy of the
the statistical output. For regular statistical administrative data source and then also during
production it is recommended that user needs the processing of the administrative data.
are re-evaluated regularly. Meetings with users

Toolkit for Quality Assessment of Administrative Data for Official Statistics 29

Eastern and Southern Africa
 Information on accuracy for the users can Finally, clarity includes aspects of whether
also include choices that have been made the data is available at different type of
for certain statistical product, where for disaggregation by e.g., region/district, gender
example timeliness has been prioritised over and age as disaggregation gives additional
accuracy. This can for example be the case value to the statistics output.
with some short-term economic statistics, the
covid pandemic and others. In these cases, d. Comparability and coherence
the users need the statistics as quickly as Comparability refers to the degree to
possible, which potentially comes at the cost which the statistical output can be com-
of accuracy. This is no problem for the overall pared, both over time and region or do-
quality of the statistical product as it is what main. To compare statistical output over
the user needs, if this is explained properly time, data needs to be available for an
in the documentation that accompanies the adequate period of time. It depends on the
statistical product. type of statistical output what an adequate
period of time is. For annual statistics, 5
c. Clarity years could be a good time period, while
Clarity of statistical outputs refers to the avail- for quarterly statistics 5 quarters might be
ability of appropriate documentation for the more relevant.
data so users can learn more about the meth-
Comparability is also relevant in geograph-
odological choices and limitations of the sta-
ical terms. E.g., district level administrations
tistical output. Such documentation will help
are always interested in statistics on their
the user understand the statistical output. It is
district and find it useful to compare with
also useful for interpretation and comparison.
neighbouring or similar districts. Com-
Therefore, published official statistics should
paring official statistics across countries
always be accompanied by metadata, which
is important for national benchmarking,
describe the content, sources, and methods.
however, to compare across countries the
Also, if such documentation includes a de-
statistical output needs to be produced
scription of the quality of the statistics, the
according to international standards.
user will have a much better understanding
on any caution they need to take when using
e. Timeliness and punctuality
the data. Changes to statistical methodologies
should be communicated to users. Timeliness refers to the length of time
between the date of publication, and
To enable users to determine whether outputs
the reference period (the time), which
meet their needs, it is recommended that
the statistical results are collected and
output producers report quality in terms of the
calculated. The longer the time, the less
quality dimensions of the UN NQAF, which are:
relevant the data is for the user. Therefore,
Relevance, Accuracy, Reliability, Timeliness,
a way to pay appropriate attention to the
Punctuality, Accessibility, Clarity, Coherence
issue of timeliness is to set up targets for
and Comparability. A good summary of quality
how long time the production of official
should contain quality measures and indicators
statistics must take. It is often so that a
for each of these quality dimensions.
long production time improve accuracy
Also, if errors have been detected in published but reduces timeliness, while a shorter
statistical output and when those errors production time reduces accuracy but
are corrected, it is a good practice to point improve timeliness. Different users will
out updates e.g., on the website where the have different preferences for the trade-off
publication is available. between timeliness and accuracy.

30 Toolkit for Quality Assessment of Administrative Data for Official Statistics

Eastern and Southern Africa
Punctuality refers to the time lag between Since users are different and have different
the actual date of publication and planned needs (there is a difference between the
date of publication of the statistical needs of a student, a policy maker, and a
output. To meet the quality dimension of researcher), statistical output can with benefits
punctuality, an up-to-date time schedule be disseminated through different means and
(i.e., publication calendar) on upcoming formats. First and foremost, the published
publications should be made publicly statistics should be available in a format
available to users e.g., on the website which is easy for users to access but also to
where the statistical output is published. download so they can reuse numbers.
It is internationally recommended that a
To understand the statistical output, it is a
publication calendar contains information
good idea to explain terms and concepts
on all publication for the coming year.
together with the numbers, and to illustrate
Changes to the expected release data
the data with figures and graphs to facilitate
should be announced well in advance so
an understanding of the statistics. Also,
users are informed about any delays and
users might find it necessary to explore older
can plan accordingly. At the end, the aim is
publications and thus previous releases should
of course to publish the statistical output in
be made easily available and accessible. If
accordance with the planned release date..
the responsibility of producing the official
f. Accessibility statistics shifts from one authority to another
(or from one department to another), links
Accessibility refers to the ease with which
should be made so users can easily find older
users can access the statistical output. It is
publications.
also about the format in which data is available
and the availability of supporting information.

Toolkit for Quality Assessment of Administrative Data for Official Statistics 31

Eastern and Southern Africa
 Other useful information

a. Engaging with users and data “fit for purpose”. At the same time, regular
holders engagement with users ideally increases mutual
trust and understanding. In the end, good
Before actually carrying out a quality
quality is to match the user needs as closely as
assessment of an administrative data source
possible with the resources that are available
information on the needs of the users should
(striking a balance between resources and
be available to find out what the final statistical
needs).
product is supposed to look like. The users
provide input to which statistical products Meetings between the data holders and
are needed which in turn provides the NSO ensure that everyone is informed about
specifications as to which administrative data each other’s work and changes in processes
sources could be used for the production of that could affect the statistical product are
them. communicated to all in a timely way (e.g., a
change in pension age, a change in the age
There can be two ways to go about the
limits of who is classified as a child and an
identification of administrative data sources.
adult, a change in what is reported as a violent
First, one can have a clear list of needed
crime and more).
sources for a specific statistical product and
then engage with the data holders in trying to These meetings can happen in many different
identify the needed sources. Another approach ways and at a frequency that is considered
is to engage in dialogue with a data holder suitable in the country context, e.g., every 6
to learn more about all their administrative months, every year or for some statistics every
data and once there is a complete list find 2 years.
out internally, what could be used for which
statistical products. b. Linking to the national quality
assurance framework
Regular engagement with the users is key to
quality assessments and assurance in order to The quality assessment of administrative data
know what the final statistical product needs should be connected to the overall NSS quality
to look like to match needs. Also, establishing assurance procedures. Thus, what is suggested
a good cooperation with the data holder is key in this guidance is ideally customised to match
to maintaining and improving the quality of the the national NQAF where needed.
data. In an ideal scenario, there is a national quality
Meetings with users are important to know steering group that meets regularly to discuss
which statistical products are needed and and agree on all matters related to statistical
if what is being produced at the moment quality.
meets the needs of those using the statistics The MDAs that own administrative data and the
(e.g., is the format, level of complexity, ones producing statistics should be included in
access according to user needs). Needs can this coordination mechanism, in some countries
change. Therefore, checking in with the users the users are included as well. If possible, each
regularly ensures that the statistical outputs MDA appoints a quality focal person that serves
remain relevant for the users and therewith as the contact for all quality related matters.

32 Toolkit for Quality Assessment of Administrative Data for Official Statistics

Eastern and Southern Africa
There are a number of examples from countries The mapping can be done sector by sector and
that have set up such kinds of structures (see in will thus give an overview of how different data
part 3 of the guidance). Having a short version sources within the same sector can possibly be
of the national quality assurance framework can linked. This overview is useful as the different
be a great idea for an NSO, as can it be to run public institutions within the same sector might
workshops on the NQAF and its benefits for all not themselves have this overview.
NSS members.
Once the mapping is done within a sector, it is
c. Mapping possible to identify those administrative data
already being used for statistics, and which
After identifying what the users’ needs are, and
administrative data sources could be further
having established a relationship with the data
investigated for their potential.
holder, the first step in assessing the quality of
administrative data is to identify which data (Templates for mapping of data sources can be
sources exist. found in part 3 of the guidance in the example
boxes related to source data)
Some administrative data sources are already
being used for producing official statistics, d. Prioritising
while other administrative data sources could
There are different models to follow when
be a potential source for producing statistics –
prioritising or deciding on which sources to
either replacing a part of, or an entire survey, or
assess in depth as well as the level of quality
producing new information on society which is
assessment for a certain statistical product.
not yet published.
When prioritising, it is very important to involve
To identify available sources a mapping tool is
the users. And once the choices are made, they
recommended. Such a tool provides a template
need to be documented in a quality report, so
for detailing information about the source
the users have access to them.
such as information about the data holder,
which variables the data source contains, These processes should ideally always follow
the population demarcation, process of data the same structure. This structure and reasons
collection, data validation procedures, which for going further with a source and prioritising
statistics it can produce etc. are also described in this section. There are
examples available in Part 3 of the guidance.

Toolkit for Quality Assessment of Administrative Data for Official Statistics 33

Eastern and Southern Africa
34 Toolkit for Quality Assessment of Administrative Data for Official Statistics
Eastern and Southern Africa
 Part 3
Country examples
and checklists

Toolkit for Quality Assessment of Administrative Data for Official Statistics 35

Eastern and Southern Africa
 1 GSBPM and administrative data

GSRBPM developed by the Latin The processes that are part of the GSRBPM
American countries model, like the GSBPM model, do not need to
In a joint initiative, the Latin American countries be executed in a strict order. Although it follows
have adjusted the GSBPM model to have a the logical sequence shown in the figure, it
process model for the work with administrative may vary depending on the circumstances
data for statistical production. The result and and some of these processes or sub-processes
amended steps can be seen in the figure below. could be executed several times, forming an
iterative loop.

36 Toolkit for Quality Assessment of Administrative Data for Official Statistics

Eastern and Southern Africa
2 Assessing and improving the
statistical system and institutional
environment
Example from Ghana on development on an NQAF by NSO and MDAs
With the passing of the new statistics bill in 2019, quality management of official statistics has been
made mandatory in Ghana. Prior to it there has not been systematic quality management in the
National Statistical System (NSS) of Ghana.

To meet this need, a quality steering group has been established with members from the NSO as
well as other producers. A 14-person team has been tasked to develop a code of practice for Ghana
that closely follows the international guidelines as well as establish quality principles. This team,
consisting of member of NSO as well as some of the main producers of official statistics in Ghana,
has sought inspiration internationally and produced a Ghanaian NQAF together.

The work has just been concluded. One of the main preliminary outcomes besides the documents is the
greatly increased communication and cooperation between the different stakeholders in the NSS.

Also, some first metadata documents have been produced and courses on the issues of quality and
metadata (in relation to SDGs) have been held.

Read more here: https://unstats.un.org/capacity-development/admin-data/

DetailedView/60d9ed2b63e8753688667ced

Example from Denmark on quality assurance peer review system within the NSS
To accomplish the coordination of activities in connection with the development, collection, production,
and dissemination of official statistics a set of guidelines for official statistics has been developed by the
NSO in cooperation with other national authorities producing official statistics (ONAs).

These guidelines reflect the European Code of Practice (CoP) with its 16 principles but are a shorter
and more reader-friendly version. Also, the NSO started visiting the producers (ONAs) to learn more
about how they were producing statistics and to introduce them to the Generic Statistical Business
Process Model (GSBPM). As a result of these visits, the national guidelines were adjusted to make it
easier to use for the ONAs.

As part of the quality management system, compliance with the guidelines is monitored yearly by the
ONAs filling in a self-assessment questionnaire based on the criteria in the guidelines. All producers
of official statistics will be subject to peer reviews. Each producer can expect to be reviewed every
5 years. All reviews start with the self-assessment questionnaire. This is followed by a visit of the
evaluation team, conversations between the team and producer under review. Each review is finalised
by a report with recommendations. The evaluation team consists of a team from Statistics Denmark
and representatives from 2 other producers. The team gets staffed by rotation principle and each
producer is part of the team for 2 years at a time.

As part of the system, the NSO offers workshops for the ONAs. These can be about the process, going
through the questionnaire as well as any other topics that are suggested by the ONAs.

Read more here: https://unstats.un.org/capacity-development/admin-data/DetailedView/

XIxIB1QTKrPOA9FqjyxO

Toolkit for Quality Assessment of Administrative Data for Official Statistics 37

Eastern and Southern Africa
 3 Assessing and improving
source data quality
ONS Draft Guidance on Quality of Admin Data in Statistics
This framework is designed by the ONS to help statistics producers assess the quality of
administrative data for use in the production of official statistics. It is divided into an introduction
as well as input and output quality assessment.

The framework defines input quality as the data coming in the door and suggests an assessment
of how suitable it is for what it is intended for. The second assessment step is the assessment of
output quality. The framework defines this as the quality of the analysis you have produced and
how well does it meet producer’s and users’ needs.

Therefore, the first step in any quality assessment is deciding what quality looks like for this
specific scenario. This decision factors in not only what high quality is to producer and user, but
also the time & information the producer has, as well as any costs associated with conducting
assessments or improving aspects of quality.

On demand there are questions and tools available from the ONS to guide statistics producers
through deciding whether this aspect of quality is important for their purpose. The purpose of the
ONS framework is not to answer these questions, but to point the producer towards what they
should be thinking about (and doing) in order to understand whether the data and/or the output
is fit for use.

https://best-practice-and-impact.github.io/admin-data-quality-stats/index.html

Example from Armenia on improving the quality of birth and death

registration data
Armenia has through close cooperation between many different ministries established an
interlinked CRVS system with unique identifiers and improved the quality and coverage of the
administrative data.

In 2015, an interagency working group was established by order of the Minister of Justice to
establish a CRVS system. Next to the Ministry of Justice it contained the Ministry of Health, Ministry
of Social Affairs, Ministry of Regional Government and Local Self-Governing Bodies, State Statistics
Committee, and Passport Department. Since 2017, the Ministry of Justice has been leading CRVS
mechanisms through the Office of the Prime Minister.

The successful implementation of the project can be traced back to a number of interlinked
initiatives lead by the different institutions that were involved. Some of the initiatives were training
of doctors in cause of death registration according to national/international guidelines, the
introduction of a Unique Identification Number, sensitisation of registration officials as well as
digitisation of the government IT systems.

More information about the Armenian CRVS system and its development can be found here:
https://crvssystems.ca/sites/default/files/assets/images/CRVS_Armenia_e_WEB%20(2).pdf

38 Toolkit for Quality Assessment of Administrative Data for Official Statistics

Eastern and Southern Africa
Checklist from Statistics Netherland’s Guidance on quality evaluation of
administrative data sources
Statistics Netherlands has in an article from 2009 divided the assessment process in 3 steps,
focusing on the assessment of source, metadata, and data. The table below indicates their
proposal for dimensions, indicators, and methods for assessing a source.

See the full article here: https://ec.europa.eu/eurostat/documents/64157/4374310/45-Checklist-

quality-evaluation-administrative-data-sources-2009.pdf/24ffb3dd-5509-4f7e-9683-
4477be82ee60

Toolkit for Quality Assessment of Administrative Data for Official Statistics 39

Eastern and Southern Africa
 Example from ISTAT of measurement method for accuracy and completeness
of input data quality
The tables below have been developed by the Italian Statistics Institute ISTAT. They provide
suggestions for measurement methods for accuracy and completeness indicators.

They have been published in an article on Evaluating administrative data quality as input of
the statistical production process from 2014. The article also contains suggestions for other
dimensions.

The full article can be accessed here: https://www.istat.it/it/files/2014/10/Articolo-7_Evaluating-

adfministrative....pdf

40 Toolkit for Quality Assessment of Administrative Data for Official Statistics

Eastern and Southern Africa
Example from Uganda on the improvement of the criminal justice data system

In 2020, with the support of UN Women, the Uganda Bureau of Statistics undertook an in- depth
analysis of the legal, policy and data environments for VAW administrative data in the justice,
law and order sector. Analysis included review of data collection instruments (registers) and
assessment of current information management systems and processes (including existing
information system infrastructure and human resources) of data capture and processing by the
data-producing institutions.

Analysis of the data environment identified the need to streamline and harmonize VAW data
collection in the justice, law and order sector. In response, modified administrative registers for
data collection with the recommended VAW minimum data set were drafted and reviewed by
focal points from each institution. The registers for data collection were then field tested (by
police officers, judicial clerks etc.). The field-tested registers were then revised again based on
service providers’ feedback. The revised registers were approved by the highest- level responsible
authorities in each institution before implementation.

Implementation of the revised VAW data collection registers by the judiciary and police began
in 2021. After adapting the VAW data collection tool for gender- based violence criminal and
civil cases, the judiciary conducted a census of VAW cases in four High Court Circuits and
corresponding Chief Magistrate Courts. In total, the data collection tool was applied to 2,890 VAW
criminal case files and 328 civil case files
Source: UN Women WHO guidance admin data on VAW 2022: https://www.unwomen.org/
en/digital-library/publications/2022/05/global-technical-guidance-for-collection-and-use-of-
administrative-data-on-violence-against-women

Toolkit for Quality Assessment of Administrative Data for Official Statistics 41

Eastern and Southern Africa
 Example from Namibia on improving CRVS data reporting through
collaboration of NSO and Ministry of Home Affairs on regional office trainings

In 2015, after a comprehensive assessment of the CRVS system in Namibia, a CRVS steering
committee consisting of the NSO, the Ministry of Home Affairs (MHAISS) and at the time the
Ministry of Safety and Security, as well as the Prime Minister’s Office developed a 5-year plan to
improve the system.

As a result, the system is now functioning more smoothly and the NSO can receive data from the
MHAISS to use for the production of annual vital statistics reports. These provide the basis for
improving the overall CRVS system in the country and information on potential errors in the data
and other observations. For example, the data shows clear differences between the regions in
terms of timeliness of registration and quality of data. This information is used by the Ministry of
Home Affairs to target training efforts - and inviting the NSO along to explain the importance of
correct data inputting.

As part of the five-year plan, the web-based system (NPRS) first developed in 2011 was also
improved which led to electronically capturing of births (2015) and deaths (2016). The web-based
registration system is government-owned, developed and maintained by the Office of the Prime
Minister. This means that it is very flexible, and changes can be made when necessary, timely and
without external costs.

Dialogues between stakeholders took a long time to be fruitful, especially in the beginning. More
time was spent on advocacy and awareness on CRVS to solicit support from key stakeholders by
sharing the common benefits from the system. After a while, the benefits became more visible,
positive results (e.g., improved registration processes, production of statistics from CR data etc.)
started to show and the level of support was increasing. Also, the high-level commitment was
increasing, as clear benefits for policymaking became visible.

An MoU has been signed between the NSO and the Ministry for Home Affairs to focus on areas
of collaboration towards improving CRVS in Namibia, which increases stability. The MoU has been
a great achievement to secure consistency of data transfer, even after management/government
changes. The Statistics Act allows the NSO to produce official statistics from different sources
in the country. The Births, Marriages and Deaths Registration Act limits the sharing of data with
private entities, but the limitation does not extend to ministries, regional councils, local authorities,
statutory institutions or bodies established by or under any law. This Act will be repealed by
the enactment of a new law (Civil Registration and Identification Act) which comprehensively
provides for privacy and data sharing.
Read more here: https://unstats.un.org/capacity-development/admin-data/
DetailedView/60e6fa8c286e603108105045

42 Toolkit for Quality Assessment of Administrative Data for Official Statistics

Eastern and Southern Africa
4 Assessing and improving data
quality during processing

Example from Cameroon on improving the administrative data quality

together with MDAs
The NSO has together with a national statistical training center developed courses in statistics
for the statistical units of the other ministries. These courses have been developed based on
the needs of each institution. After the courses, the NSO has offered support to each institution
for the production of their annual statistical reports. The aim of the initiative was to increase
availability and quality of the administrative data in Cameroon.

The training sessions were organized in waves of 22 staff for a module, i.e., 4 to 5 staff from each
administration. A training module lasted 2 weeks and the teachers were from ISSEA and a few
professionals from INS. Three modules were administered: a module for basic statistics, a module
for mid-level statistics and a more advanced module.

Regarding the strengthening of material capacities, several pieces of equipment were acquired
and distributed to the ministries to boost statistical production. These included computers,
tablets, printers, photocopiers, scanners and for some structures means of transport.

The project activities followed a three-step logic:

TRAIN — COMPLETE RESOURCES FOR PRODUCTION — PRODUCE.

After the end of the program (2014/2015), activities continued:

(i) administrations which had not yet completed their projects were able to complete them

(ii) statistical products to be published on a regular basis (directory statistics) continue to be so

despite serious funding difficulties

(iii) the design of the Statistical Information System has been strengthened and continues to be
carried out by the administrations with the technical support of the INS

(iv) many others statistical products are made available (outlook notes, sectoral data analysis
reports, dashboards, etc.)

(v) statistical culture has been improved, though still further improvements could be beneficial

Read more here: https://unstats.un.org/capacity-development/admin-data/

DetailedView/60d9ed2b63e8753688667ce6

Toolkit for Quality Assessment of Administrative Data for Official Statistics 43

Eastern and Southern Africa
 Example from Malawi: Recommendations for the improvement of
administrative data in health sector through data analysis
In a cooperation between the NSOs of Malawi and Norway, health administrative data was
analysed in order to improve the overall quality of the data.

The analysis found that missing data is a severe threat to the Malawian HMIS reported for
2013. The study showed difference in data registration at the different facilities. A facility not
registering data will, in most cases, have a negative impact on data quality. It is recommended
to identify where limited resources ought to be spent to achieve cost-efficient improvements of
quality. Improving completeness of Malawian HMIS data is an important first step. Further, it is
recommended to prioritize identifying health facilities that have errors influencing the end result
at aggregated level.

Improving quality by analysing already collected data to point out weaknesses in the data is
an efficient approach. By doing this, a highly recommended process of working in a “circular”
manner to improve quality will be established, which over time allows to improve the data further
and further.

https://www.ssb.no/en/helse/artikler-og-publikasjoner/the-health-management-information-
system-in-malawi?fane=om

Example from UK Office of National Statistics on why agreement on definitions

This example has been presented by the ONS Data Dissemination Unit. The unit is
responsible for harmonizing concepts and definitions in the ONS internally between
divisions and across government institutions.
As the example underneath illustrates, without someone responsible for harmonizing the
concepts and definitions, statistics on the same thing might be published as different
ones or the other way around. There might be misinterpretations and misunderstandings
that affect output quality.

Source: ONS Data Dissemination Unit Presentation 2022

44 Toolkit for Quality Assessment of Administrative Data for Official Statistics

Eastern and Southern Africa
Example from UNSD on an SDG metadata reporting template

Countries with metadata templates for the MDAs? (Ghana?)

Example classifications library/transformation list (Chile)

Toolkit for Quality Assessment of Administrative Data for Official Statistics 45

Eastern and Southern Africa
 5 Assessing and improving
the output quality

Example from Denmark on quality reports on administrative data that are

accessible to all users
At Statistics Denmark, every statistical product is published together with a quality report on the
website. The most basic form of the quality report gives a short description of the product and
how it has been compiled and also potential uncertainties/ problems with the data.

Different users might be looking for different levels of detail. So, for those looking for more
information it is possible to extend the description and get several levels more of information on
the statistical product.

View an example at: https://www.dst.dk/en/Statistik/dokumentation/documentationofstatistics/

urban-areas

46 Toolkit for Quality Assessment of Administrative Data for Official Statistics

Eastern and Southern Africa
Example from Denmark on half-yearly user committee meetings
Statistics Denmark is consulting users in a systematic and regular way as a means to
increase user engagement and to include the users more in the NSOs work – this way
also ensuring that the statistical products remain relevant.

Currently there are 8 user committees. They are organized around the statistical topics
that have found to be of highest interest to the users. At the moment these are: labour
market statistics, demographic statistics, research, food statistics, municipalities and
regions, welfare statistics, knowledge society and lastly economic statistics.

Each committee consist of between 20 and 30 members from different parts of society
relevant to the topic of the committee. This can be different ministries, academia, the big
business associations, association of municipalities and others. They meet on a regular
basis (e.g., once or twice a year) and discuss everything related to the topic of the
committee.

Next to the user committees, Statistics Denmark has a committee for the businesses
that by law are obliged to share certain information with the NSO. The purpose of this
committee is to consult the business representatives (e.g., the big business associations)
regarding the respondent burden and their experience reporting, to make the process as
convenient as possible. At the same time, they are users of the statistics and can in those
meetings voice their needs.

Toolkit for Quality Assessment of Administrative Data for Official Statistics 47

Eastern and Southern Africa
 6 Prioritising

Example on how to prioritise quality assessment according to risk level from

the ONS (UK)
The Office of National Statistics of the UK (ONS) has developed a QA model that is divided in 4 main
areas. For each area, different levels of QA are outlined, going from none to a high level of QA actions.
To decide on the appropriate level of QA for the respective statistics, the ONS has also developed a
risk matrix.

Level of risk of data quality concerns

Low risk – the data may have a low risk of data quality concerns in situations in which there is
a clear agreement about what data will be provided, when, how, and by whom; when there is a
good appreciation of the context in which the data are collected, and the producer accepts that
the quality standards being applied meet the statistical needs..

Medium risk – the data may be regarded as having a medium risk of data quality concerns when
high risk factors have been moderated through the use of safeguards, for example, integrated
financial audit and operational checks, and effective communication arrangements. It is also
appropriate to consider the extent of the contribution of the administrative data to the official
statistics, for example, in cases where the statistics are produced in combination with other data
types, such as survey or census data.

High risk – the data may have a high risk of data quality issues when there are many
different data collection bodies, intermediary data supplier bodies, and complex data collection
processes with limited independent verification or oversight.

Source: https://osr.statisticsauthority.gov.uk/wp-content/uploads/2019/02/
qualityassurancetoolkit_updated_Feb19_2.pdf

48 Toolkit for Quality Assessment of Administrative Data for Official Statistics

Eastern and Southern Africa
Toolkit for Quality Assessment of Administrative Data for Official Statistics 49
Eastern and Southern Africa