Hlavní práce75132_wilg01.pdf, 573.3 kB Stáhnout

PRAGUE UNIVERSITY OF ECONOMICS AND BUSINESS PRAGUE, CZECH REPUBLIC

MASTER´S THESIS

2020 G. Alex M. Williams

II

Prague University of Economics and Business Faculty of Informatics and Statistics

Study Program: Quantitative Methods in Economics Field of study: Official Statistics

NEW APPROACH TO LIBERIA POPULATION AND HOUSING CENSUS:

FROM TRADITIONAL BASED TO COMBINED APPROACH

Master’s Thesis

Author: G. Alex M. Williams, BSc Supervisor: Ing. Petr Mazouch, Ph.D.

Prague, December 2020

III

I hereby declare that I am the sole author of the thesis entitled “New approach to Liberia population and housing census: from traditional based to combined approach”.

I duly identified all citations. The used literature and sources are stated in the attached list of references.

Prague December 7, 2020 Signature

G. Alex M. Williams

IV

I at this moment wish to express my profound thanks and appreciation to the supervisor of my thesis, Ing. Petr Mazouch PhD for his incredible and countless support and guidance from my enrollment at VSE to present. Special recognition also to the supervisor of my project during the internship, Mgr. Ing. Martin Zeleny, PhD.

Lastly, I appreciate all those who assisted and motivated me one way or the other during this challenging but amazing academic journey.

V

Liberia has relied on the traditional-based approach in its population censuses. However, there have been different variables and sensitive issues that the nation would be addressing in its 2021 census. Some of the major problems include the migrant status of the population and employment characteristics. It is contended that the digitized field-enumeration census through the CAPI approach in 2021 would provide standardized variables and data that could be used for future censuses beyond 2021. Considering the financial constraints and resource limitations, the nation might opt to adopt a combined census approach or a register-based census approach for its future censuses. The thesis aimed to determine the feasibility and roadmap for future population censuses in Liberia after 2021. The traditional method, despite digitized instruments of data collection, faces numerous challenges. At the same time, the modern census approaches offer a reliable model that maximizes the accuracy and quality of socioeconomic and demographic data. The thesis used a systematic review approach regarding the implementation status of the combined and register-based census approach in different nations, along with their pitfalls and utilitarian values. The thesis depicts that a combined approach would be appropriate for Liberia because it will combine data from surveys and national records that would help to address the issues related to migration, delocalization, and under-coverage. The nation should continue to verify the data collected from registers and sample surveys in future surveys with spatial data to supplement the population and housing- grid models. Such initiatives would increase the accuracy and reliability of the census data that would be created through combined or register-based census approaches.

Keywords

Population census, Traditional-based approach, Combined-approach, Systematic review, Liberia, beyond 2021

VI

LIST OF TABLES………. …..………. VIII LIST OF FIGURES……….………. IX LIST OF ABBREVIATIONS……….……….X

1. INTRODUCTION ... 1

2. MOTIVATION AND AIM ... 2

3. BACKGROUND ... 4

4. REVIEW OF APPROACHES TO POPULATION& HOUSING CENSUS ... 7

4.1. Domain of Different Census Approaches ... 7

4.2. Limitations on Reliability and Methodology ... 10

4.3. Prerequisites for Transitioning to a Register-Based Census or Combined- Approach ... 11

4.4. Difficulties Associated with Transitioning to Combined or Register-Based Census 13 4.5. Preconditions that should be considered for Using Administrative Registers 18 4.6. Assessing the Quality of Census Data Including Data Cleaning ... 19

5. BEST PRACTICES ... 22

5.1. Case Study: Ireland ... 23

5.2. Case Study: Estonia ... 24

5.3. Case Study Austria... 24

5.4. Case Study: Slovenia ... 25

5.5. Case Study: Germany ... 26

5.6. Case Study: Italy ... 28

5.7. Case Study: Portugal ... 29

5.8. Case Study: Poland ... 29

5.9. Case Study: England and Wales ... 30

5.10. Case Study: The Netherlands ... 31

5.11. Case Study: Arabian Countries ... 32

6. DISCUSSION ON THE FEASIBILTY AND ROADMAP FOR COMBINED OR REGISTER-BASED CENSUS BEYOND 2021 IN LIBERIA ... 33

6.1. National Identification Register: ... 33

6.2. Registers that should be created for Future Censuses Beyond 2021 ... 34

VII

6.3. Variables that Should Be Added and Updated for Future Censuses Beyond

2021 35

6.4. Generating the Unique Identifier: Addressing Over coverage and Under

coverage in future censuses ... 35

6.5. Special Provisions for sensitive variables: Ethnicity and Religion in Population Census ... 36

6.6.

Data Cleaning Related to Foreign and Internal Mobility Data in Census .. 36

6.7. Reference for Sample Surveys for future censuses and Pilot Study: Provision of satellite images and Postal inquiry ... 37

7. CONCLUSION ... 39

8. SUMMARY ... 40

9. REFERENCES ... 42

VIII

Table 1: Registers currently in Liberia……….34 Table 2: Topics for Liberia Census Beyond 2021………35 Table 3: Data Sources and Variables……….38

IX

Figure A: Census Approach in Germany………27 Figure B: Census approach in Poland……….30

X

BSPIN: Branch-Specific Personal Identification Number for Official Statistics CAPI: Computer-assisted personal interviewing

CASI: Computer-assisted self-interviewing CATI: Computer-assisted telephone interviewing CPR: Central Population Register

CR: Civil Register

CSSIS: Census and Social Survey Integrated System DSE: Dual System Estimation

EA: Enumeration Area GEOSTAT: Geodetic Satellite

GIS: Geographic Information Systems

LISGIS: Liberia Institute of Statistics and Geo-information Survey LRA: Liberia Revenue Authority

NBIS: National Biometric Information System NDI: National Data Infrastructure

NIN: National Identification Number

NSDS: National Strategy for the Development of Statistics NSO: National Statistical Office

PAR: Person Activity Register

PASI: Paper questionnaire with self-enumeration PIN: Personal Identification Number

SOL: Signs of Life

TUSOL: Testing under coverage using signs of life

1

The primary objective of a population census of any nation is to obtain and update its demographic and socioeconomic data for improving developmental planning and economy.

The demographic data that is of relevance include the age structure of the population, living conditions and economic profile, as well as the spatial distribution of the population, and migration patterns.

These parameters are often impacted by socioeconomic levels that include employment status, occupation, education level, and health security. In developing nations, a population census is the primary source of demographic data because the major sources of demographic information such as vital registration, routine socioeconomic sample surveys, availability and accessibility to administrative registers, the level of community or civil registration, and immigration-emigration data are very poor in Anglophone Africa and Liberia is no exception to such a phenomenon.

As a result, these nations cannot adopt fully register-based approaches for population censuses, unlike developed nations or nations that are not affected by genocide and civil war.

Instead, field-based enumeration or traditional approach through paper-pen or CAPI methods have remained the most popular and appropriate choice for undertaking population censuses in developing nations such as Liberia. These assumptions are substantiated by the fact that the previous population censuses that were conducted before 2008 failed to preserve any demographic statistics or socioeconomic status for the referred period.

Hence, it was challenging to conduct the population census in the nation during 2008 due to the exhaustive data and efforts required to enumerate the population, based on demographic and socioeconomic backgrounds. It is predicated on these and many other drawbacks this paper seeks to provide tested and proven solutions to improve the overall population statistics in Liberia.

2

The author of this Thesis has passed the internship at the Czech Statistical Office as an intern in the Demographic and Social Statistics Section under the supervision of Mgr. Ing. Martin Zeleny, PhD. This Thesis is an expansion of the author Internship Project while working as an intern at the Czech Statistics Office. While at the Czech Statistical Office, the completed Internship paper on the comparative analysis of approaches to population and housing census in Europe and Africa in general with emphasis on Liberia, Croatia, Czech Republic, and Malawi provided ample motivation for further expansion. It was exciting to find the different approaches practiced for population census in the respective nations based on the availability of specific variables and administrative documents. More interestingly, it was fascinating to understand the ways different countries transitioned to combined-approach and fully register- based census from the traditional approaches. Hence, the author endeavored to identify the administrative documents, variables, and statistical measures that could help Liberia to transition to a combined-census approach for censuses in the nation beyond 2021. He is motivated to explore the feasibility of the combined census approach because of the financial and resource limitations of Liberia for continuing with the traditional field-based census approaches, especially an under-developed country like Liberia.

Although the population census of 2008 in Liberia was robust, there were different bottlenecks and biases over the census data. One of the major drawbacks of the 2008 census was the lack of human and infrastructural resources. Since transportation was difficult because of harsh weather conditions, it was difficult to collect and synchronize census data in a timely manner.

2021 would be a paradigm shift from previous censuses conducted through traditional paper and pencil enumeration (Liberia Institute of Statistics and Geo-Information Services ..., no date). The use of the CAPI approach would eliminate the temporal and spatial barriers related to the collation and synchronization of data with other parallel agencies involved in the Census. The digital format would also help to access various databases for aiding decision- making to the enumerators in the absence of operational administration at the local levels. It is contended that the young democracy of the nation does require robust and authenticated socioeconomic data. As a result, the digital method of enumeration, would not only help to ease out the enumeration process but would also help to create a database that would help Liberia to transition to a combined approach (field enumeration and administrative registers) and to a complete register-based approach from the next Census which would be both cost- effective and fast. Different points of reference for the 2021 census in Liberia are the joint effort of LISGIS and NSDS. In the 2008 census, in most places, there was no operational administration at the local level. As a result, the enumerators had to apply personal judgment for tracking units and individuals within a given location. The EA maps provided a good backup for identifying the census population, the probability of bias on the part of the enumerators cannot be ruled out. Such issues might have led to overrepresentation and underrepresentation of certain strata of the population in the census data.

One of the major limitations for the 2008 census in the nation was that there was no point of reference for addressing the overrepresentation and underrepresentation of data due to the unavailability of administrative registers. The population census of 2008 in the nation helped to develop various policies related to poverty reduction. However, one of the controversial issues related to erroneous census data and policy changes include reapportioning of citizens representation in the parliament. In most instances, personal biases and difficulty in accessing the census units led to the inaccurate reflection of the population in the constituencies.

Likewise, there are radical differences in population densities across various electoral districts.

Such bottlenecks could have been addressed by redrawing the census districts from the EA maps. The 2008 census was based on various surveys that included the demographic and health survey, the core welfare indicator questionnaire, and the population poverty assessment data. Moreover, certain individual organizations did ad-hoc surveys on their own,

3

and such data were also considered for the Census. Surveys and ad-hoc data cannot be perfect because they could only help to provide estimates but not the actual data on demographics and socioeconomic parameters. Sampling strategies for such a census is not appropriate, rather erroneous. The present thesis would evaluate the feasibility and terms of reference for undertaking a combined census approach for the population census of Liberia in 2031.

This thesis would be based on the background that the population census of the 2021 census would offer a starting point for using the administrative sources (registers) and sample surveys for new variables for censuses beyond 2021. The paper would explore the historical data and the terms of reference of other countries such as Czech Republic while transitioning from the field-based approach to the combined-approach. The thesis would further explore the mechanisms and identify the relevant data that would help in transitioning from the combined approach to a fully register-based census for the nation in subsequent censuses beyond2021.

This thesis would help to identify the data that should be mandatorily obtained through CAPI enumeration for the 2021 census for its subsequent use. The referred data would help to develop the administrative registers and more importantly update the national population register. The present thesis would incorporate various theoretical and conceptual frameworks for identifying the appropriate administrative registers so that issues of overrepresentation and underrepresentation are significantly minimized. This is because it was one of the major bottlenecks that reduced the reliability of the 2008 population census. The position paper would also provide an appropriate roadmap for addressing the inconsistent representation of the constituencies as per the electoral districts in the census data as well as those related to emigration and immigration and lack of local operational support. In other words, this position paper would help to identify the terms of reference as well as the best approaches for conducting the census beyond 2021. The quality of this document might be regarded as a roadmap for a paradigm shift in the census approach of nations in the future as it would pave the roadmap for adopting a fully register-based approach.

The study aims to define the future of population statistics in Liberia. The census approach adopted should capture the various government needs and define the specific type of data to be obtained from enumeration and surveys and the ones to be sourced from administrative records. The country has already planned for its 2021 population census, and, therefore, this study will inform changes for the future after 2021. Enumeration offers crucial data about population, housing, income, and education, but lacks reliability owing to Liberia's porous borders. Administrative data offer limited data, which could be used to achieve data saturation of the Census and improve its reliability. The country faces the risk of encountering content and coverage errors if it relies on the traditional method to be used in 2021 as was done in the past. However, if it borrows a leaf from European countries that have implemented innovative census strategies in its future censuses beyond 2021, Liberia will benefit from the modern Census's effectiveness and efficiency.

4

Liberia has been sub-divided into various administrative areas to ensure complete coverage and preparation of a comprehensive census manual for supervisors and enumerators. The Census would include demographic and household attributes such as age, gender, literacy, migration, births, deaths, education, disability, and household characteristics. The census report should be strengthened with economic parameters and more robust health data related to vaccination and the identification of at-risk populations concerning life-threatening diseases. Nevertheless, the 2021 census would incorporate various quality assurance initiatives from cartography and GIS to inter-census household survey program, to ensure the reliability and reproducibility of the Census both spatially and temporally. GIS helps to document the population density as a function of dwelling units across various topographic locations which helps to verify the enumerated data on population census.

Since 2004, Liberia has witnessed significant fluctuations in its population due to health crisis, socioeconomic crisis, and changes in political and cultural factors, which have offset the post- war developmental gains. However, with an increased sense of security and resurgence of economic activity, people have started to move back since 2008. Notwithstanding, there were significant bottlenecks in obtaining quality socioeconomic and demographic data due to a mix of low socioeconomic and human development(Liberia Institute of Statistics and Geo- Information Services ..., no date). The data gaps were further exacerbated by the destruction of databases and statistical capacities and resource personnel loss. The characteristics of dwelling units that would be investigated have incorporated the load-bearing strength of the residential units that would add to the housing grid statistics. A gap exists in exploring the internet connection and ownership of computers, major parameters driving sustainable development (Liberia Institute of Statistics and Geo-Information Services ..., no date).

Before the 2021 census is undertaken, one of the major plans is to update the digitized enumeration area maps produced by the 2008 census through field cartography. The CAPI would be transferred from the legacy system to a central LISGIS server to troubleshoot and finalize census data. The enumeration areas would be visited by the specific teams for capturing the data. Moreover, a pilot census was to be undertaken in March 2020 to test the new CAPI tool but was never done. For ensuring census data quality, post enumeration surveys should be undertaken on a selected EAs. The inter-rater variability between the Census conducted by the third party and LISGIS on the same EAs would be used to streamline and rectify the CAPI data. However, LISGIS should take every measure to ensure that the quality of the census data is not compromised relative to poor maps, untested methodologies, inadequate supervision, lack of adequate training, and inadequate support from the public, and impaired data processing (Liberia Institute of Statistics and Geo-Information Services ..., no date).

Liberia's next round of National Population and Housing census is all set to be initiated in March 2021. The previous population censuses in 1962, 1974, 1984, and 2008 have witnessed population estimates of 1.1 million, 1.5 million, 2.1 million, and 3.5 million, respectively (Liberia Institute of Statistics and Geo-Information Services ..., no date). The delay in the present round of population census violated Article 39 of the Liberian constitution which mandates the legislature to cause a census of the Republic to be undertaken every ten years.

The 2021 census is unique because it would be the first-ever Census to use digital technology in all phases related to preparation, geographic planning, collection, management, processing and tabulation, publication, archiving, and dissemination of census data. Although the Census as per the mandate of Article 39 was scheduled to be held in 2018, it got delayed due to untimely and inadequate funding that led to the violation of the United Nations Principles and Recommendations on the frequency of conducting the Census (Division, 2017). The United Nations Fund for Population has developed a comprehensive census project document that

5

would serve as the point of reference and resource mobilization for all the deliverable attributed to the 2021 census. The Government of Liberia has received substantial funding from the Swedish Embassy which is estimated to be 70 million Swedish Krona for undertaking the population census of 2021 (Liberia Institute of Statistics and Geo-Information Services ..., no date). The financial support and mobilization of resources indicate the importance of the 2021 Census of Liberia from the perspective of both the Government of Liberia and the United Nations that is committed for preventing another civil war in the nation in the distant future.

One of the major barriers for population censuses in 1984 as well as in 2008 in Liberia was the lack of relevance of certain demographics post-civil war era because a major population either emigrated or dislocated because of the war and this greatly affected the unit size and population density. Such assumptions were substantiated by the EA maps which confirmed the relocation or dislocation status of the population. The 2008 census in Liberia addressed various issues that limited data quality and data authenticity that were witnessed in the previous four censuses. The Census was preceded by a mapping exercise that canvassed the entire nation. This initiative helped to frame the boundaries related to administrative hierarchy and geographic positioning of the localities.

The 2008 population census of Liberia bridged the statistical gaps that were evident in the previous censuses because it included baseline statistics at the national and sub-national level.

These baseline statistics acted as the points of reference along with updated demographic indicators that were available from the limited administrative registers. The major change in the census method for Liberia in 2008 compared to the previous methods was to shift from a de jure approach to de facto approach. As a result, there was no question regarding the reliability, validity, and reproducibility of the enumerations. The 2008 population census also included the agriculture module as the point of reference for designing and implementing agricultural surveys in the future. The de-jure method of enumeration is complex because it considers the records of the usual residents of the household. On the contrary, the de facto approach is easier to conduct because its records persons spending a reference night in the household. The inclusion of the agricultural module was justified because the nation's economy is largely dependent on agriculture. The provisions were made so that special and marginalized populations such as street children who did not have a formal household as well as for inmates and transients in hotels, airports, and seaports (Liberia Institute of Statistics and Geo-Information Services ..., no date).

Therefore, the field enumeration in the 2021 census in Liberia would be more structured than it was during the previous four censuses. Available records show that there would be 10,500 enumerators and that one supervisor would manage five of them. Moreover, there would be 15 counties coordinators for authenticating and preparing the LISGIS data and six regional supervisors. Hence, the quality of field enumeration for the 2021 census would be comprehensive and reproducible. The principal objective of the 2021 Census in Liberia would be to enumerate and characterize the population in terms of its spatial distribution and location in Counties, Districts, Clans, and as well as their urban and rural distribution. The data would be sorted in terms of age, gender, race, income, ethnicity, migratory status, disability, geographic location, and other national interest attributes. The Liberia Institute of Statistics and Geo-Information Services, with support from the United Nations Population Fund, have created the terms of the census data. LISGIS is resource-limited; therefore, UNFPA will provide technical assistance and human resources for the 2021 census. The upcoming Census would be based on Computer Assisted Personal Interviewing. This is a paradigm shift from previous censuses conducted through traditional paper and pencil enumeration (Liberia Institute of Statistics and Geo-Information Services ..., no date). Liberia would be moving from a purely PAPI enumeration approach to CAPI in its 2021 census. However, it is contended that the nation should completely rely on field-based enumeration through CAPI,

6

and no administrative registers should be considered. Although resource-intensive, the CAPI approach would ensure the accurate documentation and coding of census data. Such an approach is desirable when the nation is eyeing on sustainable development, and there has been a significant movement across the population during the last ten years. However, the 2008 population census has generated certain administrative registers which could be either used or used as terms of reference in combination with the CAPI approach. This thesis would explore the rationality and feasibility of transitioning to a combined census approach and the statistical measures that would help to align with the same.

7

4. REVIEW OF APPROACHES TO POPULATION& HOUSING CENSUS 4.1. Domain of Different Census Approaches

The traditional definition of a census pivots around the principles of enumeration, universality, and periodicity. Population and housing census collect, process, and disseminate detailed statistics on population, its composition, characteristics, spatial distribution, and organization across families and households for a small and defined area (Division, 2017).

During the past few years, various population census methods with a view of increasing quality and reducing time and cost have been adopted. For example, some countries have adopted administrative registers in preference to real-time field enumeration for conducting population census. Population census is referred to as the operation that provides the official counting or the benchmarking of the population in the territory of a country and its smallest geographical sub-territories in a regular interval with information on a selected number of demographic and social characteristics of the total population (UN/ECE, 2014). Likewise, a housing census is defined as the official enumeration of all housing stock in the territory, a nation, and the smallest geographical sub-territories and information on a selected housing number (UNECE, 2015). Individual enumeration refers to the information on every enumerated person and to record their characteristics separately (UN/ECE, 2014).

Simultaneity refers to the information obtained on individuals or housing in the context of a defined reference period. Universality refers to the Census on the total number of individuals and households in a defined territory. Small area data refers to the data on population and housing from the smallest geographical areas of a nation such as small population subgroups.

This form of data collection is consistent with the overriding requirement of protecting individual confidentiality. Finally, defined periodicity refers to the undertaking of a census at regular intervals. Census data must be or are obtained every ten years. Few countries are an exception in terms of ten years. However, the census topic pivots around various strategic objectives that include the demand and availability of data at the national and local levels, the constraints of collecting data, the user's need, and the authenticity of the data sources (UNECE, 2015).

Traditionally, the National Population and Housing Census involve the enumeration of individuals and housing. However, they extend to include the individuals' social and economic background. The primary demands of the modern census process are obtaining information on different characteristics that may result in exceedingly long census forms, which could reduce the quality of the data due to the respondents' burden. The short-form consists of selected questions, while the long-form consists of more questions on specific topics. The long form is administered correctly through sample surveys rather than selecting the entire population (UNECE, 2015).

Three methodological approaches are used in the UNECE region for conducting a census. The UNECE region not only includes countries in Europe but also countries in North America, Central Asia, and Western Asia. The first method is the traditional method of full enumeration, which could be supported by the registers as the frame of control. In this context, field operation refers to the collection of data using field enumerators and supervisors who conduct personal interviews with the residents by Pen-Pencil and Paper Interviewing (PAPI).

Moreover, there are also alternative data collection methods under this traditional method, including other online or telephone methods. The second method of the Census involves the combined approach of full enumeration supported by data obtained from administrative registers. Finally, the third approach involves using registers and other administrative sources

8

exclusively (UNECE, 2015). Related studies reveal that the Netherlands has been practicing the third approach for a relatively long period. According to Statistics Netherlands, nearly all the necessary data for its 2011 census was collected from the administrative registers. The Census was conducted with only 15 staff and a budget of just 1.4 million Euros. The registers and other administrative sources have become a viable alternative to the traditional census approaches, especially when the veracity and volume of census data are very large. A register is defined as the systematic collection of unit-level data that is organized in such a way to enable updating census data. Sample surveys conducted to enrich data in the long-form are unable to provide comparable data like administrative registers. However, they are used in combination with field enumeration for supplementing census information on certain topics (UNECE, 2018).

From research, we see a significant evolution in the methods of censuses in the UNECE region for the past two decades. During the 2000 round of Census, 40 nations opted for a traditional census approach while five nations and three nations opted for a combined census approach (registers and collection) and the only register-based approach respectively, while seven nations did not undergo a census. On the contrary, during the 2010 round of Census, 34 nations opted for a traditional census approach followed by ten and nine nations that opted for a combined census approach (registers and collection) and the only register-based approach, while three nations did not undergo a census. Finally, in the 2020 round of Census, 22 nations opted for a traditional census approach followed by 12 and 14 nations that opted for a combined census approach (registers and collection) and the only register-based approach, while eight nations do not plan to undergo a census. These findings suggest that the European nations are shifting from the traditional enumeration approaches to register-based or combined approaches (UNECE, 2018).

The significant advantage of a traditional approach is to provide a snapshot of the whole population for a specific period along with the potential availability of multivariate data from small populations and subgroups. However, the traditional approach is not only elaborate and complex; it could also be biased. The self-completion method is warranted in literate countries. However, the chances of respondent error are also high in the self-completion method. The timing and duration of the traditional Census should be such that the selected period should provide appropriate coverage of the entire population. The yearly updates are combined with the traditional approach when the changes in the characteristics follow a regular pattern (UNECE, 2015). These annual updates include health insurance renewal, migration data, and tax data.

In nations such as France, a rolling census approach is adopted. It depends on the frequency of update of census data and the sampling rate that depends upon the budgeting and the geographical locations that are disseminated. Although the vital advantage of this approach is to provide a higher frequency of updated data, the essential attribute of simultaneity is often lost because it does not provide a snapshot of the total population at a given time. As a result, the point of reference for comparison suffers setbacks, especially from the perspective of policy decisions (UNECE, 2018). On the other hand, various nations use the combined approach of field enumeration and administrative registers. The combined approach is to ensure cost- effectiveness to the census process and reduce the response burden on citizens. The combined approach is complemented with detailed field data for improving the accuracy of the Census.

One of the significant differences between the combined approach and the combination of the administrative registers approach is that the variables that are not available in the latter are obtained through field enumeration. The primary technical and legal conditions underpinning the combined census approach are the availability of a population register (UNECE, 2018). The hallmark of combined and register-based censuses is the administrative registers.

9

A register is defined as the systematic collection of data at the unit level, which should be organized in such a manner that there is no difficulty in updating them whenever possible and required. Updating is referred to as the process of identifiable information to establish, correct, enriching with up-to-date information, and extending the register by tracking changes in the data that was used for describing the population units and their respective correlates.

Administrative data sources refer to the data holdings that contain pieces of information which are collected primarily for administrative purposes but not for research or statistics.

Administrative data is collected by various departments of the government as well as private organizations for registration, transaction, and documentation while providing a specific service or a series of services to the residents. The administrative registers that are popular include unique identifiers as well as those without unique identifiers. The statistical registers are prepared by transforming data from the administrative registers or other administrative data sources. In most nations, administrative data is synonymous with register-based data, while other nations also consider administrative sources apart from the registers. It is necessary to identify whether administrative registers would be sufficient for the register- based approach or any other administrative sources should be used for the register-based approach.

The administrative registers essential for the combined approach include tax lists, social security files, unemployment profile, and educational qualifications. The specific legislation that governs the combined approach census is that the Census should be updated but should preserve the statistical confidentiality of the census-related information (UNECE, 2018). The population count obtained by the combined approach is better precise and sorted under various variables, unlike the traditional approach. The 2021 population census for the Czech Republic is planned as a multisource census based on field enumeration and more extensive administrative registers. The field enumeration approaches of the 2011 census were based on the population grid of usual residents. These were matched with registered residents from the CPR for the microregions, municipalities, and NUTS3. NUTS3 (Nomenclature of Territorial Units for Statistics category 3) refers to the regions of the third level in Europe (Šanda, 2011).

The usual place of residence was obtained from the census forms collected during enumeration, while the registered residential data would be obtained from the central population register (CPR) (Šanda, 2011). A CPR is a record that stores the demographic, dwelling, and socioeconomic data of the population and is updated from time to time. The nation opted for combined census approach, including enumeration and administrative registers. A total register-based census is useful if population censuses are conducted with a long-term aim when the registers get updated with time. The population census of 2021 would be obtained from a diverse set of administrative registers apart from the CPR. These registers include central insurance register (indicates the number of insured persons), social security register (number of employed and self-employed persons), pension beneficiary register, maternal beneficiary registers (individuals on maternity leave), register of unemployment (unemployed and job seekers), social supports register (individuals getting social support such as parental allowance and child care support), the System of Assistance (persons receiving social support) and several education registers that capture the data of individuals from the elementary school level to the university level (Šanda, 2011). The data from field enumeration and the administrative registers would be tallied through overlapping to prevent population data omission. One of the fundamental reasons for the use of multiple registers in the upcoming 2021 census would be to address the over coverage in CPR. The terminated insurance data on health insurance would be tallied with the CPR, and matched pairs would be excluded. Identification of the rest of the records from CPR that is not matched to any other data sources and that are unchanged over a defined period would be considered for the Census.

The availability of statistical business registers further supports the shift towards a predominantly register-based census in 2021. Including, register of census districts and buildings (that include territorial and spatial identification, month and year of completion and

10

the date when it was cleared, the ownership status of the house, the number of dwellings, ground coverage, floor space, and type of load-bearing structure and connection to sewage);

the farms register (individuals engaged in agricultural practices and their material possessions (Liberia Institute of Statistics and Geo-Information Services ..., no date).

There are different modes of electronic data collection based on technological innovations.

The electronic documentation of raw data provides timely and high-quality statistics and efficiency and cost savings related to the Census. The various types of data collection process include pencil/pen and paper questionnaire with face-to-face interviews (PAPI), computer- assisted personal interviewing (CAPI), computer-assisted telephone interviewing (CATI), a paper questionnaire with self-enumeration (PASI), and computer-assisted self-interviewing (CASI). The paper-based interview is a lengthy and time-consuming process, which has the same accuracy levels of CATI. On the contrary, CASI and PASI are more effective when the target population is knowledgeable and sensitized regarding the importance of the Census.

Another approach of census data collection is the multimodal approach which may include the combination of PAPI, CASI, CAWI, CAPI, and others. The other aspects for selecting a data collection approach for a census include the fiscal resources, human resources, availability of updated administrative registers, accurate GIS mapping of the households, and security options.

4.2. Limitations on Reliability and Methodology

One of the major issues for any census is to ensure the reliability and reproducibility of the census data in terms of demographics and socioeconomic parameters. It is not surprising that overrepresentation is one of the major issues with the combined census approach, while underrepresentation is a common feature of field-based enumeration irrespective whether a PAPI or CAPI approach was undertaken. Although the latter addresses the bottlenecks associated with the synchronization and collation of census data, the barriers related to data collection and authentication remains a major issue even with the adoption of digitized technologies while conducting population census. In this regard, various nations such as the Czech Republic and New Zealand have been able to link census forms to the CPR which have helped them to transition from a field-based enumeration to combined approach as well as exclusively register-based approach for subsequent censuses. The statistical approaches that are often undertaken to access and update CPR concerning census data include deterministic and probabilistic approaches. The variables that are used for linking census data to CPR include a personal identification number, name and surname, and birth date. The evidence suggests that the Czech Republic has 99% of its census forms linked to CPR in the 2011 census, which depicts the reliability and authentication of the census data of the nation (Šanda, 2011).

The referred percentage might be considered relatively high because of the high rates of migration within the European Union (Šanda, 2011). One of the significant drawbacks of the 1984 and 2008 census of Liberia was that the migrant and delocalized populations were not appropriately captured in the census data. Hence, the 2021 census should try to address the loopholes related to the identification of demographics of its migrant and delocalized population.

The field-based enumeration approach is insufficient to capture such data because the population census of the Czech Republic showed that the number of individuals identified in the population census through the census forms (n=1271K) and Central population register records (n=6935K) was much lesser than the census forms linked to the CPR records (n=100216K). The total number of individuals that were recognized by any source population census in the Czech Republic amounted to 108422K (Šanda, 2011). These figures suggest the importance of linking or verifying field enumeration data with CPR based census. Even after the conduct of the Census, the findings suggest that the total number of individuals registered

11

in the CPR is higher than that enumerated through an entirely field-based approach (107151K versus 101487K, respectively). Situations of such could be a point of argument that whether the CPR based method leads to overrepresentation or whether the field-based enumeration method has led to the underrepresentation of the population. These assumptions certainly create a dilemma that which administrative registers should be used in the absence of central population registers or along with field-based enumeration data to address the issues related to overrepresentation and underrepresentation as well as inconsistencies in constitutional data based on electoral areas.

In this regard, the VIP ADMIN GRANT (VAG) PROJECT for the preparation of the 2021 population census for the Czech Republic is worth mentioning. The project was initiated in January 2016 for identifying and validating the administrative registers that could be used for the population census of 2021. The project was undertaken for a specific territory, the Hradeck Kralove District, which represents only 2% of the population, but the terms of reference and the data that was selected were robust and reliable. All the records related to the administrative registers that were evaluated included exclusive and personal identifiers that were continuously updated. The referred project not only considered the CPR as the significant frame of reference but tallied the same with different administrative registers. The number of individuals identified through the CPR in the VAG (VIP ADMIN GRANT) report was 164,834 while the central insurance register exhibited a count of 183,541 (Šanda, 2011). The finding supported Central Population Register as a significant administrative register for point of reference for population census because the central insurance register had issues of overrepresentation. After all, the number of individuals who recently died was not captured in the insurance registers. On the contrary, the social security register exhibits a count of 154, 032 individuals that included the counts related to pension beneficiary register, maternity benefit register, register of unemployment (if the government extends the social security) (Šanda, 2011). Also included are the social support register that includes social support beneficiaries, parental allowance, and child-care support, as well as the system of assistance in material need register which documents individuals receiving essential social support such as housing and minimal healthcare provisions (Šanda, 2011).

These findings suggest that social security register might be an underrepresentation of population data in comparison to the central population register because the population count through the former is lesser than the latter. The high population count in the central population register could signify that the migrant populations who do not have social security are appropriately captured in the CPR records. The findings of the 2011 census of the Czech Republic substantiated this finding because the population counts through the CPR register and the CAPI based method was almost comparable, the population census undertaken by the latter approach exhibited a gap of around 0.6% compared to the former (Šanda, 2011). This discrepancy could be attributed to a migrant force that has still not enrolled for social security.

From the findings of the Czech Republic, Liberia could adopt certain administrative registers (although it does not have any CPR) to prepare a CPR equivalent along with the field enumeration data that would be obtained through the CAPI approach. Ideally, the nation should consider social security register along with the migrant data and individuals who are marginalized or homeless. Liberia should also access unemployment data if such individuals are not socially secured through any of the social security schemes such as health care plans.

4.3. Prerequisites for Transitioning to a Register-Based Census or Combined-Approach Various conditions should be made before transitioning to a register-based census from a traditional census. The major factors for the same are as follows:

12

Censuses should most often have legal underpinnings for the chosen methodological approach if there is a Statistical Act or not. In certain nations, complex legal requirements might limit access to the necessary administrative data sources that are either highly confidential or for those that can invoke legal ramifications. Often appropriate legal bases are prepared before accessing the administrative registers by the statistical offices in the absence of a National Statistical Act. Hence, the legal base should precede the adoption of combined-approach or register-based approach for population census. On the contrary, if there is no legal base on National Statistical Act, combined-approach or register-based Census should select only those variables or a selection of those variables and their derivatives that were obtained through the traditional Census. For example, combined-approach or registered-based Census should not provide much emphasis on ethnicity and religious attributes in preparing household and population census because it could increase political unrest and genocide. This is important from the perspective of Liberia that has witnessed civil wars and religious clashes.

Public Approval

Even if the national statistical offices are provided with legal approval for conducting combined and register-based censuses, public approval or political representation should be sincerely acknowledged. Apart from gaining access to sensitive information, the public in general might support the combined-approach or register-based Census because there would be fewer privacy concerns, unlike the traditional method which often suffers from poor response rate due to the same. As the register-based or combined approach inquire very few personal questions if at all, the concerned stakeholders are more likely to advocate for the referred approach. On the other hand, the public might be uneasy and apprehensive as to which data is accessed for population and housing census. Such apprehensions need not only be privacy concerns but also on the authenticity and reliability of the data. Such apprehensions are not irrational because the lack of data cleaning techniques or the inability to identify appropriate administrative sources could lead to errors in Census. Apart from the issues related to underrepresentation or overrepresentation, there could be an inconsistent representation of the constituencies. On the one hand, public approval would help to ensure that the public accepts the census data for future administrative and economic reforms or share their valid apprehensions through their elected representatives. For example, the 2008 census in Liberia invoked disputes from people's representatives regarding the representation in the national assembly. If such issues could surface from traditional approaches (as Liberia used the traditional census approach in 2008), it could be possible that the inability of census data to appropriately represent the population would invoke protests from the public in general as well as political parties. For example, the Government of India is preparing two administrative registers for identifying those who migrated from Pakistan and Bangladesh after 1949, and they should be sent back to their home countries. After nearly 70 years, such an initiative has neither rationality nor utility from the perspective of socioeconomic development of its residents. Nonetheless, the exercise has invoked futile protests from the public and political parties. Such attributes could also surface in countries such as Liberia whose migrant communities are often religion or ethnic-centric. The mass protests in India are related to the identification of minorities rather than streamlining the population register (UNECE, 2018).

Stakeholder Approval

Stakeholders or data users often want that each Census should provide the same set of information that was provided by the previous Census. Nevertheless, such requirements are not mandatory when the methodology of the Census changes. Therefore, the stakeholders

13

need to be assured that there would be no data erosion which was collected in the previous censuses. A transparent environment regarding the plans of the National Statistics Office should be encouraged. The end-users would be assured that their requirements would be taken into consideration while preparing the register-based or combined-approach with Census. The utilitarian effect of combine-approach or registered-based Census should be conveyed to all the concerned stakeholders, including the political parties and the public in general. Since Liberia witnessed protests on the 2008 census regarding the inaccurate representation of the constituencies, it will be possible to have the necessary buy-in from the relevant stakeholders for a registered-based census when they would be convinced that the population census would not be affected by the subjective bias of the enumerators due to the unavailability of the local organization to help them out in the enumeration process. Peoples representatives elected to the parliament should be sensitized on the importance of the combined-approach, and their voice should also be acknowledged (UNECE, 2018).

Comprehensive and reliable register system

This is one of the major steps for transitioning to a register-based or combined-approach.

Although the administrative quality is often not adequate, the constant interaction of the NSO and the data holders can significantly improve it over time. Statistical registers which are comprehensive and reliable in terms of accurate and timely data is necessary. Administrative resources such as the population register are not normally set up for contacting a registered- based or combined-approach for census. As a result, a suitable transformation process is necessary to create reliable statistical register systems. One of the priorities for statistical registers is a data integration and data matching. In most nations, administrative data sources are used by different public authorities. Such dynamics of data access is beneficial for developing statistical registers because the data would be updated from different sources that could range from health insurance to unemployment and social security to migration status (UNECE, 2018).

Knowledge of Administrative sources and unified identification systems

It is not advisable for the national statistical offices of different nations to follow the same set of action steps in transitioning from the traditional method to a combined approach and to a registered-based census. This is because of the diversity of the administrative resources that are available in the respective nations. Moreover, the variables related to population and housing, which are unnecessary for the respective nation should either be excluded or must not be made mandatory register-based data. For example, countries with health security and social security should not be overwhelmed with the collection of health statistics while conducting population census with the traditional and combined approaches. On the other hand, the nations which do not have the provision for health and social security should consider health statistics and relevant epidemiological data while conducting the field-based enumeration.

4.4. Difficulties Associated with Transitioning to Combined or Register-Based Census Dependence on Public Authorities

Transitioning to the register-based or combined-approach might cause significant dependence on the public authorities for the various administrative records. For example, Liberia needs to depend upon health data and unemployment registers for adequate transitioning to the combined approach. The dependence on public authorities by the statistical offices is further

14

mandated due to the concepts and definitions of the units prevailing in a country (UNECE, 2018).

Timeliness and Reference periods

Public authorities who would be entrusted to share the necessary data might not store the data in statistical formats that would cause a delay in the identification of relevant administrative data. On the other hand, the timeliness of the delivery of updated data from different sources might vary significantly. Under such instances, it would be difficult to integrate and identify the necessary data for the combined approach. Likewise, the administrative data that is sought from the public authorities in terms of statistical units might have different reference period, which further imposes a significant challenge to data integration and data matching. Ideally, the reference points for the combined approach or register-based should ensure uniformity and harmonization of the reference period to prevent the probability of overrepresentation and underrepresentation. There are certain variables which are relatively stable over time and is not impacted by a specific reference period. Contrary to this, there are a plethora of sensitive and economic variables that should have a uniform reference period to ensure utilitarian aspects of a population census (UNECE, 2018).

Difficulty in Identifying the Sub-populations

Although register-based Census or combined-approach holds the promise of universality, the range and detail of the outputs could be often limited to the variables that could be accessed only from the administrative sources. It is possible that the variables might not be pertinent to the entire population. On the contrary, certain variables that are intrinsic to a specific sub- population might not be available in the administrative registers. Under such circumstances, the quality and authenticity of the register-based census data could be argued. For example, it would be challenging to identify appropriate and relevant data on minorities or immigrants from the perspective of lingual, ethnic, and religious identity. Similarly, there would be inadequate data regarding the homeless population in most of the nations that do not have any social security scheme. The GIS systems might be helpful in developing approximation for the part of the population that is present in niches and are underrepresented. In this regard, regression and forecasting models should be implemented to enumerate marginalized populations (UNECE, 2018). Nations which are not versant with the digitized and IT-based census approaches should ensure that the IT systems are up to date.

Common Framework for Combined or Register-Based Censuses

The common framework for transitioning to combined or register-based census approach involves five key steps: identification of the appropriate data sources, transformation process, construction of the statistical registers, disseminating the outputs including data cleaning, and quality measurement and assurance (UNECE, 2018). The data sources that are commonly used for the register-based Census must include the central population register or any national register that document its population on the basis of Unique Identification number, other administrative data including administrative resources and resources that are stored by public authorities, census surveys for elections, new surveys or pilot studies for capturing a new variable or sub-populations whose representation is poor in the National Population register and other administrative sources. The major data source for register-based Census or combined-approach is the National Population register or the CPR. However, nations which do not have an established population register are exploring ways to combine different administrative sources to create an equivalent statistical population register. Although Liberia has a National Population register which would get further fine-tuned during the 2021 census, the nation could combine other administrative resources such as the social security register

15

and unemployment register to match the data in the National Population register during the 2031 census. However, the nation should delineate data on the migrant communities separately in combination with the referred registers. It might be possible that an individual who has not disclosed his or her migrant status for availing social security, they could have registered with another register.

Nevertheless, accessing various administrative resources would help to ensure the quality of the population registers as well as for updating the same. One of the major reference points for identifying an individual or a household, even in the absence of a National Population register is the Unique Identification number. The unique identifiers are not only restricted to individuals, households, and localities but also include identifiers for enterprises. Unique identifiers for enterprises and their respective employees are a major instrument for identifying the migrant workforce.

Even if there are no unique identifiers, different deterministic and stochastic models could be adopted to identify the population of interest from a dearth of administrative resources. The stochastic method would also help to create a unique identification number that could be uploaded to the national population registers. For example, the Czech Republic had started to work on a plethora of administrative resources immediately upon completion of the 2011 census. The national statistics office did set up ministries representatives who are also the important data owners of the administrative sources. Administrative data sources that are a part of the VAG project include social security registers to educational registers (Šanda, 2011).

The pilot project helps to identify the necessary administrative sources that would be essential for conducting the 2021 census. The reports of the VAG project have already reflected the registers that should be used for the Census even if CPRs are not maintained (Šanda, 2011).

The results already showed that social security registers and unemployment register are the two most important administrative documents that could be combined as estimates of the central population register. It might also be possible that certain variables that are needed to be captured in a future population census are not available in the CPR or any other administrative sources. Under such circumstances, the case study from the Dutch population census of 2011 could act as a viable reference point (Schulte Nordholt, 2018). For example, there is no register on the variable occupation. The nation has decided to adopt new software for repeated weighting. Likewise, the variable educational attainment would be based on the educational attainment file, which would be imputed. Population grid squares would be adopted as a novel method for the register-based approach. After the identification of appropriate data sets, it is necessary for analyzing the data sets for the possible variations and bottlenecks. For example, most of the registered owners in the Czech Republic were not willing to provide administrative data of external individuals (Šanda, 2011). As a result, ad-hoc procedures were implemented for the VAG project through outsourced services. In this regard, different non-government organizations could be utilized in Liberia for developing data sources on marginalized communities and sub-populations.

Data Transformation

Since data in the population registers and other administrative sources do not only involve data from the perspective of a census, the accurate transformation of said data for producing a reliable statistical register is strongly mandated. One of the major prerequisites for transforming administrative data to the statistical registers is data linking or data integration.

Unique identifiers are popular choices for the process of data integration. Nevertheless, individual nations might not have mandatory unique identifiers for all the administrative resources. Even in the absence of unique identifiers, nations could still create a unique identification code for each of its residents from a set of standard identifiers including name, sex, age, date of birth, and numerical address. It would then result in matches that could be used for producing outputs of comparable accuracy to those prepared by linking the

16

administrative sources in terms of a unique identifier. The stochastic methods provide an opportunity for the national statistical registers to create a unique identifier to ensure the integration of data from diverse administrative sources (UNECE, 2018). Therefore, there could be two approaches during the data linking process; creating a new set of unique identifiers for diverse administrative data to select the population that was not represented in the CPR or field-based traditional approaches and cleaning of the necessary data based on the unique identifiers that are already available with that of the records available and updated in the administrative sources. The VAG pilot study for the 2021 census in the Czech Republic have depicted that most of the samples carried extra records that were not valid for the reference date of initiating the pilot census (Šanda, 2011). For example, terminated studies were identified and excluded from further processing. Several person identification numbers were missed in the data sets along with the error in the PIN.

In most cases, the PINs were rectified and identified various administrative registers were integrated. However, a small number of duplicate records were found in the CPR, which was deleted. These findings suggest that the CPR would have minor problems related to deduplication and such issues primarily stem from errors in a PIN. Once, the PINs are rectified, the incidences of deduplication get minimized automatically. Although most of the administrative sources that were used for linking data with the CPR for the VPN project exhibited a match higher than 95%, the social security registers exhibited a match of only 70%.

This was not a major surprised because the Czech Republic has faced bottlenecks while conducting the previous Census due to the erroneous deletion of its residents based on language and ethnicity. On the contrary, substantial populations of Germans or Slovaks have concealed their identity to receive social security. Likewise, substantial amounts of Slovak citizens who speak Czech have been overrepresented in the population.

The gaps that are estimated in terms of linking CPR and administrative resources could be addressed by probabilistic linking methods (UNECE, 2018). Probabilistic linking methods are adopted for verifying the linked pairs (UNECE, 2018). The verification process could yield high reliability. However, the records that are not matched in various administrative sources, the national statistics office provided the Ministry of Interior with all the records that were not linked to the CPR. Although the respective ministry helped to identify 96% of the missing PINs, their reliability and validity were questionable. The primary reason for the unlinked PINs was registered place of residence of the residents outside the territory where the VAG project was conducted. Therefore, the pitfalls in the linking process were related to the difference in the place of residence rather than errors in the PINs. These findings should be translated to Liberia whereby delocalization is one of the major issues of population census in 2008. Once the data has been successfully linked, other transformation processes are installed for improving the quality of the statistical registers. The two types of approaches that are used for improving the quality of linked data include data cleaning and editing and formation of the census variables with unique algorithms. For the latter, data from different registers are effectively combined in statistical registers by adding information from additional data sources. These approaches include cleaning, deletion of erroneous values, resolving the discrepancies between sources and the reference period, coding and decoding relevant data as well as their editing, investigation, and resolution of missing values through imputation and selecting records which meet the population group of interest. These populations could include marginalized and sub-populations that are underrepresented in the administrative sources.