Demographic Data in Small Island Developing States: - - PDF document

1

Demographic Data in Small Island Developing States: State-of-the-art, Challenges and Opportunities

Alessio Cangiano & Andreea Torre The University of the South Pacific, Fiji Islands Paper for the IUSSP XXVIII International Population Conference, Session 72: Data quality and time trends Cape Town, South Africa, 29 October - 4 November 2017 Draft (please do not quote)

Introduction Out of 232 countries listed by the United Nations, 54 are Small Island Developing States (SIDS). Shared features of small insular economies and environments – such as their narrow resource base; reliance on a limited number of industries; costly per capita production, service provision and infrastructure; remoteness from markets; and vulnerability to economic shocks, political upheavals and environmental hazards (House 2013) – have catalyzed a growing interest of the international community in the unique development challenges of island states. The critical development challenges facing SIDS are compounded with recurring demographic characteristics and dynamics shared by insular populations: high population growth and density associated with rapid and concentrated urbanization; numerous examples of delayed demographic transitions caused by stagnating life expectancies and/or stalling fertility; and some of the highest emigration rates worldwide, especially highly skilled. Yet, analysis of the development implications of demographic trends has been undermined by SIDS’ limited statistical capacity and specific challenges of demographic data collection – for example, some

• f the highest per capita cost of census data gathering worldwide and the reliance of statistical

systems on foreign donors’ funding and technical assistance. As a result of the lack of an appropriate evidence base, policy formulation, implementation and evaluation, including the monitoring of the Millennium Development Goals (MDGs) and of ICPD Program-of-Action, has suffered. This poses critical challenges for the transition from the MDG framework to the post-2015 Sustainable Development Goals (SDGs) which will require significant expansion in the availability and use of demographic statistics in SIDS. Demographic research focusing on SIDS has equally been lagging behind, and has somehow lost momentum. Over the recent decades, demographers have devoted limited attention to the study of small and isolated populations – especially in comparison with other academic disciplines such as economics, geography and environmental sciences. Demographic research on SIDS has been carried out in relative isolation, with little or no systematic exercise to learn from comparative analyses of island states located in different world regions.

2

This paper reviews the state-of-the-art of population and development data in SIDS, along with the situational challenges experienced by SIDS in the production and dissemination of demographic statistics. In doing so, we attempt to identify good practices and priority actions that could be pursued to enhance the development of SIDS statistical systems. This paper relies on a combination of data and information sources. A systematic mapping of statistical capacity in SIDS is attempted based on existing quantitative measures, namely the World Bank Statistical Capacity Indicators and available data on the completeness of vital event

• registrations. OECD data (PRESS database) on international aid allocated to the development of

demographic and social statistics are also used to investigate reliance of SIDS statistical systems

• n donor assistance. Only independent countries are considered in the analysis. For

comparative purposes SIDS across regions are referred to according to geographical groupings (Pacific, Caribbean and AIMS1) and classified according to level of economic development2. Further insights on specific data sources and issues were obtained through a review of technical documents and research reports on the production and use of demographic data in SIDS. Finally, reflections provided in this paper are based on the author’s exposure to Pacific Island NSOs daily work practices over five years in his role as coordinator of the Population and Demography and Official Statistics programs at the University of the South Pacific; as member of the Pacific Statistics Steering Committee; and as researcher visiting National Statistical Offices (NSOs) around the region to access unit-level Census and immigration data. Mapping of statistical capacity The World Bank’s Statistical Capacity Indicator (SCI) provides a quantitative assessment of the capacity of a country’s statistical system. It is a composite score summarizing statistical systems’ performance in three areas: methodology; data sources; and periodicity and timeliness3. According to this measure, SIDS as a group have an average statistical capacity lower than all developing countries combined (figure 1). There is however great variation across SIDS. Some of the largest countries such as Mauritius, Dominican Republic and Jamaica have SCI scores that are higher than the developing country average. Trends of statistical capacity over the last ten years are equally diversified. While few SIDS with very low SCI scores in the mid-2000s have since experienced remarkable increases in statistical capacity (e.g. Timor-Leste and Solomon

1 Atlantic, Indian Ocean, Mediterranean and South China Sea. 2 Based on the World Bank’s classification by income groups (low, lower-middle, upper-middle and high income). 3 The first dimension, statistical methodology, measures a country’s ability to adhere to internationally

recommended standards and methods. The second dimension, source data, reflects whether a country carries out regular data collection activities, and whether key administrative data are used for statistical estimation purposes. The third dimension, periodicity and timeliness, captures the accessibility and periodicity of key socioeconomic indicators, including child and maternal mortality. This dimension aims to report the extent to which timely statistical outputs are timely disseminated for the benefits of the users. Countries are scored against 25 criteria in these three areas, using publicly available information and/or country input. The overall Statistical Capacity score is then calculated as simple average of all three area scores on a scale of 0-100. More details can be found at http://datatopics.worldbank.org/statisticalcapacity/files/Note.pdf

3

Islands), this is not a consistent outcome – with countries such as Comoros and Maldives experiencing significant capacity loss. Interestingly, the performance of SIDS in terms of statistical capacity seems to be only partly related to their level of economic development. Amongst high income SIDS, only Seychelle has a SCI score in line with the developing countries’

• average. On the other hand, there are exceptions at both ends of the spectrum, i.e. lower-

middle income countries such as Cabo Verde, Timor-Leste and Sao Tome and Principe – all former Portuguese colonies – that perform better than the SIDS average and an upper-middle income country (the Marshall Islands) that has the lowest statistical capacity of all SIDS. Figure 1 – Overall Statistical Capacity Score in selected SIDS, 2006 & 2016.

Source: World Bank Statistical Capacity Indicators

Disaggregated analysis for the three dimensions of the World Bank’s SCI reveals that the area where SIDS’s statistical capacity is lowest compared to the developing country average is the periodicity and timeliness of data dissemination (figure 2). SIDS performance in this dimension has even worsened between 2006 and 2016. On the other hand, significant improvements have been recorded for SIDS average methodology score (up from 41 to 48 in 2016), resulting in a reduced gap with other developing countries. This suggests that some progress has been made in implementing international standards in the data production cycle.

4

Figure 2 – Dimensional Statistical Capacity Scores in SIDS and all developing countries, 2006 & 2016.

Source: World Bank Statistical Capacity Indicators

A significant challenge of building statistical capacity in SIDS is the high reliance on international

• aid. According to the OECD Partner Report on Support to Statistics (PRESS) database, SIDS

countries that in the first five years of the present decade (2010-2014) have received the largest financial support for statistical activities were Haiti (USD 34 million, mostly for the preparation

• f its next Census scheduled in 2018), Guyana (USD 16 million, including a major program for

the collection of crime- and violence-related data), Guinea-Bissau (USD 8 million, primarily for strengthening nutrition statistics for vulnerable population groups), Timor-Leste (USD 6 million, 2 million of which for its 2015 national population census) and the Solomon Islands (mostly for the 2012-13 Household Income and Expenditure Survey and the 2015 Demographic and Health Survey). Yet it is when levels of support for statistical activities are evaluated on a per capita basis that SIDS countries stand out as the main beneficiaries worldwide. In 2010-2014 SIDS as a group received 2 USD per person, i.e. five times as much as the average for all developing countries (USD 0.40). Similar to the distribution of total international aid, Pacific micro-states receive the highest per capita donor assistance for statistical capacity building. For example, with a population of only 10-20 thousand people, Tuvalu, Nauru, Cook Islands and Palau received per capita statistical assistance between 40 and 90 USD. Analysis of the PRESS database also reveals that SIDS countries with lowest statistical capacity have not be receiving, in relative terms, higher support (PARIS21 2016).

5

Demographic data systems In most politically autonomous SIDS Population and Housing Censuses, supplemented by periodic household survey programs, provide the core data infrastructure for demographic and socio-economic statistics (including fertility, child and adult mortality, education, labour force and a range of development indicators used for SDG reporting). Most SIDS have a long- established track record of census enumeration dating back to colonial history. Early census counts started in the 18th and 19th centuries with the main intent of mapping the availability of manual labour in plantations and infrastructural works. By the mid-twentieth century, modern censuses were conducted in Fiji and all larger Polynesian islands – although the development of statistical sources was later-occurring in Melanesia (Lewis 1994). In the Caribbean islands of the British empire census planning in the 1960s and 1970s was a regional undertaking, with a Regional census committee taking responsibility for the harmonization of census procedures and centralized data processing hubs in the two largest islands (Trinidad and Jamaica)4. The first column of Table 1, reporting the last Census year, shows that the vast majority of SIDS completed (at least) one census enumeration in the 2010 census round. The exceptions are Haiti and Comoros, where the last censuses were carried out in 2003. On the other hand, few countries have completed census enumerations at five-year intervals (Kiribati, Samoa, Timor- Leste, Tonga). In Vanuatu, a Mini Census was carried out in 2016 to update the list of households affected by cyclone Pam and collect disaster-related information. In both the Pacific and the Caribbean concerted efforts were made to adopt harmonized census methodologies under the aegis of regional organizations with a mandate for statistical development – the Pacific Community (SPC) and CARICOM. This included common sets of core questions on internal migration, education, economic activity, disability and housing characteristics. Fertility and mortality data are also primarily obtained from population censuses using a combination of indirect methods and model estimates. Due to a combination of funding and human resource constraints quality control practices have not been systematically in place in Pacific censuses (Haberkorn and Jorari 2007). Procedures to ensure quality management of the collection and production of demographic statistics in small Caribbean countries have also been patchy (OECS 2015). However, more countries have undertaken post-enumeration surveys for the last census than for the 2000 round, especially in the Caribbean and AIMS states. In the Pacific, Palau (2015) and Tonga (2016) pioneered the regional efforts to validate census coverage. However, results of these post-enumeration surveys are typically for internal validation purposes, so a comparative assessment of the coverage of census data in SIDS remains beyond the scope of this study. Household survey programs are also an important pillar of demographic data systems in SIDS. In particular, Demographic and Health Surveys and UNICEF’s Multiple Indicator Cluster Surveys have been relied upon in many countries to estimate infant and child mortality, maternal mortality, fertility and various other social and health indicators used for SDG reporting.

4 See website of the Central Statistical Office of Trinidad and Tobago (http://cso.gov.tt/census/censushistory/)

6

Table 1 – Availability of demographic data in SIDS

Last population census Health Surveys CRVS Population (000) Year PES(a) Birth registr. (% complete) Death registr. (% complete) VSPI(b) Caribbean Antigua & Barbuda 85 2011 >90% (2001) >95% (2014) High Bahamas 351 2010 Yes 75-89% (2003) >90% (2003) Very high Barbados 226 2010 No MICS (2012) 99% (2012) >95% (2013) High Cuba 11,367 2012 MICS (2000,2006, 2010-11,2014) 100% (2014) >95% (2014) Very high Dominica 71 2011 Yes >90% (2000) >95% (2014) High Dominican Republic 9,445 2010 Yes DHS (1996, 1999, 2002,2007, 2013); MICS (2000, 2014) 88% (2014) 63% (2013) Medium Grenada 106 2011 No >95% (2015) High Guyana 747 2012 DHS (2009), MICS (2000,2006-07, 2014) 89% (2014) 91% (2012) High Haiti 8,374 2003 DHS (1994-95, 2000, 2005-06, 2012) 80% (2012) 2% (2004) Very low Jamaica 2,670 2011 Yes MICS (2005,2011) 100% (2012) 83% (2011) Medium

• St. Kitts &

Nevis 46 2011

• St. Lucia

166 2010 No MICS (2001) 92% (2012) >95% (2014) High

• St. Vincent &

the Grenadines 110 2012 Yes >90% (2001) >95% (2015) Very high Suriname 535 2004 Yes MICS (1999-2000, 2006, 2010) 99% (2010) 81% (2014) Medium Trinidad & Tobago 1,244 2011 Yes DHS (1987), MICS (2000,2006,2011) 91% (2010) >95% (2010) Very high Pacific Cook Islands 17 2016 No >90% (2013) >90% (2013) Fiji 837 2007 No 90% (2012) Low Kiribati 110 2015 No DHS (2009) 94% (2009) 50% (2011) Very low Marshall Islands 53 2011 Yes DHS (2007) 96% (2007) 70% (2006) Very low Micronesia (FSM) 103 2010 No 57% (2003) Very low Nauru 10 2011 No DHS (2007) 83% (2007) Palau 18 2015 Yes Papua New Guinea 7,254 2011 No DHS (1996,2006, 2016-17) 50-74% (2004) Medium Samoa 192 2016 No DHS (2009) 59% (2014) Very low Solomon Islands 516 2009 No DHS-MICS (2015) 80% (2007) Timor-Leste 2015 Yes DHS (2009-10,2016) 55% (2010)

7

Table 1 – Availability of demographic data in SIDS (continued)

Last population census Health Surveys CRVS Population (000) Year PES(a) Birth registr. (% complete) Death registr. (% complete) VSPI(b) Tonga 101 2016 Yes DHS (2012) 93% (2012) 77% (2004) Very low Tuvalu 11 2012 No DHS (2007) 50% (2007) Vanuatu 234 2009 No DHS-MICS (2013), MICS (2007-08) 43% (2013) AIMS Cabo Verde 492 2010 Yes DHS (2005) 91% (2010) >95% (2012) Very low Comoros 576 2003 Yes DHS (1996,2012), MICS (2000) 87% (2012) Maldives 408 2014 Yes DHS (2009), MICS (1995,2001) 93% (2009) >95% (2013) Medium Mauritius 1,237 2011 Yes >90% (2013) >95% (2014) Very high Sao Tome & Principe 179 2012 No DHS (2008-09), MICS (1996,2000, 2006,2014) 95% (2014) Very low Seychelles 91 2010 Yes >95% (2014) Notes: (a) Post-Enumeration Survey; (b) Vital Statistics Performance Index. Sources: Author's compilation from Bureau of statistics; DHS program; UNICEF-MICS; World Bank; GBD Mortality Contributors (2017).

Some SIDS especially in the Caribbean have a long-established track record of health surveys, having participated since the 1990s in multiple rounds of either the DHS (Dominican Republic, Haiti) or MICS (Cuba, Suriname, Trinidad and Tobago Sao Tome & Principe) survey programs. Most Pacific and AIMS countries have conducted their first DHS in the last decade. PNG is an exception with three surveys conducted at ten-year intervals. Vanuatu (2013) and the Solomon Islands (2015) carried out a DHS integrating some components of MICS (e.g. questions on child labour), resulting in a promising approach to maximize the efficiency of survey data collection and reduce respondent burden. Overall, since the 2000s there has been an increase in both the number of SIDS countries conducting health surveys and their frequency. Not so long ago the neglect of Civil Registration and Vital Statistics (CRVS) has been referred to as “the single most critical failure of development over the past 30 years” (Horton, 2007: 1526) and as a “scandal of invisibility, which renders most of the world’s poor unseen, uncountable and hence uncounted” (Setel et al. 2007: 1569). With significant re-prioritization and international coordination efforts deployed over the last decade, improvements in civil registration systems have been recorded in many countries. However, progress remains patchy and discontinuous (Mikkelsen et al. 2015). Worldwide, one in three children aged 5 years or younger have not had their births registered and an estimated two-thirds of deaths are not counted in the vital statistics system; more than half of WHO member states obtain either no

8

data for mortality and cause of death, or obtain data of such poor quality that they are of little value for public health policy and planning (AbouZahr et al. 2015). In Caribbean and AIMS countries significant progress has been made in extending the coverage

• f CRVS systems. Most countries have achieved complete or near-to-complete coverage of birth

registrations, with some effectively maintaining birth registers since the 1990s. The same can be said for death registrations, although making few exceptions – in particular Haiti where coverage was very low in the 2000s and with no evidence of improvement ever since. Levels of completeness are not equally high across the Pacific, where coverage of birth registration is only partial except in few countries. Death registration in Pacific countries are mostly incomplete or not routinely processed for statistical purposes – the Cook Islands being the main exception. The fitness-for-purpose of CRVS as a source of statistical data does not only require comprehensive coverage of vital events, but also the collection of high quality information and effective data processing and dissemination to make the data available to the end users. The last column of the table reports a summary assessment of the quality of CRVS systems provided by a composite metric, the vital statistics performance index (VSPI). The VSPI assesses CRVS performance through use of mortality data – on the grounds that birth registration levels are generally higher than those of death registration – taking into account six components: completeness of death reporting, quality of death reporting, level of cause-specific detail, internal consistency, quality of age and sex reporting, and data availability or timeliness, each of which captures a different aspect of data accuracy or utility (Mikkelsen et al. 2015). The index supports the argument that Caribbean countries have the most performing and Pacific countries the least performing CRVS systems among SIDS. The differences in coverage of CRVS between Pacific Islands and other SIDS may be ascribable to the more unfavorable geographic configuration whereby vast distances between islands compromise the availability of reliable and continuous communications. In addition, the low analysis and dissemination capacity may jeopardise the availability and use of birth and death

• statistics. Even where relatively complete CRVS registers are available, compilation of data for

analysis, dissemination, and policy purposes is not routinely undertaken. As a result, CRVS are rarely used in the Pacific to generate fertility and mortality statistics5, that remain primarily estimated with indirect methods from census data and model life tables. The risk of relying exclusively on model life tables is that adult mortality estimates (and life expectancy) highly depend on the reliability of the entry points (infant mortality). To shed light on some of the factors associated with incomplete CRVS it is useful to analyse the differences between population groups. Figure 3 shows the breakdown of birth registration coverage by respondent socio-economic background for selected countries with lowest coverage in their respective regions. For all the countries the primary source of data is a DHS, which should ensure a high degree of comparability across countries. In all three Pacific countries – where total coverage is in the range 40-60% – the completeness of birth

5 One notable exception is the 2011 Census report from Nauru, where birth and fertility data from vital

registrations are compared with census-derived fertility estimates for coherence check and validation purposes.

9

registrations is higher in urban than in rural areas (especially in Tuvalu). Most remarkably, there is also a significant gap between the poor and the rich. In Samoa, for example, there is a thirty percentage points difference in birth registration coverage between households in the poorest wealth quintile (47%) and those at the top of the income distribution (77%). In the Caribbean country with lowest CRVS coverage (Haiti), the completeness of birth registrations is also affected by income inequality (fig. 3.d). Further analysis of DHS data could provide deeper insights on the determinants of birth underreporting to better inform CRVS development plans. Figure 3 – Completeness of birth registrations (%) by residence and wealth quintile in selected SIDS countries

Source: UNICEF Global Database (2016) based on DHS Surveys

The statistical exploitation of other administrative records is equally underdeveloped. In particular, very few countries systematically process immigration data from arrival and departure records (that are mostly used to produce visitor/tourism statistics). For Pacific and Caribbean countries, most migration statistics are obtained from the main receiving countries of North America, Europe and the Pacific Rim, while data at the point of departure are generally limited to the estimation of the inter-censal net migration balance. To make up for this information gap various countries with high levels of outmigration – e.g. St. Lucia and Tuvalu – have included modules on emigrants (members of households living overseas) in their Census questionnaires.

10

Challenges and opportunities for demographic data systems In SIDS, the establishment of appropriate demographic data systems is undermined by specific challenges that are mostly associated with to insufficient resource allocation and the presence

• f diseconomies of scale. For example, SIDS bear some of the highest per capita cost of census

data gathering worldwide because of proportionally higher fixed costs of the data collection

• infrastructure. This is also exacerbated by the geographical configuration of large archipelagos

with many remote and geographically isolated islands scattered across vast ocean masses. A well-known constraint for National Statistical Offices in SIDS is the small human resource base. NSOs in smaller SIDS typically employ less than 20 staff – see for example OECS (2015: 63). As few as 4 or 5 staff work at NSOs in micro-states such as Niue or Nauru. This is a major structural constraint in terms of range of expertise needed to develop the full set of statistics produced by large statistical offices. In addition, Pacific NSOs have been experiencing high staff turnover, including “losses” of most senior staff members to more rewarding positions in other government departments or international statistical organizations. Haberkorn and Jorari (2007) report that in the preparation of the 2010 round Pacific censuses only three NSOs (Fiji, Samoa and Cook Islands) had staff with previous census experience while other countries (Kiribati, Tonga) had undergone 100% turnover amongst NSOs senior staff. As noted in the OECS Statistical Assessment report, «there is a fragility associated with small staff numbers, making the statistical office and its work programme vulnerable to the loss, or even temporary unavailability, of just one or two personnel. Making available, and maintaining, the entire gamut

• f skills sets within each country will almost certainly be a prohibitively expensive undertaking»

(OECS, 2015: 66). The poor development of administrative statistics is often compounded by the lack of statistical infrastructure and of a conducive authorizing environment. Much of the administrative data collected by different government ministries are either inaccessible to the NSOs or are not compiled for statistical use (OECS, 2015; SUSTINEO 2016). Data sharing between agencies and the national statistical offices is not a widespread practice due to a combination of logistic and institutional barriers. Small statistical offices often lack resources and the technical capacity to

• versee and coordinate the administrative data collection by line ministries. In addition,
• utdated statistical legislations may not provide an adequate institutional framework clearly

establishing mandates and regulating data sharing practices. This has also detrimental effects on quality of administrative data because internationally recognized frameworks for the assessment of data quality are not applied (OECS, 2015). The underutilization of CRVS systems for statistical purposes in Pacific countries calls for urgent measures to raise CRVS amongst the regional priorities of the 2030 agenda for sustainable

• development. Because of the absence of reliable vital statistics, Pacific countries are now facing

a rapid health transition but do not have timely information to guide the development of health priorities for their populations. In response to this need the Pacific Vital Statistics Action Plan has been developed under the aegis of the Brisbane Accord Group (BAG) – an international

11

partnership bringing together development agencies, academia and donors. The BAG has been supporting Pacific countries and territories in undertaking a comprehensive assessment of CRVS and developing country-specific improvement plans. The high reliance of SIDS on household surveys for the production of social statistics, and especially for SDG monitoring and reporting, is also problematic. Costs of national household surveys are proportionally higher because in small populations relatively larger samples are needed to obtain estimates that are statistically representative. This not only means higher budget requirements, but also implies a higher respondent burden. By way of comparison, in Kenya (pop. 44 million) 31,079 women were interviewed for the 2014 DHS, i.e. 2.6 respondents per 1,000 women in the target age group (15-49). With a total population of 10.3 million, 14,287 women were sampled in the 2012 DHS (5.3 per 1,000). In the Comoros (pop. 720 thousand), the 2012 DHS sample included 5,329 women, corresponding to 3% of the target group. Moving on to the smallest SIDS countries, respondent burden further increases, e.g. 12.7% of women were interviewed in Tonga, and more than 40% in Tuvalu (2007). This exemplifies how the transfer of standard sampling techniques to small populations has resulted in very high load on respondent populations. SIDS’ small size also poses another type of challenge, i.e. the statistical recording of rare demographic events such as maternal or infant mortality. This is especially problematic in small countries without comprehensive vital registration coverage, where indicators are typically derived from sample surveys (i.e. DHS or MICS). The most significant example is the maternal mortality ratio (MMR). Due to the rare occurrence of maternal deaths this indicator is typically estimated per 100,000 births, i.e. with a denominator that can be 50 or more times as high as the annual number of births occurring in smallest SIDS populations. In countries with population size of about hundred thousand or less (such as Dominica, St. Kitts and Nevis, St. Vincent and the Grenadines, Marshall Islands, Federated States of Micronesia and the smallest Pacific island countries) there are typically less than five maternal deaths recorded annually. This implies that in a year where one or two more maternal deaths are recorded, the MMR could easily increase by 50%. Survey-based estimates for some SIDS countries can produce confidence intervals of the MMR that are so large to make point estimates almost meaningless6. In fact, House (2013) notes that the increase in the MMRs observed in some SIDS is actually to be ascribed to large annual fluctuations of this indicator rather than to higher risks of maternal death. Similar spikes can occur in the trends of infant mortality due to specific and localized disease outbreaks rather than to a general worsening of child or maternal health. One way to cope with random variations of these indicators is to report multi-annual averages rather than single-year data, even if this implies compromising on the timeliness of the available statistics (Haberkorn and Jorari, 2007). A long-standing challenge for SIDS statistical systems is that there is a wealth of collected data that is poorly disseminated and insufficiently utilized to inform public and private decision

6 For example, WHO estimates for Tonga’s MMR report a lower boundary (57 per 100,000 births) and an upper

boundary (270) of the confidence interval that are, respectively, less than half and more than twice the point estimate (124 per 100,000 births).

12

• making. Back in the 1990s, it was rare for more than a single report to be published out of

national censuses in the Pacific. Census reports were typically outsourced to overseas consultants and hardly designed to be understood by users. As a result, Census data were rarely used by national development planners within the country (Lewis 1995). Delays in census data dissemination were also very common. Over the last two decades considerable progress in the dissemination of census data has been made by SIDS across regions. Countries with the more developed statistical systems have now fully-fledged dissemination strategies featuring a wide range of data dissemination products catering to different users (e.g. preliminary census counts, thematic releases, web-only data tables, infographics) and various user engagement activities (e.g. newsletters and social media

• utreach). Queryable online dissemination platforms have also adopted to allow users to

generate customized data outputs. REDATAM is widely used in the Caribbean for the processing and dissemination of Population and Housing Census data. The system generates user-defined tabulations, graphs or maps through microdata online processing. A flagship initiative in Pacific census data dissemination is the SPC’s PopGIS system7. The interface allows users to map hundreds of Census-derived indicators with fine levels of territorial granularity (down to enumeration areas), as well as to download the underlying data tables for further elaboration. For larger geographical units the system also includes survey-based indicators (such as poverty indicators based on Household Income and Expenditure Surveys). Availability in the public domain of Census and survey microdata for research has also improved with the advent of international initiatives for microdata dissemination. The International Household Survey Network (IHSN) maintains a central survey data catalog to inform data users

• f the availability of survey and census data from multiple sources8. As of October 2017, it lists

19 2000-round censuses and eight 2010-round censuses from Caribbean and Pacific countries. Some of the largest SIDS (Dominican Republic, Fiji, Haiti, Jamaica, Trinidad and Tobago) have also made readily available 10% samples of anonymized records from their national population census through the IPUMS International platform9. In spite of these improvements, the dissemination of demographic data remains constrained by various factors. Data availability is undermined by limited dissemination capacity of small national statistical offices (NSOs). Timing between collection and publication of data was identified as a significant issue of concern for data users both in the Caribbean and in the Pacific

7 See http://prism.spc.int/regional-data-and-tools/popgis2 8 The International Household Survey Network (http://www.ihsn.org) is a network of international agencies that

aims to improve the availability, accessibility, and quality of survey data within developing countries, and to encourage the analysis and use of this data by national and international development decision makers, the research community, and other stakeholders. Its central survey data catalog does not provide direct microdata access, but includes documentation, metadata and dissemination tools and recommendations to agencies that own such data.

9 IPUMS-International is an archive of anonymized samples of national population censuses from around the world.

The data are coded and documented consistently across countries and over time to facilitate comparative research and made available to qualified researchers free of charge through a web dissemination system. More information can be found at: https://international.ipums.org/international/index.shtml

13

(OECS, 2015; SUSTINEO, 2016). In smaller SIDS there is almost exclusive reliance on external donors for analysis and dissemination of the primary demographic data collections. In both Pacific and smaller Caribbean states (OECS) analysis of census and survey data is conducted thanks to the financial and technical support of international and regional organizations (UNFPA, the Pacific Community and Caribbean development bank). Nonetheless, the potential for secondary statistical exploitation of demographic data (e.g. Demographic and Health Surveys) remains largely unfulfilled. This is also due to the small domestic pool of researchers working in the countries and the general scarcity of demographic research focusing on SIDS, which has lost momentum after early interest in the peculiarity of island demographic features (McElroy 2001). Another major gap in the availability of demographic data identified by users both in the Caribbean and the Pacific is the dearth of disaggregated data for specific population groups (e.g. youth, elderly, disable, ethnic minorities). With the advent of the SDGs and their stated

• bjective of “leaving no one behind” there is growing need for such data with demographic

breakdown aligned to the development targets. As a matter of fact, this relates more generally to the disproportionate compliance costs experienced by SIDS for the international statistical reporting obligations (Cook 2016). Therefore, making greater use of national statistics and improving the dissemination of data disaggregated by sex, age and other demographic variables were identified as priorities for statistical development in the main SIDS’ Action Document (SAMOA Pathway) and will be a necessary condition to fulfil for SIDS’ transition from the MDG framework to the post-2015 SDGs – which, with 17 goals, 169 targets and 300+ monitoring indicators, will require much enlarged data availability in the public domain. The way forward The picture portrayed in this study of demographic data systems in SIDS is one of light and

• shade. SIDS as a group have lower statistical capacity than the average of developing countries,

facing significant structural constraints for demographic data development due to their geography, volatile economic performance, small skill base in the workforce and user demand. However, generalizations are inappropriate given the significant differences between SIDS in the Pacific, Caribbean and AIMs regions that reflect their regional location, size, history, institutional capabilities, technology infrastructures, regional and institutional partnerships and the place and role of NSOs in the country. Significant progress has eventuated especially in some of the largest countries thanks to better quality control procedures in national censuses, periodic household surveys, more complete CRVS registrations, and the use of more comprehensive data dissemination strategies. In turn, smallest SIDS may simply lack the scale needed to maintain the capability to develop the full range of demographic statistics produced by large countries. The investment required in these small countries to meet UN and other international statistical standards such as the SDG monitoring system would be a prohibitive share of their government budgets and, most importantly, may not be fully justified by domestic needs for development

• planning. Reliance on international donor funding and capacity supplementation – at least in

the most specialized areas of expertise – are likely to continue to be a necessity (Cook, 2016).

14

Despite the significant challenges, the small scale and logistic disadvantage of SIDS might also provide fertile ground for devising new strategies and experimenting technological innovations. Some SIDS are making significant investments in the use of technology to streamline and cut the costs of data collection. For example, the last Census in Tonga (2016) has been entirely conducted with tablet-based Computer Assistant Personal Interview (CAPI) applications (a project supported by Australian aid). This has enhanced not only the quality of the data (e.g. with information on GPS locations), but also the timeliness of data processing and

• dissemination10. The use of GPS technology and GIS systems in the collection and analysis of

population data has also become a widespread practice in many SIDS. Address registers have been maintained in several countries (e.g. Cook Islands and Samoa), paving the way for the establishment of population registers that could considerably enhance the quality of population estimates between censuses and reduce respondent burden. These examples showcase how SIDS countries can be perfectly suited to adopt cutting edge technology in the data cycle and even to pilot new IT-intensive operational practices (Cook, 2016). Other opportunities for SIDS to lead the way in the route to data-driven development are being pursued to tackle SIDS’ specific areas of disadvantage, e.g. in the use of demographic data and techniques for vulnerability assessment. The scope for applying integrated methodologies to inform disaster relief interventions in SIDS is enormous. This is demonstrated for instance by the damage assessment maps for cyclone Pam and cyclone Winston recently developed by the Secretariat of the Pacific Community using GIS mapping of Census data and satellite imagery. An example is showed in figure 3. A promising approach to generate updated information is also provided by the linking of conventional demographic data with ‘Big Data’ sources. This has been done, for example, to track population movements before and after the 2010 Haiti earthquake (Bengtsson et al. 2011) using Digicel data (a leading mobile network provider in SIDS). The incorporation of demographic and big data into environmental risk assessment frameworks and disaster management strategies is still in its infancy, but provides tremendous scope to achieve the ‘Data revolution’.

10 Preliminary population counts of Tonga’s 2016 Census were disseminated in the first quarter of 2017 within only

four months since the data collection. Another example of timely dissemination of preliminary census release comes from the Cook Islands 2016 Census (conducted in December 2016).

15

Figure 4 – Damage assessment map for cyclone Winston in Fiji, 2016. Bibliography Bengtsson, L., Lu, X., Thorson, A., Garfield, R., & von Schreeb, J. (2011). Improved Response to Disasters and Outbreaks by Tracking Population Movements with Mobile Phone Network Data: A Post-Earthquake Geospatial Study in Haiti. PLoS Med, 8(8), e1001083. Bureau of Statistics, Department of Health and the Pacific Community (2015). Nauru Vital Statistics Report 2008-2013. Nauru: Bureau of Statistics and Department of Health.

16

Cook, L (2016). Agenda 2030 and the SIDS: Strengthening Statistical Capacity and Readiness. OECD-PARIS21 SIDS Forum, Paris, May 2016. Haberkorn, G. & Jorari A. (2007). “Availability, Accessibility and Utilization of Pacific Islands Demographic data – Issues of Data Quality and User Relevance”. Asia Pacific Population Journal, 22(3): 75-95. House, W.J. (2013). Population and Sustainable Development of Small Island Developing States: Challenges, Progress made and Outstanding Issues. Population Division Technical Paper No. 2013/4. New York: United Nations Department of Economic and Social Affairs. Iorangi, T., Tangimetua A. and the Pacific Community (2015). The Cook Islands Vital Statistics Report 1999‐2013. Rarotonga, Cook Islands: Ministry of Health. Lewis, L. (1994). “Statistical Services in the Pacific: Some Personal Reflections”. Discussion Paper

• No. 5, UNFPA COUNTRY SUPPORT TEAM Office for the South Pacific. Available at:

http://www.un.org/popin/regional/asiapac/fiji/disc/disc5.htm Lewis, L. (1995). “Hitch-hiking Along the Super-Highway: Redirection for Population Statistics in the Pacific”. Discussion Paper No. 15, UNFPA COUNTRY SUPPORT TEAM Office for the South

• Pacific. Available at: http://www.un.org/popin/regional/asiapac/fiji/disc/disc15.htm

Mikkelsen, L., Phillips D., AbouZahr C., Setel P., de Savigny D., Lozano R. and Lopez A. (2015). “A global assessment of civil registration and vital statistics systems: monitoring data quality and progress”. The Lancet, 386(10001): 1395-1406. Naidu, S., Buttsworth M. and Aumua A. (2013). “Strengthening civil registration and vital statistics systems in the Pacific: the Fiji experience”. Working Paper No. 35., Health Information Systems Knowledge Hub, Brisbane: The University of Queensland. OECS (2015). Assessment of the Regional Statistical System (RSS) of the OECS Region to inform an OECS Regional Strategy for the Development of Statistics (RSDS). St. Lucia: OECS. PARIS21 (2016). Partner Report on Support to Statistics 2016, Partnership in Statistics for Development in the 21st Century, Paris: OECD. Setel, P., Macfarlane S., Szreter S., Mikkelsen L., Jha P., Stout S. and AbouZahr C. (2007). “A scandal of invisibility: making everyone count by counting everyone”. The Lancet, 370(9598): 1569-1577. SUSTINEO (2016). DFTA Pacific Statistics Assessment - A survey of user and producer

• perspectives. Canberra: SUSTINEO

United Nations Statistics Division (2010). Post Enumeration Surveys – Operational Guidelines. New York: UNDESA.