Chapter 3. Understanding the methodological framework in Georgia
In order to provide an empirical foundation to the analysis of the links between migration and policy, the Interrelations between Public Policies, Migration and Development (IPPMD) project used three evidence-gathering tools: household surveys, community surveys, and interviews with representatives of public, international and local organisations to provide additional qualitative information about the migration context in Georgia.
This chapter explains how the sampling for the survey was designed, as well as the statistical approaches used in the chapters that follow to analyse the impact of migration, return and remittances on key policy sectors. The chapter includes a brief overview of the survey findings, including differences across regions and between migrant and non-migrant households. It outlines some of the gender differences that emerged among migrants, particularly in terms of the country of migration, and the reasons for leaving and returning.
The Interrelations between Public Policies, Migration and Development (IPPMD) project framework is empirically based. In order to provide evidence-based analysis on the interrelationship between migration and the various sectors under study, the project carried out data collection in Georgia from July to November 2014. The OECD Development Centre developed three analytical tools for the fieldwork, each tailored to the Georgian context in collaboration with the CRRC-Georgia. These were:
-
Household surveys, of 2 260 households. The household questionnaire gathered information about individual and household characteristics related to five key development sectors: i) the labour market; ii) agriculture; iii) education; iv) investment and financial services; and v) health and social protection,1 as well as household members’ experience with emigration, remittances and return migration. It also asked about their experience of specific public policies, which may affect their migration and remitting patterns.
-
Community surveys, of 71 communities (the same communities in which the household surveys took place, as a complement to them). Respondents were district and locality leaders. The questionnaire gathered information on the community’s demographic, social and economic background as well as the existence of policies and development programmes.
-
Stakeholder interviews: 27 interviews held with representatives of government ministries, public institutions, non-governmental organisations and international organisations based in Georgia. These interviews were used to collect qualitative information on trends, policies, opinions and predictions related to various aspects of migration in the country. The information they provided helped enrich and interpret the quantitative surveys by including additional details about the specific context in Georgia.
This chapter describes how these tools were implemented, and provides a descriptive overview of the data collected. It presents the sampling design for the household and community surveys and stakeholder interviews, and outlines the analytical approach adopted in this report. It uses the survey findings to paint an initial picture of Georgia’s migration experience in terms of geography, gender and perceptions.
How were the households and communities sampled?
Households and communities were sampled using multi-stage stratified cluster sampling. The Georgian Central Election Commission’s list of 3 605 voting precincts,2 last updated before the 2013 presidential elections, was used to develop the sampling frame. The voting precincts defined the primary sampling units (PSUs).
Georgia is organised administratively into nine administrative divisions (mkhare) and two autonomous republics (Abkhazia and Adjara). The sampling frame excluded the territories of Abkhazia and South Ossetia,3 due to access issues. Precincts predominantly inhabited by ethnic minority populations (located in the administrative divisions of Kakheti, Kvemo Kartli and Samtskhe-Javakheti) were also excluded, as the survey was only conducted in the Georgian language. This reduced the potential number of PSUs to draw from by 405, for a total of 3 200 possible PSUs (voting precincts). The survey is therefore representative of Georgian-speaking households, which equates to approximately 1.8 million households in the country (90% of the country’s population) and a nearly nationally representative geographic coverage.
The precincts were grouped into strata. First the country was divided into urban (comprising two separate substrata: Tbilisi and all other urban settlements) and rural areas.4 Second, the rural and urban regions – excluding Tbilisi – were divided into four geographical quadrants: north west, north east, south west and south east.5 The project set a target of interviewing 2 000 households, consisting of 1 000 migrant households and 1 000 non-migrant ones. As the average recent response rate in non-political surveys in Georgia has been about 70% in recent years,6 a bigger sample was needed to reach the target of 2 000 completed household interviews. The targeted sample size was thus increased to 2 890, to account for projected response rates: 1 445 migrant households and 1 445 non-migrant households.7
The target sample of PSUs was set at 80, out of a possible 3 200 precincts.8 PSUs were divided into rural and urban strata, and within the urban stratum, the number of PSUs was equally divided between the capital and other urban settlements. Then, apart from Tbilisi, which formed its own geographic (urban) region, the other PSUs were distributed in proportion to the number of registered voters by the four geographic quadrants – in both urban and rural settlements.
Since data were not available on which to base a sample of households with either an emigrated household member or a returned one, all households in the 80 sampled PSUs were block listed prior to data collection. Block listing allowed all households to be classified as having a migrant or not, and ensured representative sampling of households from the lists produced.
Nine of the block-listed PSUs had fewer than seven migrant households. These 9 were dropped, leaving an overall sample of 71 PSUs.9 The PSUs were randomly selected from all nine administrative divisions as well as Tbilisi and the Autonomous Republic of Adjara (Table 3.1). Table 3.A1.1 in Annex 3.A1 presents more detailed information on PSUs sampled, broken down by geographical and rural/urban status.
Household surveys
The last stage of the sampling design involved selecting households for interview. A household was considered to be a migrant household if it had at least one current or returned migrant member who had spent at least three consecutive months in another country (Box 3.1).
A household consists of one or several persons, irrespective of whether they are related or not, who normally live together in the same housing unit or group of housing units and have common cooking and eating arrangements.
A household head is the most respected/responsible member of the household, who provides most of the household needs, makes key decisions and whose authority is recognised by all members of the household.
The main respondent is the person who is most knowledgeable about the household and its members. He or she may be the head, or any other member (aged 18 or over). The main respondent answers the majority of the modules in the questionnaire, with the exception of the return migrant module, which was administered directly to the returnees themselves. As it was not possible to interview migrants who were abroad at the time of the survey, questions in the emigrant module were asked of the main respondent.
A migrant household is a household with at least one current international emigrant or return migrant (Table 3.2).
A non-migrant household is a household without any current international emigrant or return migrants.
An international emigrant is an ex-member of the household who has left to live in another country, and has been away for at least three consecutive months without returning.1
An international return migrant is a current member of the household, who was born in Georgia, had previously been living in another country for at least three consecutive months and returned to the country.2
International remittances are cash or in-kind transfers from international emigrants. In the case of in-kind remittances, the respondent is asked to estimate the value of the goods the household received.
A remittance-receiving household is a household that has received international remittances in the past 12 months prior to the survey. Remittances can be sent by former members of the household as well as by emigrants who have never been part of the household.
← 1. Migration surveys often consider individuals to be migrants only after they have been away for either 6 or 12 months. Including shorter migration spells ensures that seasonal migrants are included in the sample (however temporary trips such as holidays are not considered in this definition). The survey also captures migration experiences that date back in time as the definitions do not put any restrictions on the amount of time that has elapsed since emigration, immigration or return migration. However, it is likely that more recent migration experiences are better captured in the survey as emigrants who left long ago are less likely to be reported by the household.
← 2. This does not include individuals who are currently in the country on vacation and/or to process their papers to work/go abroad again. However, household members who are in Georgia for the same reasons and have been in the country for at least a year are considered to be return migrants.
Both migrant and non-migrant households were sampled randomly from the list produced via block listing. The target number of households to be interviewed per PSU was 36:10 18 migrant and 18 non-migrant households. In PSUs with fewer than 18 migrant households recorded, all of the migrant households and an equal number of non-migrant households were interviewed.11 To compensate for this smaller sample size, larger samples were then randomly taken from other PSUs.
The households that were not sampled were put on a reserve list, to substitute for any sampled households where interview attempts failed. A summary of the quantitative sampling strategy is included in Table 3.A1.3 in Annex 3.A1.
The household survey was conducted by 37 interviewers and 7 supervisors from CRRC-Georgia. It took place between 18 July and 13 September 2014, following a week-long training seminar and pilot field tests led by CRRC-Georgia and the OECD. As the data collection was done electronically using tablets, extensive testing, including in the field, was done to ensure the software worked appropriately. The interviewers worked during weekdays and weekends and were instructed to visit a household at least three times before recording a non-response. A short description of the modules included in the survey is included in Table 3.A1.2 in Annex 3.A1.
Following the fieldwork, the data were tested for coherence and errors. Overall, 2 260 households were interviewed. Among them, there were slightly more urban households (1 219) than rural ones (1 041), and more non-migrant households (1 288) than migrant ones (972).12 Table 3.3 summarises the final sample.
Community surveys
In each of the 71 PSUs sampled, a community questionnaire was administered to a local government representative who was knowledgeable about the community and migration issues.13 The community surveys were conducted from 9 August to 22 October 2014, after the household surveys and after new local administration staff was appointed in the communities in question following elections of local self-government bodies in June 2014. The surveys were conducted using paper questionnaires.
The community survey included questions about the share of households that currently have a family member living in another country and their most common country of residence, as well as the most common occupational activities of those living in the community.
Obtaining accurate community-level data was a challenge. Data were most often entirely based on the opinions and estimations of the respondents because official data were only rarely available. The PSUs cover relatively small areas and statistical data is not normally gathered or analysed at this level. For example, urban municipalities covered areas much bigger than the defined PSUs – although the interviewers asked respondents to only concentrate on the geographical limits of the PSU, it was not always possible to gather data at this level. In order to account for this issue, the interviewers were asked to specify the geographical area for the questions referred to in the questionnaire.
Stakeholder interviews
In order to capture a wide range of information and opinion on the topic of migration and sectoral policies, semi-structured interviews were conducted from 17 July to 7 November 2014 using a guide developed by the OECD.
The guide was divided into five topics:
-
general awareness of migration
-
actions, programmes and policies directly related to migration
-
main actions, programmes and policies likely to have a link with migration
-
perceptions of migration-related issues
-
co-ordination with other stakeholders on migration.
Three versions of the discussion guide were developed, targeting three types of respondents: representatives of i) state institutions; ii) international organisations and iii) local NGOs and other types of organisations. Questions for each topic were modified according to whether the institution was working on migration issues directly or indirectly, and its role vis-à-vis migration policy. All versions of the discussion guide were available both in Georgian and in English and were sent to respondents on request in advance of the interviews.
The OECD and CRRC-Georgia put together an initial list of potential respondents, based on the knowledge of experts working in the field and institutions which are members of, or consult, the Georgian State Commission on Migration Issues (SCMI). During the interviews, a snowball sampling approach was employed, with all respondents asked to name other experts working in the field. When deciding which organisations to approach for an interview, CRRC-Georgia ensured that representatives of all types of relevant organisations were covered, and that none were over-represented. The original goal was to interview experts working both in the capital and outside of it but in the end only one respondent from outside the capital was interviewed. The final 27 interviewees consisted of 9 representatives of public institutions, 8 from international organisations,14 and 10 from local NGOs or academic institutions (Table 3.4).
Overall, 25 interviews were conducted in Georgian and 2 were conducted in English.15 The interviews conducted in Georgian were not translated, and were analysed in Georgian during the first phase. For the second phase, the OECD prepared a joint codebook based on preliminary analysis of the data which was then used as a conceptual framework. The codebook includes general themes (main themes and subthemes) which are common to all countries taking part in the project, but left room for adding new themes specific to a country. All transcripts were then coded according to the codebook and analysed. The results were then used in the analysis to make sense of and complement the findings.
How were the data analysed?
Having described the tools used to collect data for the project, this section provides an overview of how the data were analysed. Statistical analysis assesses the statistical significance of an estimated relationship, that is, how likely it is that a relationship between two variables is not random. The analyses in this report incorporate both statistical tests and regression analysis. Statistical tests, such as t-tests and chi-squared tests, test the correlation between two variables, without controlling for other factors. A t-test compares the means of a dependent variable for two independent groups. For example, it is used to test if there is a difference between the average number of workers hired by agricultural households with and without an emigrated member. A chi-squared test is applied when investigating the relationship between two categorical variables, such as private school attendance (which only has two categories, yes or no) by the children living in two types of households: those receiving remittances and those not. Statistical tests determine the likelihood that the relationship between two variables is not caused by chance.
Regression analysis is useful to ascertain the quantitative effect of one variable upon another, controlling for other factors that may also influence the outcome. The household and community surveys included rich information about households, their members, and the communities in which they live. This information is used to create control variables that are included in the regression models in order to single out the effect of a variable of interest from other characteristics of the individuals, households and communities that may affect the outcome, such as the household’s business investments or an individual’s plans to emigrate.
Two basic regression models are used in the report: ordinary least square (OLS) and probit models. The choice of which one to use depends on the nature of the outcome variable. OLS regressions are applied when the outcome variable is continuous (i.e. can take on an infinite number of values). Probit models are used when the outcome variable can only take two values, such as owning a business or not.
The analysis of the interrelations between public policies and migration is performed at both household and individual level, though this depends on the topic and hypothesis investigated. The analysis for each sector looks at two relationships:
-
The impact of a migration dimension on a sector-specific outcome
;
-
The impact of a sectoral development policy on a migration outcome
.
The regression analysis rests on four sets of variables:
-
Migration, comprising: i) migration dimensions including emigration (sometimes using the proxy of an intention to emigrate in the future), remittances and return migration; and ii) migration outcomes, which cover the decision to emigrate, the sending and use of remittances and the decision and sustainability of return migration.
-
Sectoral development policies: a set of variables representing whether an individual or household took part or benefited from a specific public policy or programme in four key sectors: the labour market, agriculture, education and investment and financial services.
-
Sector-specific outcomes: a set of variables measuring outcomes in the project’s sectors of interest, such as labour force participation, investment in livestock rearing, school attendance and business ownership.
-
Household and individual-level characteristics: a set of socio-economic and geographical explanatory variables that tend to influence migration and sector-specific outcomes.
What do the surveys tell us about migration in Georgia?
The migration dimensions of emigration and return were left to chance in the sampling of migrant households. Their numbers therefore reflect their relative importance. Figure 3.1 shows the prevalence of emigrant and return migrants by area, based on the household-level data. The capital Tbilisi and to a lesser extent rural areas have a relatively larger sample of return migrants compared to urban areas.
Source: Authors’ own work based on IPPMD data.
Overall, the 2 260 household surveys collected data on 8 754 individuals, as well as another 980 former household members who had emigrated. A total of 804 households had emigrants – 36% of all households in the sample (Figure 3.2, left-hand pie chart). Among the individuals currently living in the country, 308 were return migrants, and specific data about their migration experience were also collected. The 258 households with return migrants formed 11% of all households in the sample (Figure 3.2, right-hand pie chart). Ninety households (4% of the sample) have both emigrants (one or more) and return migrants (one or more).
Source: Authors’ own work based on IPPMD data.
Table 3.5 shows how household characteristics differ depending on their migration status. Households with emigrants have typically fewer members than other households, which is not surprising given that they have lost at least one member. Households receiving remittances are more likely to be in rural areas than other households. In addition, households with return migrants have lower dependency ratios than all other groups. They are also less likely to have a female household head, because most return migrants are men, who generally re-assume the position of household head on their return. Among households with no migration experience, a higher share of households have at least one member who has completed post-secondary education compared to households with emigrants, but this is probably because people who emigrate tend to be the most educated in the household.16
For the purposes of this project, a household-level wealth indicator was constructed based on questions in the household survey concerning the number of assets owned by the household, ranging from cell phones to real estate. The wealth indicator is created using principal component analysis (PCA)17 and suggests that households with migration experience tend to be wealthier.
The IPPMD survey also included a question on whether individual household members aged 15 or over planned to emigrate. The data show that plans to emigrate are more prevalent when households have migration experience. A large part of this difference can be attributed to returned migrants themselves, as 20% of them plan to emigrate again within the next 12 months.
Table 3.6 summarises the characteristics of adult individuals (15+) from the sampled households, broken down by whether they are non-migrants, returned migrants or current emigrants. Non-migrants are the oldest group, with an average age of 47, compared to current emigrants (42) and return migrants (44). Women made up 53% of the sample. While emigration seems to be a gender-balanced phenomenon with an equal share of men and women, return migrants are more often men; only 35% of returnees are women. More men than women plan to emigrate, with women accounting for only 43% of those planning to emigrate.
Among individuals without migration experience, 33% have finished post-secondary education. The share is slightly higher for emigrants (34%), while 36% of return migrants have completed post-secondary education. This may be because some of them have received education in the country of destination. Those planning to emigrate have the highest education levels, with 43% of them having completed post-secondary education.
Emigration patterns are different for men and women
Data collected on emigrants included their current country of residence, the time since they emigrated and the reason they left. Emigrants’ destination countries vary by gender (Figure 3.3). While most men had left for Russia, women chose a more diverse set of destinations – mainly Greece, Turkey and Italy (64% of women, compared to 16% of men). The main countries in the “other” category are Germany, the United States of America and Ukraine.
Source: Authors’ own work based on IPPMD data.
The main reasons given for emigrating were to look for work, take a job, or to support the family financially (together accounting for more than 80% of respondents; Figure 3.4). Emigrants who left to study abroad mainly went to Russia (29%) and Germany (29%), followed by the United States (9%) and the United Kingdom (9%).
Note: Respondents were given the chance to provide two reasons for emigrating, but only the first reason was taken into account. Countries are ordered according to the share of emigrants in that country amongst all sampled households.
Source: Authors’ own work based on IPPMD data.
About 25% of emigrants had left Georgia less than two years before the survey, 25% had left between two and five years before, and 50% more than five years before. The average time since migration was very similar for men and women, even though the percentage of migrants that are seasonal among men is twice as high as for women, at 8% and 3% respectively.
More than one in four households receive international remittances
Although emigration and remittances are closely linked, one does not necessarily imply the other. In the sample, about one in four households receive international remittances (27%). Most – but not all – households receive remittances from a former household member who emigrated; 103 (17% of remittance receiving households) receive remittances from another source. Among households with an emigrant member, 66% receive remittances, compared with 5% of households without an emigrant member. Overall, in rural areas, 29% of the households received remittances compared to 25% in urban areas (Figure 3.5).
Source: Authors’ own work based on IPPMD data.
Information was also collected on financial decisions made by households receiving remittances from a former household member. The most common activity was to repay a loan (Figure 3.6). Rural areas were particularly likely to do so, with 41% of rural households repaying loans, compared to 22% of those in urban areas.
Note: The sample only includes households that receive remittances from a former household member. The figure displays the seven most common activities reported by households. Households could specify different activities undertaken after a migrant left the household from the following list: taking a loan from a bank, paying for health treatment or schooling of a household member, accumulating savings, repaying a debt/loan, building or buying a home, investing in agricultural activities, taking out a loan from informal sources, accumulating debt, setting up a business, building a dwelling to sell to others, buying land, and restoring or improving housing.
Source: Authors’ own work based on IPPMD data.
The survey also collected detailed information on the remittances received from former members. On average, a remittance receiving household received GEL 4 310 (Georgian lari: equivalent to USD 2 450) from former household members in the year prior to the survey. The average amount sent home per emigrant who remits is GEL 4 000 (USD 2 270) per year.18 This average differs between men and women: women remit GEL 4 530 (USD 2 570) on average, while men remit GEL 3 350 (USD 1 900). Moreover, women remit more than men at different levels of formal education (Figure 3.7).
Note: The number of emigrants with either primary education only or no formal education at all was too low to be included in this figure. Remittance amounts were provided by respondents in Georgian lari (GEL), the exchange rate at 1 July 2014 was used to calculate the amount in USD. Lower secondary education includes basic education (grade 7 to 9), and upper secondary education includes general secondary education (grade 10-12) in the Georgian education system. Significance tests are indicated as follows: ***: 99%, **: 95%, *: 90%.
Source: Authors’ own work based on IPPMD data.
Most return migrants are happy to have come home
The shares of return migrants living in Georgia who had returned from Greece and Turkey are higher than the shares of emigrants currently living there. As with emigrants, the former countries of residence among return migrants differ by gender (Figure 3.8). Men return from a wider range of countries; three quarters of women return from a total of three countries, whereas only 58% of men do. The share of women who had returned from Russia is slightly higher than for men, and is much higher than the share of female emigrants living in Russia. Men mainly return from Russia, but the share is much lower than the share of emigrants currently living in Russia (as a percentage of all emigrants).
Note: The main countries in the ‘other’ category are Germany, Spain, Ukraine and the United States of America.
Source: Authors’ own work based on IPPMD data.
The reasons return migrants had emigrated are similar to those mentioned by current emigrants. The majority of returned migrants emigrated for work or financial reasons. On average, return migrants spent almost four years abroad before returning. About half of the return migrants came back because they preferred to be in Georgia. This includes returning for family reasons, for marriage, to retire or for health reasons (Figure 3.9). The survey asked return migrants whether they were satisfied to be back in in Georgia; 60% of returned migrants claim to be satisfied, although around 9% of those that claimed so plan to emigrate again in the coming year. Among those return migrants who are not satisfied, this figure rises to 37%.
Source: Authors’ own work based on IPPMD data.
Conclusions
This chapter has presented three tools – household surveys, community surveys and qualitative stakeholder interviews – used to collect data to analyse the interrelation between migration, public policies and development. The following chapters take a sector-by-sector approach to presenting the results of the data analysis: the labour market, agriculture, education and investment, and financial services.
Notes
← 1. The module on health and social protection is not used in this report.
← 2. The Central Election Commission is the supreme body of the Election Administration of Georgia. It manages and controls all levels of election commissions since 1999.
← 3. South Ossetia is not an official administrative region (mkhare) in Georgia and covers parts of the following four administrative regions: Imereti, Mtskheta-Mtianeti, Racha and Shida Kartli.
← 4. Urban and rural settlements are defined according to their official status in Georgia, as defined by the Georgian government. The division has historically been defined by the primary type of economic activity (agricultural and non-agriculture) as well as the level of infrastructure available. Changes in status are rare and the last revision was in 2013, when a number of villages around the municipality of Tbilisi changed status from rural to urban.
← 5. The north east segment includes the regions of Kakheti, Mtskheta-Mtianeti and Shida Kartli. The north west segment includes the regions of Imereti, Racha-Lechkhumi-Kvemo-Svaneti and Samegrelo-Zemo Svaneti. The south east segment includes the regions of Kvemo Kartli and Samtskje-Javakheti. The south west segment includes the regions of the Autonomous Republic of Adjara and Guria.
← 6. In the Caucasus Barometer survey, the response rate in Georgia was 68.7% in 2013, 74.6% in 2012, and 69.7% in 2011; in the 2012 Georgian Labour Survey the response rate was 82%.
← 7. The actual response rate was lower than expected: 65% among migrant and 86% among non-migrant households. To compensate for the lower non-response rate, an additional reserve sample of 59 migrant and 628 non-migrant households was selected.
← 8. The decision to sample 80 PSUs was a compromise between what was feasible and the desire to obtain the most geographically widespread sample possible.
← 9. Five of the nine dropped PSUs were in Tbilisi, two in Samtskhe-Javakheti, one in Imereti and one in Guria.
← 10. The number of 36 households per PSU was set as a function of the sample objective of 2 000 households in 80 PSUs, as well as historical non-response rates.
← 11. Both households with emigrants and households without emigrants were sampled from the lists produced after block listing, using simple random sampling. Overall, 1 430 migrant households were sampled randomly from the selected (block-listed) PSUs and in each PSU the number of sampled migrant households was proportional to the number of overall migrant households in that cluster (as detected through block listing). If in any of the PSUs proportional distribution produced a number less than six, the sample was forced at six migrant households. Because of this, the overall migrant household sample increased to 1 445. After determining the number of migrant households sampled within each PSU, the same amount of non-migrant households were selected randomly.
← 12. Due to non-responses, it was not possible to achieve an equal ratio of migrant and non-migrant households in each PSU in the final sample. Because additional households were included from randomly selected PSUs, the number of total households sampled per PSU varied.
← 13. In one PSU in Tbilisi, the community interview was conducted with a resident knowledgeable about the area rather than with a representative of local government.
← 14. One of the interviews was conducted with two respondents simultaneously, following the request of the respondents.
← 15. The stakeholder interviews were conducted by three interviewers who were trained in the methods and discussion guidelines elaborated for this project. One of the interviewers was fluent in English and, whenever needed, conducted interviews in English. All but one interview was audio recorded and transcribed in the language of the interview. One interviewee did not agree to the interview being recorded, and for this interview, the interviewer provided a detailed report.
← 16. Emigrants are not considered as household members, and therefore not taken into account in the calculation of share of households with at least one member who finished post-secondary education.
← 17. The first component created using PCA is considered to represent wealth, since wealth is assumed to account for the largest variance in the assets a household owns.
← 18. Remittance amounts were provided by respondents in local currencies. The exchange rate between the Georgian lari (GEL) and the US dollar (USD) was taken at 1 July 2014.