if it affects the monetary cost of migration but not the choice of migration (Di Falco, Veronesi, and Yesuf 2011; Pizer 2016).5 πΌ is the parameter while πππis the error component across alternative j for each migration cost. With the independence assumption, it is straightforward to retain standard estimation using Ordinary Least Squares (OLS), and the general form of CF can be expressed as CF (πππ; πΏ) and πΏ is the parameter. Therefore, the error component now consists of πππ = CF (πππ; πΏ) + πΜ. ππ
Substitute CF (πππ; πΏ) into the utility function πππ. Therefore, with a given CF specification, we can have the utility function πππ as follows:
πππ = π(πβ²ππ, ππ, ππ, πΎ) + CF (πππ; πΏ) + πΜ ππ (8)
Banteay Menchey
6 90 52 142
Battambang 6 96 49 145
Siem Reap 5 89 46 135
Total 17 275 147 422
Note: Data collection was conducted from 10th December 2019 to 18th December 2019.
Since we attempt to construct the alternative specific costs of migration, a separate version of the questionnaire was employed to obtain the recruitment fees and migration costs through official channels, particularly via the PRA. We randomly select 30 registered companies based in Phnom Penh, managing and sending labour migrants to Thailand. To identify registered companies, we use a list provided by MoLVT (2019), and the Prakas on Private Recruitment Agency by MoLVT (2013) that highlight conditions and legal status of a recruiting agency sending workers abroad.
There are 17 PRAs out of 63 active PRAs participated in the survey.
Variables
Our explanatory variables of interest are the total cost of migration and its components, based on the Global Knowledge Partnership on Migration and Development (KNOWMAD) and ILO migration cost survey across different bilateral corridors (ILO 2016b, 2018, 2020f). Migration cost is measured and classified under the SDGs indicator 10.7.1. It includes: 1) the general cost, which refers to the cost borne by potential migrants on specific training programmes such as language training or the cost of preparing to work in the destination and the cost of preparation to work in foreign country. 2) The financial cost includes the contract agreement, passport and visa, health check, and travel expenses. 3) The opportunity cost is the cost borne by migrant workers in comparison to the wage not earned when migrants travel from home to training locations or when a migrant is not earning and spends time on pre-departure training. 4) The total cost is calculated as the sum of all costs (See Table A.4 for migration cost sub-components).
It is important to note that, in this study, we justify the computation of migration costs as the ratio of the reported total cost of migration to the average household's monthly consumption expenditure prior to migration. This measurement differs from ILO-KNOWMADβs approach which measures migration costs as a share of migrant monthly wages in the host country. There are at least two main reasons for this justification. First, Cambodian labour migration decisions are characterised by a collective decision which is a joint decision of the household or the parental decisions (MoP 2013). Household members jointly determine the costs and benefits of sending one or more family members abroad. Given such consideration, the monetary costs of migration are perceived by all household members including migrants themselves. Thus, inquiring households about the amount they have spent for foreign employment can be an ideal alternative to capture the actual costs of moving. Second, a household survey can provide information about more than migration costs; it can provide information about the household's socioeconomic status, such as income, expenditure, and debt, which is vital for quantifying the migration costs in terms of household economic status instead of migrantsβ wage ratio at the destination. Martin (2017) suggests that more accurate information about migrant workers can also be obtained by conducting a survey through their households. Finally, our measurement that marginally deviates from the ILO-KNOWMADβs approach may indicate new and important evidence about how relatively expensive the cost of moving is for poor Cambodian households and low-skilled migrants from Cambodia to Thailand.
The case-specific variables include household head and household characteristics β age, gender, household size, average household education, occupation (farmer, own business, public servant, and employee), level of monthly household income, dependency ratio, level of household wealth index (poorest, poor, medium, wealthy, wealthiest) constructed by the Polychoric Principal Component Analysis (See Table A.6 in Chapter 2 Appendix). Variables at village level are
considered important in determining household migration decisions, we also include the geographic variables such as irrigation system (whether or not the household is located in a village that has an irrigation system), distance from household to school, to nearest border check-point, and to the nearest immigration office.
Migrant characteristics include the length of stay at the destination and years of education, which are the variables theoretically affecting the choice of migration route (DjajiΔ and Vinogradova 2019). The regression model's control variables include gender, marital status (single, married, widowed, and divorced), occupation at the country of destination (service, factory, construction, and fishing), and migrantsβ good health status dummy variable equal to 1; otherwise zero.
We incorporate an immigration policy variable, deportation, measured by numbers of Cambodian migrant being deported (DjajiΔ and Vinogradova 2019; Mayda 2010). The data was retrieved from the National Committee for Counter Trafficking (NCCT), reporting the number of deportees from various destinations each year from 2013 to 2018 across different check-points. Cambodian migration stock in Thailand at a provincial and regional level, retrieved from the Department of Employment, Ministry of Labour Thailand in 2018, is also included in the models to capture network effects and control for the endogeneity in migration cost.
Descriptive Statistics
Table 2.3 presents the results of the descriptive statistical analysis and additional material which can be found in the Appendix in Table A. 7 and A. 8. The simple statistical test shows a statistically significant difference between regular and irregular migration costs in all cost components. The results show that costs for irregular migration are relatively lower than for regular migration at both cost components and aggregate. The total reported cost of migration through regular channels
is 2.6 months of household consumption expenditure on average, which equals USD 458, while the total cost for irregular migration is 1.274 months of monthly household consumption expenditure, equivalent to USD 217. The result also reveals that regular migrants spend a large proportion on financial costs which constitute 76 % of the total cost including travel document fees, visas, medical test fees, contract fees, and internal and international transportation costs. In contrast, irregular migration cost is equivalent to USD 176 on average.
Table 2.3 Average Migration Costs per Migrant by Channel
VARIABLES Regular Migration Irregular Migration
Diff. in Mean
Mean SD Mean SD
Total cost of migration 2.659 1.194 1.274 1.405 1.386***
General cost of migration 0.253 0.436 0.006 0.041 0.247***
Financial cost of migration 2.098 1.244 1.016 1.426 1.083***
Opportunity cost of migration 0.127 0.221 0.010 0.064 0.117***
Notes: The cost of migration is measured by the ratio of the reported total cost of migration per person and calculated into a USD 2014 constant to the householdβs monthly consumption expenditure prior to migration. Wald test is performed to test the null hypothesis of equal means. *** p<0.01, ** p<0.05, * p<0.1
Table A.8 in Chapter 2 Appendix displays the differences in means between irregular and regular migrantsβ characteristics. The statistical test shows that our variables of interest indicate statistically significant differences within migration choice at 1 % and 5 %, respectively. First, irregular migrants tend to stay in the destination country longer than regular migrants by 45.12 months on average. This longer length of stay could result in difficulties when passing through immigration check-points if migrants want to return and re-emigrate. Furthermore, the length of stay for irregular migrants appears longer since regular migrants are contracted to a fixed time work permit of only two years with a possible two-year extension. Finally, there is a statistically significant difference between regular and irregular migrants in terms of years of education. This
irregular migrants rely on their kinships such as family/friends, informal brokers, and experienced migrants to facilitate their migration.