Winners and Losers Among a Refugee-Hosting Population *CORE and Université catholique de Louvain, *Fund for Scientific Research (FWO), University of Antwerp 17 January 2010, Royal Economic Society Fifth PhD Presentation Meeting
11.4 Mio refugees (UNHCR 2007) : a visible phenomenon...... but few economic analysis Beyond some popular misleading views on refugees Not temporary protracted refugee emergencies Not passive refugees bear economic functions Research questions Does the establishment of a refugee camp affect the local population? By which channels? Do people adopt coping strategies?
Research Question 1 Research Question 2 Research Question 3 Identifying assumptions Relaxing assumptions
Inflows from Burundi and Rwanda Kagera Quasi-natural experiment Unexpected in 1993 (Burundi) and 1994 (Rwanda) More than one half of Kagera population Movement restrictions to 4 km Good markets + demand : + food prices Labour markets Wage effects Business boom transport costs; Improved health and sanitation services Negative externalities: Environmental degradation; spread of disease; security issue
3 Research Questions Problem : No idea about the magnitude, the persistency and the distribution of these reported effects RQ1: Do economic benefits compensate for negative externalities? RQ2: Is the impact differentiated? RQ3: Non-monotonic relationship between proximity and welfare?
RQ1: Net economic benefits? Research Question 1 Research Question 2 Research Question 3 Identification Strategy KHDS dataset, selecting people interviewed before 21 October 1993 and re-interviewed in 2004 : Distance ( to refugee camp ) and refugee population RI v(h),t = log 13 c=1 pop c /d v,c log(v h,t ) = β 0 + β 1 (RI v(h),t ) + β 3 Z h,t + β 4 Q v(h),t + α t + α h + α v + ɛ h,t (1) V h,t : cons per adult equi., correcting for price differences Z household characteristics, Q village characteristics
Research Question 1 Research Question 2 Research Question 3 RQ1: Results (1) (2) (3) (4) POLS VFE HHFE HHFE(+) Log(V h,t ) Log(V h,t ) Log(V h,t ) Log(V h,t ) Log(RI) -0.0985*** 0.0669** 0.0616* 0.0829** (0.0264) (0.0328) (0.0320) (0.0338) Activity h,1991 incl. drop drop drop Z h,t incl. incl. incl. incl Q v,t incl. incl. incl. incl α t incl. incl. incl. incl. α v incl. incl. α h incl. incl. Obs. 4220 4220 4220 4220 R 2 0.259 0.370 0.279 0.309 On average, positive impact from refugee inflows
RQ2: Impact differentiated? Research Question 1 Research Question 2 Research Question 3 Theoretical model predicts: Losers: Agricultural worker Winners: Self-employed in agricultural and non-agricultural activities log(v h,t ) = β 0 + β 1 (RI v(h),t ) + β 2 Activity h,1991 RI v(h),t + β 3 Z h,t + β 4 Q v(h),t + α t + α h + α v + ɛ h,t (2) Net gains unevenly distributed Losers: Agricultural worker Winners: Self-employed farmer Discrepancy with theoretical model: self-employed in non-agricultural activities
RQ2: Results Introduction Research Question 1 Research Question 2 Research Question 3 (1) (2) (3) (4) POLS VFE HHFE HHFE(+) Log(V h,t ) Log(V h,t ) Log(V h,t ) Log(V h,t ) Log(RI) -0.118*** 0.0481 0.0426 0.0623* (0.0265) (0.0329) (0.0323) (0.0339) Agri. Self 0.00810** 0.00578 0.00751* 0.00796** *Log(RI) (0.00399) (0.00378) (0.00397) (0.00402) NonAgri. L -0.00970* -0.00965* -0.00664-0.00658 *Log(RI) (0.00549) (0.00537) (0.00549) (0.00552) Agri. L -0.00778-0.00937** -0.00975** -0.00878* *Log(RI) (0.00513) (0.00468) (0.00487) (0.00493) NonAgri Self -0.0212*** -0.0160*** -0.0164*** -0.0179*** *Log(RI) (0.00422) (0.00398) (0.00420) (0.00424) Z h,t incl. incl. incl. incl Q v,t incl. incl. incl. incl α t incl. incl. incl. incl. α v incl. incl. α h incl. incl. Obs. 4220 4220 4220 4220 R 2 0.285 0.386 0.290 0.320
Research Question 1 Research Question 2 Research Question 3 RQ3: Non-monotonic relationship? (1) (2) (3) (4) (5) (6) Log(V h,t ) Log(V h,t ) Log(V h,t ) Log(V h,t ) Log(V h,t ) Log(V h,t ) Log(RI) -1.304*** -1.341*** (0.451) (0.475) Log(RI) 2 0.486*** 0.504*** (0.168) (0.176) Log(Prox) 0.193*** 0.236*** 2.156* 1.989* 0.612*** 0.639*** (0.0724) (0.0754) (1.171) (1.210) (0.198) (0.205) Log(Prox) 2-0.431* -0.385[0.151] (0.259) (0.268) Z h,t incl. incl. incl. incl. incl. incl. Q v,t incl. incl. incl. incl. incl. incl. α t incl. incl. incl. incl. incl. incl. α v incl. incl. incl. α h incl. incl. incl. incl. incl. incl. Obs. 4238 4238 4238 4238 4238 4238 R 2 0.277 0.307 0.278 0.308 0.280 0.311 Controlling for the average refugee population, inverted-u shape relationship between distance and welfare Controlling for proximity, existence of a scale effect
Identifying assumptions Relaxing assumptions Exogeneity Quasi-natural experiment (Friedberg and Hunt 1995: 36): Episodes where the timing and location of immigration may be politically rather than economically motivated, [...] reduces the problem of immigrants choosing location based on their labor market conditions >< Attrition : (Traced) remaining households more able to adjust? >< Baez (2008: ): endogenous location? Poorer regions that are highly disadvantaged in some unobservable domains may be relatively more likely to [...] host more refugees (e.g. limited institutions and systems to control their arrival or assist them). Common trend assumption: similar trajectory in absence of refugees? (Lee and Kang 2006)
Attrition? Introduction Identifying assumptions Relaxing assumptions Figure: Own calculation based on KHDS P(Migration = 1 x) = Φ(β 0.547 (RI v(h),t ) + β n.s. Activityh,1991 RI v(h),t + βzh,1991 + βq v(h),1991 + αt + ɛh,t) (3)
Exogeneity? Introduction Identifying assumptions Relaxing assumptions Baez >< Border out of control of local authorities Educated guess? (1) (2) (3) (4) (5) (6) Log(RI) Log(RI) Log(RI) Log(RI) Log(RI) Log(RI) Log(V h,1991 ) -0.205*** -0.103*** -0.241*** -0.137*** (0.0197) (0.0191) (0.0232) (0.0215) 93 91 Log(Cons) 0.00940-0.0773*** -0.0198-0.0665*** (0.0154) (0.0175) (0.0141) (0.0157) Z h,1991 incl. incl. incl. Q v,1991 incl. incl. incl. Obs. 2360 2360 2318 2226 2318 2226 R 2 0.047 0.232 0.000 0.051 0.219 0.240 Robust to sample restricted to the two bordering districts
Identifying assumptions Relaxing assumptions Common trend assumption? Placebo test (1) (2) (3) (4) (5) (6) (7) (8) Log(V h,t ) Log(V h,t ) Log(V h,t ) Log(V h,t ) Log(V h,t ) Log(V h,t ) Log(V h,t ) Log(V h,t ) Log(RI v,1993) -0.0169-0.0169 0.0182 0.0194-0.0205-0.0205 0.00141 0.00264 (0.0250) (0.0250) (0.0244) (0.0244) (0.0401) (0.0401) (0.0243) (0.0244) True Log(RI) 0.0566* 0.0633* 0.0307 0.0384 (0.0313) (0.0324) (0.0314) (0.0323) Agri. Self(91) 0.0184*** 0.0184*** 0.0236*** 0.0236*** *Log(RI v,1993) (0.00530) (0.00530) (0.00332) (0.00334) NonAgri. L(91) 0.00788 0.00788 0.00192 0.00195 *Log(RI v,1993) (0.00689) (0.00689) (0.00385) (0.00387) Agri. L(91) -0.000820-0.000820-0.00375-0.00367 *Log(RI v,1993) (0.00677) (0.00677) (0.00427) (0.00429) NonAgri Self(91) -0.00484-0.00484-0.00463-0.00461 *Log(RI v,1993) (0.00584) (0.00584) (0.00335) (0.00337) Act h,1991 idem idem *Log(RI) idem idem α t incl. incl. incl. incl. incl. incl. incl. incl. α v incl. incl. incl. incl. α h incl. incl. incl. incl. incl. incl. incl. incl. Obs. 4771 4771 6649 6649 4771 4771 6649 6649 R 2 0.693 0.693 0.839 0.841 0.696 0.696 0.841 0.844
Identifying assumptions Relaxing assumptions Identifying Assumptions Exogeneity of the camp location Qualitative argument: No endogeneity! Unlikely endogenous choice Lower-bound estimate Common trends assumption Placebo Test Relaxing theoretical assumptions Allowing for geographical mobility Refugee camp : not a dispersion force Including migrants: Lower-bound estimate and geographical adjustment of labour Allowing for occupational mobility: Lower-bound estimate
Introduction On average, positive effect... but unevenly distributed: Looser: Agricultural worker Winner: Self-employed farmer (conditional on common trend assumption) Self-employed in non-agricultural activities? possible reconciliation: Increased competition (selection) following entry of new businesses Migration as a coping strategy Minimum mass of refugees needed for positive externalities Further works: open the black box External validity? Provision of local public goods: impact on heath status? (with Ph Verwimp, Antwerp) Agglomeration economies may play a role, with G. Duranton (Toronto) and J. Thisse (UCL)? Impact on social cohesion or trust?