Civil War, Migration and the Effect on Business Cycles: The Case of Sri Lanka

Abstract

We estimated a stochastic growth model for Sri Lanka over the 1962–2015 period and found that permanent productivity shocks constituted the largest proportion of the variance in output growth. We computed correlation coefficients between permanent productivity shocks and some variables and found a negative correlation between the shocks and the growth in migration outflows. The findings are consistent with the negative effect of civil war on economic growth from an exodus of skilled workers. We further investigated the factors that drive migration from Sri Lanka to a major migration destination, namely Australia, for educated and skilled Sri Lankans. We regressed the growth rate of the Sri Lankan-born population in Australia on a civil war dummy variable, per worker Australian gross domestic product (GDP), adjusted for the restrictiveness of Australian immigration policy and an estimated labour wedge that represents labour market inefficiency in Sri Lanka. We found that Sri Lankan immigrants in Australia increased with per worker GDP in Australia, civil war and labour market inefficiency in Sri Lanka. The finding suggests that excessive protection of insiders in the Sri Lankan labour market should be abolished to mitigate migration outflows that have continued since the end of the civil war.

Keywords:

• Civil war
• migration
• labour wedge
• business cycles

JEL Codes:

• D74
• E32
• F22
• F41

Acknowledgements

The author is grateful to the Japan Society for the Promotion of Science (JSPS) for the financial support. The author would like to thank two anonymous reviewers for their constructive comments and suggestions. The author would also like to thank Enago (www.enago.com) for the English language review.

Notes

1. The source of data on GDP measured in constant local currency and the total population is the World Development Indicators on the World Bank website. The source of data on military expenditures by country as a percentage of GDP is the SIPRI Military Expenditure Database on the Stockholm International Peace Research Institute website. The estimation was carried out using RATS 6.2 and its MONTEVAR.PRG.

2. The Official Language Act of 1956 effectively prevented Tamils who were not fluent in Sinhala from being employed in the public sector. Nieto (2008) mentioned that the proportion of Tamils in civil service fell remarkably from 60% in 1956 to 10% in 1970.

3. DeVotta (2014) argued that the Official Language Act of 1956 and the 1972 constitution contributed to a superior and arrogant mind-set among new Sinhalese recruits in the universities, the military and bureaucracy.

4. This description of the Jaffna library burning comes from Feith (2010), who criticised the incident as symbolic violence by government forces against the Tamil culture and language.

5. Samaranayake (2007) noted that all LTTE members were Sri Lankan Tamils.

6. The LTTE’s bombing in October 1994 killed not only the UNP’s presidential candidate but also several former and future cabinet members and the general secretary of the party. The LTTE was known for suicide bombings. Samaranayake (2007) noted that the LTTE carried out suicide attacks far more frequently than major foreign military groups, such as Al-Qaida, Hamas and Hezbollah.

7. In April 2004, the UNP-led government lost parliamentary elections to the SLFP-led coalition. Orjuela (2005) argued that Sinhalese nationalist groups condemned the government’s concessions to the LTTE and curbed the peace process. Uyangoda (2005) argued that Sinhalese opposition parties in the parliament also undermined the ruling party’s efforts towards a negotiated compromise with the LTTE.

8. Sambanis (2002) also surveyed the literature and identified risk factors for civil war: poverty, ethnic polarisation, natural resources, ethnic diasporas, geographical dispersion, low levels of democracy and so forth.

9. Fearon and Laitin (2003) similarly pointed out rugged terrain as a risk factor for civil war because rebels could hide there from government forces. Although the Sri Lankan civil war was mostly fought in the flat Northern province, their argument may apply to the civil war in Sri Lanka for which the dense forest may have served as a hideout.

10. Brück, de Groot, and Bozzoli (2012) pointed out the difficulty in comparisons across those studies given the lack of a coherent framework of analysis.

11. Ganegodage and Rambaldi (2014) presented a Sri Lankan model in which households optimally chose their consumption path. However, instead of solving the model, they estimated an equation derived from the production function.

12. The source of the data on GDP and the population is mentioned in Note 1.

13. The source of the data on household final consumption expenditure, gross fixed capital formation and exports and imports of goods and services is the same as that of the data on GDP and the population mentioned in Note 1. The series are measured in constant local currency. The estimation was carried out using Dynare 4.4.3.

14. The small contribution of depreciation rate shocks is consistent with our expectation in the introduction: the destructive effect of civil war on the economy is negligible.

15. The source of the data on R&D expenditures as measured in percentage of GDP is the World Development Indicators on the World Bank website. The estimation was carried out using R 3.3.1 and running the MCMC simulation five thousand times.

16. The source of the data on net FDI inflows in percentage of GDP is the World Development Indicators on the World Bank website.

17. The source of the data on the population of foreign-born individuals by country of birth is the International Migration Database on the OECD website.

18. We excluded Canada from the sample because the country published the population of Sri Lankan-born individuals only in 2001, 2006 and 2011.

19. Hugo and Dissanayake (2014) referred to the 2011 Australian census and mentioned that Sri Lankan immigrants were highly educated: the ratios of immigrants with a bachelor’s degree to the Sri Lankan-born population and to the total population was approximately 30 and 15%, respectively.

20. We set the initial value of the dummy variable to 1. When the immigration policy becomes more restrictive in a year, the dummy variable decreases by 0.1 in that year and remains in that level until a further change in the immigration policy.

21. An uninformative prior distribution is specified for the coefficient parameters. An inverse gamma prior distribution with a shape equal to 1 and a scale equal to 0.0005 is assumed for the variance of the disturbance term. The estimation was carried out using R 3.3.1 and its MCMCPack.

22. The source of the data on GDP per person employed is the World Development Indicators on the World Bank website.

23. Shastri (1997) noted that the UNP’s policy against the July 1980 General Strike resulted in a decrease in the number of functioning trade unions. Teitelbaum (2007) found a significant decrease in industrial disputes in the manufacturing sector.