Resource discoveries and FDI bonanzas Gerhard Toews University of - - PowerPoint PPT Presentation

Resource discoveries and FDI bonanzas

Gerhard Toews University of Oxford Pierre-Louis Vézina King’s College London March 27, 2017

Intro Identification and Estimation Data Results Appendix

New discoveries of natural resources in several African countries including Ghana, Uganda, Tanzania and Mozambique raise an important question: will these windfalls be a blessing that brings prosperity and hope, or a political and economic curse, as has been the case in so many countries? Joe Stiglitz (2012)

Intro Identification and Estimation Data Results Appendix

What happens to a developing economy following a large oil

• r gas discovery before production starts?

Intro Identification and Estimation Data Results Appendix

What we know

The prospect of resource wealth unleashes political forces (Venables, 2016). In countries with large discoveries (public) investment increases right after the news shock hits, creating a pre-boom boom (Arezki et al., 2017). ⇒ Discoveries have economic consequences, before production starts.

Intro Identification and Estimation Data Results Appendix

What we do

1 We look at the effect of large oil or gas discoveries on foreign direct

investment (FDI) into developing economies.

2 Using Mozambique as a case study we explore the consequences of

FDI on local labour markets.

Intro Identification and Estimation Data Results Appendix

Why FDI?

FDI is a key part of economic development (Hirschman, 1957)

Important source of finance for developing countries; Transfers technology, skills, management practices; Creates higher-paid jobs.

Intro Identification and Estimation Data Results Appendix

Why we care?

1 We know little about the effect of news in developing economies

where FDI is likely the main source of investment.

2 In light of the resource curse it is important to evaluate whether oil

and gas discoveries attract or deter FDI.

Intro Identification and Estimation Data Results Appendix

Results preview

In the 2 years following a large discovery: Non-extraction FDI inflows increase by 73% Number of FDI projects increase by 37% Number of sectors and source countries increase by 20% Number of jobs created increase by 20%

Intro Identification and Estimation Data Results Appendix

Discoveries in non-OECD countries (since 2003)

Source: Horn, M. and Myron K. 2011. Giant Oil and Gas Fields of the World. Giant means at least a total of 500 million barrels of ultimately recoverable oil equivalent

Intro Identification and Estimation Data Results Appendix

Data on large discoveries

AGO AGO AGO ETH GHA GHA LBR MOZ MOZ MOZ MOZ SLE TZA TZA AZE AZE CHN CHN CHN CHN CHN CYP IND KAZ MYS VNM VNM VNM RUS RUS RUS RUS BOL BRA BRA BRA BRA BRA BRA BRA COL VEN EGY IRN IRN IRN IRN IRN IRN IRN IRN IRQ IRQ IRQ IRQ LBY SAU SAU SDN

1 3 10 40 100 500 1500

Discovery discounted net value (GDP=100)

2002 2004 2006 2008 2010 2012

Intro Identification and Estimation Data Results Appendix

Are large discoveries random?

Probability of a large discovery is 2-3%; Several studies treat large discoveries as an exogenous source of variation (Arezki et al., 2017; Tsui, 2011; Lei and Michaels, 2014). ⇒ Timing of giant oil discoveries is plausibly exogenous and unpredictable due to the uncertain nature of exploration.

Intro Identification and Estimation Data Results Appendix

Structural Basins

Intro Identification and Estimation Data Results Appendix

Drilling in non-OECD countries

Intro Identification and Estimation Data Results Appendix

Giant discoveries in non-OECD countries

Intro Identification and Estimation Data Results Appendix

Are large discoveries random?

An example of the uncertain nature of explorations: In 2010 Lundin Petroleum made the largest discovery of the year (and one of the biggest ever for Norway). It was found within 3 meters of where Elf Aquitaine drilled but failed to find oil in 1971.

Intro Identification and Estimation Data Results Appendix

Estimation

FDIit = βDit + αi + σt + ǫit Dit which is dummy equal to 1 in the year of the discovery and the two subsequent years. Country fixed effects (αi) pick up factors that vary little year-on-year such as institutions or market potential. Global factors such as the oil price are picked up by year fixed effects (σt).

Estimation

Intro Identification and Estimation Data Results Appendix

Discovery countries vs. synthetic counterfactuals

2000 4000 6000 8000 Non−extraction FDI 2003 2005 2007 2009 2011 2013 GHA synthetic GHA 1000 2000 3000 4000 5000 Non−extraction FDI 2003 2005 2007 2009 2011 2013 ETH synthetic ETH 2000 4000 6000 8000 Non−extraction FDI 2003 2005 2007 2009 2011 2013 MOZ synthetic MOZ 500 1000 1500 2000 Non−extraction FDI 2003 2005 2007 2009 2011 2013 TZA synthetic TZA

Intro Identification and Estimation Data Results Appendix

Data on FDI

1 Source: fDiMarkets (part of the Financial Times Group); 2 Data at the project level on value of investment and jobs created; 3 As a novelty, the data allows us to decompose FDI into extensive

and intensive margins (number of projects vs. average value of projects, as well as number of sectors and source countries).

Intro Identification and Estimation Data Results Appendix

FDI to discovery countries

100 200 300 400 500

USD billion

2003 2005 2007 2009 2011 2013 Extraction FDI Total FDI

Note: Extraction FDI is as defined by fDi Intelligence.

Intro Identification and Estimation Data Results Appendix

Results - Non-extraction FDI

(1) (2) (3) (4) FDI (USD million) Nb projects Avg project size Jobs created Discovery in past 2 years 0.594∗∗ 0.303∗∗ 0.314 0.549∗ (0.264) (0.126) (0.211) (0.251) N 1080 1080 1080 1080 R-sq 0.72 0.90 0.41 0.75

Intro Identification and Estimation Data Results Appendix

Results - Extensive margin

(1) (2) (3) Nb source countries Nb sub-sectors Nb sectors Discovery in past 2 years 0.188∗∗ 0.193∗ 0.158∗∗ (0.078) (0.088) (0.071) N 1080 1080 1080 R-sq 0.86 0.89 0.86

Intro Identification and Estimation Data Results Appendix

Discovery effect on FDI by business activity

Largest city Rest of country

Extraction Logistics EducationTraining DesignDevelopmentTesting Recycling SalesMarketingSupport TechnicalSupportCentre Electricity MaintenanceServicing CustomerContactCentre Headquarters ResearchDevelopment SharedServicesCentre BusinessServices Construction Retail ICTInternetInfrastructure Manufacturing

−3 −2 −1 1

Discovery effect MaintenanceServicing EducationTraining Recycling TechnicalSupportCentre DesignDevelopmentTesting ResearchDevelopment CustomerContactCentre Headquarters SharedServicesCentre Manufacturing SalesMarketingSupport Retail Logistics ICTInternetInfrastructure BusinessServices Construction Extraction Electricity

−3 −2 −1 1 2

Discovery effect

Intro Identification and Estimation Data Results Appendix

Robustness

Previous number of giant discoveries Change the counterfactual Something specific about the countries (eg openness)

Randomization

Flexible specification

Leads and Lags

Different time horizon

Time Horizon

UNCTAD Data

UNCTAD

Heterogeneity

Heterogeneity

Intro Identification and Estimation Data Results Appendix

Conclusion

Large oil and gas discoveries lead to FDI bonanzas; FDI in non-extractive sectors increases by around 70%; Driven by the extensive margin, i.e. by new projects, in new sectors, from new source countries;

Resource discoveries and FDI bonanzas

Gerhard Toews University of Oxford Pierre-Louis Vézina King’s College London March 27, 2017

Intro Identification and Estimation Data Results Appendix

Estimation

To include the zeros in the FDI data we use an inverse hyperbolic sine transformation (Burbidge et al., 1988; MacKinnon and Magee, 1990); ... or a Poisson pseudo-maximum likelihood estimator (Silva and Tenreyro, 2006). We cluster standard errors two ways, by year and country.

Back

Intro Identification and Estimation Data Results Appendix

100 placebo discovery effects (Within country shuffles)

Discovery effect .5 1 1.5 2 Density −1 −.5 .5 1 Estimated placebo coefficent

Back

Intro Identification and Estimation Data Results Appendix

Leads and lags: First discoveries

−2 −1 1 2 3 Effect on FDI −2 −1 Discovery +1 +2

Back

Intro Identification and Estimation Data Results Appendix

Discovery effect on FDI: Varying time horizons

Discovery year Discovery year + 1 Discovery year + 2 Discovery year + 3 Discovery year + 4 Discovery year + 5

−.5 .5 1 1.5 −.5 .5 1 1.5 −.5 .5 1 1.5 −.5 .5 1 1.5

FDI (USD million) Nb projects Avg project size Jobs created

Back

Intro Identification and Estimation Data Results Appendix

Robustness to UNCTAD data and longer time period

Period 1970-2014

(1) (2) (3) FDI FDI FDI Discovery in past 2 years 0.484∗∗ 0.486∗∗ 0.434∗∗ (0.185) (0.185) (0.166) N 8731 7523 6527 R-sq 0.73 0.74 0.75 Sample countries Non-OECD Exploration Discovery

Period 2003-2014

(1) (2) (3) FDI FDI FDI Discovery in past 2 years 0.488 0.460 0.525 (0.301) (0.299) (0.307) N 1992 1080 300 R-sq 0.81 0.74 0.65 Sample countries Non-OECD Exploration Discovery

Back

Intro Identification and Estimation Data Results Appendix

Heterogeneity of the FDI effects across countries

−1 1 2 3 Discovery effect on non−extraction FDI inflows 2000 4000 6000 8000 10000 GDP per capita (2005 USD) −.5 .5 1 1.5

Discovery effect on non−extraction FDI inflows

5 10 15

Nb of previous discoveries

−1 1 2

Discovery effect on non−extraction FDI inflows

−2 −1 1 2

Rule of law

−1 1 2 3 Discovery effect on non−extraction FDI inflows 1.5 2 2.5 3 3.5 4 CPIA property rights and rule−based governance rating (1=low to 6=high)

Back