Assessing economic impact of sovereignty transfer over Hong Kong: a synthetic control approach

ABSTRACT

This article compares the trajectory of Hong Kong’s purchasing power parity–adjusted GDP per capita before and after the 1997 handover with the trajectory of a weighted combination of similar economies, using weights determined endogenously by data. The synthetic Hong Kong is constructed to provide the counterfactual of what would have happened to Hong Kong economy in the absence of transfer of sovereignty. We find that Hong Kong economy is negatively affected, and the gap between actual and synthetic Hong Kong from 1997 to 2001 on average is 7% of the 1996 level if Japan receives the greatest weight. The average negative effect becomes 9% if Singapore receives the greatest weight.

KEYWORDS:

• Hong Kong
• synthetic control
• handover
• sovereignty
• GDP

JEL CLASSIFICATION:

• N95
• P52
• O47

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ An international dollar is a hypothetical currency that has the same purchasing power as the U.S. dollar does in the United States. It is commonly used to compare living costs or standards across countries.

² For a summary of recent protests see the BBC report titled ‘The Hong Kong protests explained in 100 and 500 words’ https://www.bbc.com/news/world-asia-china-49317695 A summary of Hong Kong security law can be found at https://www.bbc.com/news/world-asia-china-52765838.

³ Goldman Sachs projected that capital would flow from Hong Kong to Singapore because of the

ongoing protests, see https://finance.yahoo.com/news/hong-kong-may-lost-us-093000630.html

⁴ We use Indonesia, Malaysia, Philippines, and Thailand as control groups only in Robustness Check section since their living standards were much lower than Hong Kong before and after 1997.

⁵ Other prominent confounding events include the burst of the dot-com bubble in 2001, the 2002–2004 outbreak of severe acute respiratory syndrome (SARS), and the announcement of Mainland and Hong Kong Closer Economic Partnership Arrangement (CEPA) in 2003..

⁶ PCA is a computationally low-cost method for data reduction. The weights from PCA are computationally more robust than the highly nonlinear synthetic control method.

⁷ The pre-1997 high inflation in Hong Kong was due to labour shortage, land shortage, rapid economic restructuring as well as economic integration with mainland China, see Liu (1999).

⁸ In other words, the average GDP of the whole training period (1989–1993) would substantially under-predict the value in the validation period.

⁹ Later, we will show that total factor productivity is highly correlated with GDP, and that may explain why the weight for lagged GDP becomes 0 in Model 4 after total factor productivity is included.

¹⁰ Even though Model 5 has less RMSPE than Model 3, we believe the conclusion is not convincing without Taiwan in the donor pool. That is why we prefer Model 3 over Model 5. The results do not change qualitatively if weights from Model 5 are used to construct synthetic Hong Kong.

¹¹ Synthetic Hong Kong fails to generate a good match with actual Hong Kong in terms of the inflation rate (because Hong Kong had unusually high inflation), trade openness (because of Japan’s low value), and stock market capitalization (due to the extreme value of Hong Kong). This is expected because SCM is not designed to mimic extreme values.

¹² SCM does not allow for extrapolating and imposes bounds on weights, see the restriction of $0\le w_j\le 1$ in (2).

¹³ The DID estimator is computationally robust since it minimizes an objective function much less nonlinear than SCM.

¹⁴ We see narrowing gap in two GDPs after 2003, mainly because of the positive shocks of Closer Economic Partnership Arrangement (CEPA) and China’s joining WTO.

¹⁵ The pre-1997 average GDP is 2522, 7082, 2321 and 5211 for Indonesia, Malaysia, Philippines, and Thailand, respectively. The pre-1997 average GDP of Hong Kong is 24,877.