Annex: Technical Appendix
A Vector Error Correction Model (VECM), which can capture both long-run and short-run relationships for an empirical analysis, is used in the paper. The data set for the estimation is on a quarterly basis, of which CPI data are seasonally adjusted by the Census X-12 while quarterly government expenditure data are calculated by interpolating the annual data in a quadratic-match average way. Independent variables are foreign exchange rate (NEER), broad money, central government expenditure, oil prices, and inflation in Papua New Guinea’s major trading partners (import value weighted average of CPI in Australia, Japan, New Zealand, the United States, and developing countries). Oil prices and inflation in Papua New Guinea’s major trading partners are assumed to be exogenous variables given Papua New Guinea is a small and open economy (price taker).
Douglas K. Pearce, Sara A. Reiter, 1985, “Regression Strategies When Multicollinearity Is a Problem: A Methodological Note,” Journal of Accounting Research, Vol. 23, No. 1.
Neil R. Ericsson, Eilev S. Jansen, Neva A. Kerbeshian, and Ragnar Nymoen, 1998, “Interpreting a Monetary Conditions Index in Economic Policy,” BIS Conference Papers, Volume 6, pp. 237 –254, Bank for International Settlements, Basle, Switzerland.
Sampson Thomas, Jeffrey Yabom, Williamina Nindim, and Jacob Marambini, 2006, “Exchange rate pass-through in Papua New Guinea,” Pacific Economic Bulletin, Volume 21, No 1, Asia Pacific Press.
Prepared by Aiko Mineshima. This chapter is based on a presentation delivered at the Bank of Papua New Guinea in Port Moresby in November 2007 and benefits from comments received at that time.
The quality of CPI data for Papua New Guinea is poor: for example, the weights are outdated (based on a 1975–76 Household Income and Expenditure Survey), and prices on dwelling rentals, which compose about 3.9 percent of the headline CPI basket, have not been collected since 1991 and are therefore held constant.
Tradable goods include: food; drinks, tobacco, and betel nut; clothing and footwear; household equipment; and motor vehicles.
Nontradable goods include: rents, council charges, and fuel/power; transport and communication; and miscellaneous.
The monetary condition index (MCI) is a weighted average of changes in the effective exchange rate and interest rates relative to a base period (in this case, end-1996). Reflecting the relative influence on monetary conditions, the weights for Papua New Guinea are 75 percent for the NEER and 25 percent for the interest rate. As Papua New Guinea is a small, open economy with a shallow financial market, the impact of exchange rate movements on monetary conditions is more significant than in more developed economies.
Productivity (GDP gap or real GDP per capita) could be an explanatory variable of inflation. However, both the GDP gap and real GDP per capita do not work well for Papua New Guinea; the sign for the GDP gap is opposite to the our expectation, while NEER becomes statistically insignificant when real GDP per capita is included in the model. These results could be related to the poor quality of national accounts data. Nonetheless, given indications that productivity has not improved in recent years, inflationary pressure could be further increased if other conditions are unchanged.
Thomas and al. (2006) estimated the exchange rate pass-through to underlying inflation by using OLS. The result shows the pass-through effect is approximately 50–60 percent.
Including both broad money and government expenditures in a model causes multicollearity problem (the correlation coefficient of broad money growth and government expenditures is about 0.9). Other studies have shown that dropping one of correlated variables is the best way to solve this problem; Pearce and Reiter (1985) discusses the well-known two-stage strategy.
Private credit, related to broad money, could be an explanatory variable of inflation. However, the econometric model becomes divergent if private credit is used instead of broad money.
“Miscellaneous” includes medical and health care, entertainment, cultural goods and services, and other goods.
Peer countries are selected based on the size of impact of changes in commodity exports on current account balances, foreign exchange regimes, and income levels.
Looking at levels rather than growth rates, government expenditure and broad money as a share of nominal GDP have been higher for PNG than for its peer countries: the average of government expenditure as a share of nominal GDP for 1996-2006 is 31 percent for PNG and 23 percent for its peers, while the average of broad money as a share of nominal GDP for 1996–2006 is 34 percent for PNG and 21 percent for its peers.