Does productivity growth lower inflation in Korea?

Abstract

We investigate Granger causality between productivity growth and inflation in Korea using quarterly data for the period 1985Q1–2002Q4. Our results indicate unidirectional Granger causality from productivity growth to inflation. In light of such causality, we estimate the effect of productivity and other variables on productivity. According to our regression results, a 1% increase in labour and Total Factor Productivity (TFP) reduces Consumer Price Index (CPI) inflation by 0.07–0.08% and 0.37–0.44%, respectively. Our results also suggest that the productivity-inflation nexus became stronger in Korea since the Asian financial crisis, and that this was largely due to structural reform and technological progress.

Keywords:

• productivity
• inflation
• Granger causality
• Korea
• New Economy

JEL Classification::

• O47
• E32
• C32

Notes

¹ See, for example, Jarett and Selody (1982).

² For example, Clark (1982) and Ram (1984) find a significant negative effect of inflation on productivity for the US, while Saunders and Biswas (1990) find such an effect for the UK.

³ These studies explored the relationship between inflation and productivity in the context of the disinflation of the 1980s. They also benefited from improvements in econometric methodology, in particular paying more attention to the issue of stationarity as well as the possibility that the business cycle may simultaneously affect both inflation and productivity. Examples include Cameron et al. (1996), Freeman and Yerger (1998, 2000), and Tsionas (2003). Nevertheless, some studies such as Smyth (1995) and Bitros and Panas (2001), continue to find a significant negative relationship.

⁴ More formally, Ball and Moffitt (2001) attribute the negative impact of productivity growth on inflation to the lag between productivity growth and workers’ wage demands. Dew-Becker and Gordon (2005) and Kiley (2003) find some empirical evidence that higher productivity growth leads to lower inflation.

⁵ The Augmented Dickey–Fuller (ADF; Dickey and Fuller, 1981), Phillips–Perron (PP; Phillips and Perron, 1988) and Kwiatkowski–Phillips–Schmidt–Shin (KPSS; Kwiatkowski et al., 1992) results are available from the authors upon request.

⁶ The test results are available from the authors upon request. They indicate that a restricted constant, which allows a nonzero drift in the unit-root process, should be included in the multivariate system of equations. The lag value of the VECM is set equal to two. The null hypothesis of r = 0 is rejected at the 1% level but the null hypothesis of I ≤ 1 cannot be rejected. Consequently, the estimated likelihood-ratio tests indicate the presence of one co-integration vector and a long-run relationship in the underlying data-generating process of the time-series.

⁷ The parameter estimates are available from the authors upon request.

⁸ Relative prices in food and oil prices (FEP) and public services prices (PSP) are calculated by dividing the price indexes of these products by the CPI index derived without these products and then multiplying by 100.

⁹ Gordon's model also included a dummy to represent a period during which the Nixon administration controlled prices.

¹⁰ Brayton et al. (1999), Crary (2000), and Macklem and Yetman (2003) also use this restriction. Under the assumption of no demand and supply shocks in the long run, changes in FEP and PSP are averaged to be zero during the sampling period.

¹¹ We chose a smoothing parameter of 100 in estimating potential GDP with the HP filter.

¹² Table 2 reports only the results for CPI inflation to save space, but the results for DEF and PCE inflations were similar.

¹³ The coefficient estimates for this state–space model are available from the authors upon request.

¹⁴ Following Equation 2, we multiplied capital's share of income – – by (K/L) growth when decomposing LPD growth into TFP and (K/L) growth.