Has inflation targeting improved social welfare in practice?

Abstract

This article verifies empirically the proposition that inflation targeting can improve social welfare. To analyse this, we construct a cashless small open-economy model conforming to the New Keynesian Open-Economy Macroeconomics and use data in the United Kingdom. We estimate the feedback rule, before estimating the deep parameters using GMM. Using the estimated structural model and identified shock processes, we simulate the imaginary paths of the nominal interest rate, producer price index inflation rate and output if the central bank has not adopted an inflation targeting policy. This simulation shows that the variances of output and the inflation rate under a noninflation targeting regime are larger than those under the actual inflation targeting regime. We also calculate social loss, as defined by the second-order approximated utility function under each regime. The results indicate that social loss brought about by an inflation targeting regime is less than that associated with a noninflation targeting regime.

Acknowledgements

I have benefited from discussions with and comments from Eiji Ogawa, Hayato Nakata, Toshiki Jinushi, Tsutomu Watanabe and Yosuke Takeda. The views expressed and any errors are my responsibility.

Notes

¹ Until 1995, its inflation target, using the same measure, was in a range of 1–4%.

² The PRIX is similar to a consumer price index. Okano (2005) shows that consumer price index inflation is an appropriate target to dissolve inflation-output trade-offs, while Benigno and Benigno (2004) and Gali and Monacelli (2005) assert that a more suitable target is producer price index inflation.

³ Where , , , , ω is the openness in the SOE, α is price stickiness and δ is a subjective discount factor.

⁴ Following Nelson (2001), we choose the TB rate.

⁵ The TB rate is divided by four to obtain quarterly denominations. Real GDP is converted to logarithms and detrended. The PPI is converted to logarithms.

⁶ Inflation targeting was introduced in October 1992.

⁷ SEs are 0.4433 and 0.1392, respectively.

⁸ SEs are 0.8613 and 0.3349, respectively.

⁹ F-value is 1.2114 lower than the 5% significance level, 2.3205. Also, the feedback rule has satisfied the ‘Taylor Principle’ after inflation targeting was introduced.

¹⁰ Nelson (2001), using not the PPI inflation rate but the PRIX inflation rate, points out that the TB rate in the United Kingdom has reacted more strongly to the PRIX inflation rate, which was adopted by the Bank of England as the targeted inflation rate after the introduction of inflation targeting.

¹¹ We choose the GMM because the assumption of the GMM estimator is more mitigating than the minimum distance estimator adopted by Boivin and Giannoni (2003).

¹² The subjective discount factor δ is constrained to 0.99 following Rotemberg and Woodford (1997).

¹³ Steady state values are acquired by a reduced form of (4). See Rotemberg and Woodford (1997).

¹⁴ SEs are 0.1282 and 0.0316, respectively.

¹⁵ Gali and Monacelli (2005) interpret ω as openness of the small open economy and calibrate by using the share of imported goods on GDP in Canada. Our estimation is pertinent because the share of imported goods on GDP in the United Kingdom in this period is 0.2640.

¹⁶ This implies that the slope of the aggregate supply curve, namely, β_λ is 0.2485. Nelson and Nikolov (2002) estimated the New Keynesian Phillps Curve and reported that its slope is 0.222 from 1979 through to 1990 and it is 0.349 from 1961 through to 2000, respectively.

¹⁷ To identify the process of ξ _t , estimating (4) and calculating (6) generates the same result.

¹⁸ See Gali and Monacelli (2005).

¹⁹ ϑ cannot be estimated because it is eliminated at the process of log-linearization, therefore we set ϑ = 7.88, as assumed in Rotemberg and Woodford (1997).