ABSTRACT
Underlying, or core, inflation is likely the most important variable for monetary policy. It is considered to be the optimal nominal anchor as it is stable, excludes relative price shocks, and reflects underlying trends in the behaviour of price-setters and demand conditions in the economy. Despite its importance, there is sparse literature on estimating and forecasting core inflation in South Africa, with the focus still on measuring it. This paper emphasizes predicting core inflation from time-varying parameter vector autoregressive models (TVP-VARs), factor-augmented VARs (FAVAR), and structural break models using quarterly data from 1981Q1 to 2013Q4. We use mean squared forecast errors (MSFE) and predictive likelihoods to evaluate the forecasts. In general, we find that (i) time-varying parameter models consistently outperform constant coefficient models (ii) small TVP-VARs outperform all other models; (iii) models with heteroscedastic errors do better than models with homoscedastic errors; and (iv) allowing for structural breaks does not improve the predictability of core inflation. Overall, our results imply that additional information on the growth rate of the economy and the interest rate is sufficient to forecast core inflation accurately, but the relationship between these three variables needs to be modelled in a time-varying fashion.
Disclosure statement
No potential conflict of interest was reported by the authors.
Notes
1 Monetary policy is subject to a new debate on policy frameworks, including nominal income targeting and price-level targeting, following the global financial crisis (see, for example, Woodford Citation2014).
2 Only Norway operationally targets core inflation, while its mandate is in terms of headline inflation. There are a number of reasons why headline inflation has become the variable of choice for central banks, including that communication with the public is thought to be easier; wide public understanding; and that people care about a cost-of-living index, the basket of goods they actually consume, rather than core inflation.
3 Estimating involves using dynamic model selection to choose a value of at each point in time. For more details, see Koop and Korobilis (Citation2013, 9).
4 Unlike the normal Minnesota prior, which has two hyperparameters for own lags and other lags, we used one shrinkage parameter to simplify computation.
5 We employed numerous other methods, including Bai and Ng (Citation2002) and Onatski (Citation2010) to ensure that we were getting the correct number of factors.
6 In factor analysis, usually M ≥ 100.. See Forni et al. (Citation2009) as an example of convergence problems.
7 However, we also conducted our analysis till twelve-quarters-ahead. Our basic results in terms of the ranking of the models based on their forecast performances, continued to remain unchanged when compared till eight-quarters-ahead. Complete details of these results are available upon request from the authors.
8 We also included models with only time-varying intercept terms as an alternative TVP strategy. These models do not outperform models where all parameters are allowed to vary but do better than the constant coefficient models.
9 We also estimated the small-scale TVP-VAR models, without the approximation based on the forgetting factors. However, this version of the small-scale TVP-VAR model was outperformed by all the other small-scale TVP-VAR models reported in the paper both in terms of point and density forecasts. Complete details of these results are available upon request from the authors.