South African Reserve Bank working paper finds factor model combination of survey forecasts improves inflation forecasting accuracy

The South African Reserve Bank published a Working Paper assessing which measures of inflation expectations best forecast South African headline CPI inflation and whether combining forecasts beats relying on any single source. Using survey and financial-market data, it finds that a factor-model-based linear combination of forecasts can improve inflation forecast accuracy relative to common aggregation approaches, including the simple averaging embedded in the Bureau for Economic Research (BER) aggregate measure used in the SARB’s Quarterly Projection Model. The paper evaluates inflation expectations from three quarterly BER surveys of businesses, financial analysts and trade unions, alongside market-implied expectations from Bloomberg, across current-year, one-year-ahead, two-year-ahead and five-year-ahead horizons (with the five-year series available from 2011Q3). Forecast performance comparisons using mean absolute scaled errors and root mean squared errors generally rank the factor model best at shorter horizons, with financial analysts often next best, while longer-horizon results show closer performance across measures and indicate that trade unions, businesses and market-based indicators can add information depending on the state of the economy. The Working Paper is presented as preliminary, externally refereed research intended to elicit comment and stimulate debate, rather than a policy or operational change announcement.