A research paper released on Monday has exposed Bank of Botswana(BoB)�s inflation forecast model as not having far reaching predictions and its accuracy only relevant for a short time.

The research titled �Evaluation of the Performance of the Bank of Botswana�s Inflation Forecasting Model� was conducted by central bank�s Research & Financial Stability Department
It was authored by Deputy Director, Innocent Molalapata and economists Lesego Molefhe, Lizzy Sediakgotla and Daniel Balondi.

According to the quartet, to support its forward-looking monetary policy framework, BoB produces inflation forecasts using a quarterly projection model, known as the �Core-Model.� The inflation forecast provides guidance on the appropriate monetary policy interventions necessary to achieve and maintain the Bank�s primary objective of price stability, they wrote.

The economists further said it is critical to have a model that produces inflation forecasts that do not deviate significantly from the actual data, without justifiable accounting, as it can misinform policymaking.

�According to the results, on average, the Bank�s Model has a good forecasting accuracy in the short term, however the model has a low predictive power of medium-term inflation movements, reflecting higher uncertainty associated with economic events in the distant future. Regarding prediction bias, the Model tends to under-predict inflation, which is a result of changing economic conditions domestically and externally,� said the quartet.

The four brains merged to produce a paper titled �Evaluation of the Performance of the Bank of Botswana�s Inflation Forecasting Model� which sought to evaluate the performance of the central bank�s inflation forecasting model over a 10-year period, from December 2008, a period of the Great Recession, to December 2018.

While they were putting the central bank policy practice under the knife, the objectives of the study are to assess the Model�s forecast bias by employing the Mean Forecast Error and determination of the size of the forecast error using the Mean Absolute Forecast Error (MAFE).

The researchers explained that BoB uses a forward-looking monetary policy framework that is guided by a forecasting and policy analysis system (FPAS). According to them, the FPAS is the Bank�s main policy analysis and forecasting framework while the Core-Model, which forms part of the FPAS, is used to produce medium-term forecasts for macroeconomic variables, such as inflation and the output gap.

�The CoreModel is based on Botswana�s monetary policy transmission mechanism. It captures the essential macroeconomic relationships, primarily the impact of monetary policy on output and inflation. The model provides a comprehensive view on future economic developments and possible policy actions necessary to achieve the Bank�s inflation objective.

It also explores possible alternative scenarios to the baseline forecast. Given the Bank�s forward looking monetary policy framework, forecasts from the Core-Model are used as input into deliberations of monetary policy formulation and direction by the Monetary Policy Committee (MPC),� further explained the economists.

Going sharp in their analysis knife, the economists said the Core-Model is calibrated in accordance with the behaviour of key relationships in the monetary policy transmission process, therefore the model output is expected to be a close match to actual data, assuming an adequately functioning transmission mechanism, ceteris paribus.

However, according to the four experts, it is not easy to have a model that is able to produce forecasts that accurately fit actual or realised data due to, among others, future unanticipated changes in the structure of the economy; model misspecification; historical data measurement errors; inaccurate parameter estimation or calibration; volatility in the analysed variables; inappropriate policy design and decision; as well as domestic and external shocks in the economy.

As the quartet�s thesis was reminiscent of the monetary policy practice dating back to the Great Recession, they said some of the shocks are generated by financial imbalances, as was evident during the 2008/9 global financial crisis, as well as lack of congruence of monetary, financial stability and fiscal policies towards achieving macroeconomic stability.

�Meanwhile, a model that produces an inflation forecast path that deviates significantly from the actual data, without justifiable accounting, is not desirable as it can misinform policy making. In practice, it is essential for forecasts to be approximate to observe values as that provides reliable information that can be useful in guiding policy formulation and implementation.

In this context, it is important to assess the forecasting ability of the Core-Model regularly in order to identify and attend to discrepancies, if any, and where necessary, enhance its forecasting ability,� said the four researchers.

When delving into the �Forecast Errors for Headline Inflation� the four brains said statistically, on average, with the exception of the first two periods, BoB has consistently under predicted both short-term and medium-term headline inflation for the period December 2008 to December 2018.

The economists further elaborated that the bias is more pronounced in the medium-term forecasts, precisely from the 5th quarter to the 8th quarter. Under-prediction of short-term inflation (for four quarters) is, however, relatively small at around 0.05 percentage points on average, while medium term inflation forecasts have on average been 0.69 percentage points lower than actual inflation, they said.

Last week, after maintaining the Bank Rate at 3.75 percent, BoB changed its inflation forecast of October which projected inflation to revert back to the objective range in the third quarter of 2021. The fresh inflation forecast which was done last week during an MPC meeting is that inflation will revert to the objective range of 3-6 percent sooner, in the second quarter of next year.

According to Head of Research & Financial Stability Department, Tshokologo Kganetsano, the quick reversion of inflation is due to the increase in transport fares, adjustment in fuel prices and postal office tariffs. However RMB Botswana researchers who use a model different than that of BoB, said the effect of Kganetsano�s stated price movements is unlikely to see inflation average beyond 3 percent in 2021. They expect price growth to register an average of 1.9 percent in 2020 and 2.5 percent in 2021.

The demand side is expected to act as a drag on inflation in 2021 as the bulk of Botswana�s work force will be faced with unemployment challenges as businesses continue to reel from the effects of the pandemic, said the researchers. The RMB Botswana researchers said they expect the effect of upside pressures on the headline figure to remain limited, coupled with an uncertain economic outlook as a result of the disruption caused by covid-19.

RMB Botswana explained that the model they use is a �bottom-up approach� which is based on the baskets and data from Statistics Botswana. �We track each basket and stress it according to any recent or upcoming developments-so in the case of transport, we observe international prices and the likelihood of BERA will adjust local prices. If the likelihood is high enough, then we factor into the model.

Basically we do this for all the baskets and since they are weighing according to what Stats Bots provides, we are then able to provide an annual forecast. This is then done according to different scenarios-in the case some of our expectations don�t happen or we have surprise announcements. As the year goes along, we keep amending to incorporate any changes. For longer term forecasts, beyond two years, we use a regression and make adjustments to the outcome based on our research houses expectations as well as regional outlooks,� explained a researcher at RMB Botswana after their inflation forecast.

The other side of the coin

The four researchers who studied Botswana�s Performance of BoB�s inflation forecasting model have reached to a conclusion that, according to the results from a rolling four-year sample of forecast errors using MAFE, the forecasting performance of the model used by BoB has improved with time.

According to the research quartet, the innovations in the model structure, constant engagements between the Modelling Team and Monetary Policy Technical Committee members and Sector Specialists, as well as capacity building programmes on modelling and forecasting which the Bank continues to invest heavily in.

In the conclusion paragraph, the four economists said BoB�s forecasting framework has a high level of precision, particularly in the short term. However precision of the inflation forecasts in the medium term is limited by increased uncertainty associated with the longer horizon, they further stated in the report.

The researchers said BoB is not addressing the issue of the forecasts being less far reaching, but this is partly addressed by updates of forecasts for successive MPC meetings and Monetary Policy Reports (MPRs). They said continual efforts are being made to improve the inflation forecast performance in both the short- and medium-term.

�As part of ongoing improvements to the forecasting ability, since the last half of 2017, an assumption on the rate of crawl for the ensuing year is made, compared to the earlier approach in which the rate of crawl was assumed to be constant,� said the researchers.

The four researcher further advised that going forward it is crucial for BoB to continue to invest in capacity improvements to facilitate updating of the model (and its calibration) to reflect the evolving structure of the economy through infusion of relevant methods and skills inputs.

Another advice was in recognition of the influence of administered prices on inflation, the four saying it is important for BoB to broaden relations with all stakeholders in order to be informed of impending changes in administrative prices, to inform both forecasts and policy posture.

�Engagement of the MPC by the forecasting team in the formulation of initial conditions, external assumptions and alternative scenarios must be strengthened further and should include a discussion of the policy rate.

Finally, where possible, BoB should strive to address impediments to monetary policy transmission, particularly the credit channel, to enhance the effectiveness of the transmission of monetary policy. High precision and reliable forecasts are important for the transparency and predictability of monetary policy, which enhances policy credibility,� said the researchers.