BSP’S inflation forecasting more accurate – report

The ability of the Bangko Sentral ng Pilipinas (BSP) to forecast inflation is more than adequate, and its forecasting performance in the last 12 years has been generally “accurate, unbiased and efficient”, even outperforming forecasts from the private sector, according to an evaluation report.

Based on its latest Monetary Policy Report (MPR) on BSP'S forecasting actions for the period 2010 to 2022, the central bank considers both its monthly and annual inflation forecasting performance as better than most, including that of the International Monetary Fund (IMF) and private sector economists.

With that, the BSP said that they have "performed well and will continue to do so" after systematically assessing its inflation performance in terms of accuracy, unbiasedness, efficiency, and benchmarking with other forecasts,

“The tests indicate that the inflation forecasts produced by the BSP are generally accurate, unbiased, and efficient for the period 2010 to 2022. However, testing for a smaller sample size showed that the BSP’S inflation forecasts were systematically lower than actual outturns reflecting the significant uncertainty in global oil and food prices as well as the smaller impact of non-monetary measures that were assumed in the baseline forecasts. Nonetheless, the BSP’S forecasts performed better compared to the private sector and IMF across the different forecast horizons,” said the BSP.

For the month-ahead inflation forecast for example, from 2010 to 2022 the BSP correctly predicted that actual inflation will settle within the forecast range 143 times out of 155 months, or a batting average of 92.3 percent.

The MPR also noted that 11 out of 11 monthahead inflation forecasts in 2022 were within the forecast range announced by the BSP, which is usually the last working day of a given month.

Meanwhile, forecasting errors or the mean absolute error (MAE) from the month-ahead forecasts in 2022 were lower compared to the historical average.

Last year, the BSP said the MAE of the median and average inflation forecasts for the BSP was 0.17 percentage point (ppt) compared to Bloomberg’s 0.20 ppt. The BSP’S month-ahead forecast errors had a 0.20 ppt historical average since 2010. “The BSP’S monthahead inflation forecasts (in 2022) outperformed the private sector’s consensus estimates,” it said. The BSP also noted that the MAE of its one-month ahead inflation forecasts “has consistently declined since 2010, although it has risen to levels above the historical average in the years 2018, 2020, and 2021.” As for the BSP’S annual inflation forecasts, in 2022 the forecasts were lower compared to the actual average. The annual inflation estimates are released eight times to the public after each Monetary Board policy meeting in a given year. The BSP said the 2018-based annual inflation forecasts for 2022 ranged from 3.7 percent to 5.8 percent. By the time the BSP had one of its policy meetings on March 24, 2022, the forecasts have risen above the inflation target band of two percent to four percent. “The continued upward revisions in the annual forecast for 2022 can be attributed to the adjustments in the assumptions for crude oil prices and global non-oil prices due to global geopolitical factors as well as and the peso depreciation following the monetary policy tightening of the US Federal Reserve and other major central banks. Domestic factors such as the impact of supply constraints on food prices, higher electricity rates, and transport fare adjustments also contributed to the upward revisions in the forecasts for 2022,” said the BSP.

In 2022, the BSP, IMF and the private sector’s six-month, one-year, and two-year ahead forecasts were all lower than the average actual inflation of 5.8 percent.

The BSP said however that its estimates were more accurate. “When evaluated over a longer period, the MAE of the BSP’S full-year inflation forecasts across the different forecast horizons are consistently lower compared to the private sector and IMF,” it said.

To improve its forecasting performance, the BSP since last year has intensified efforts to acquire real-time information for it to identify emerging risks across different sectors. They developed and continue to develop new tools such as machine learning methods to

analyze non-traditional data such as internet searches, newspaper articles, and social media posts, it said.

Meanwhile, to further enhance not only its forecasting models but also surveillance, simulations and policy analysis, the BSP has ongoing consultations with the IMF’S Institute for Capacity Development (IMF-ICD) and the Japan International Cooperation Agency.

At the moment, the BSP’S Department of Economic Research is transitioning to the Policy Analysis Model for the Philippines (PAMPH), in cooperation with the IMF-ICD.

PAMPH will allow the BSP to present a more detailed outlook of inflation and growth, as an enhanced workhorse model. The new model will be extensively tested before it is used publicly.

Currently, the BSP forecasts inflation using the Multi-equation Model (MEM) but an improved PAMPH will complement MEM.

Refinements to the PAMPH include detailed blocks of aggregate demand, external sector, and the fiscal sector. “Future extensions of the model will cover macroprudential policies and capital flow management measures, along with a credit block and labor sector,” said the BSP.

The PAMPH which is based on the Forecasting and Policy Analysis System developed by the IMF, has redefined the structure of interest rates based on the policy rate and market rate to ensure a more representative transmission of monetary policy to the economy. It also re-defined the real exchange rate relative to the US dollar as the trade-weighted real effective exchange rate, and re-calibration of parameters.