​Full press release + figure

·         The research presents a new model developed at the Bank of Israel for “Nowcasting” quarterly GDP growth—forecasting growth that has occurred but for which an estimate has not yet been published. The model is expected to reduce the forecasting error by approximately 10 percent compared to the existing models at the Bank.

·         The model also generates monthly estimates of the state of activity, which are consistent with the quarterly forecast. These estimates described well the sharp fluctuations in activity during the waves of COVID-19 morbidity. Their output makes it possible to receive data earlier relative to existing indicators such as the Composite State of the Economy Index. This is extremely important in assessing the state of the economy in real time.

·         According to the model, in 2021 the economy returned to high growth in October–November with the end of the fourth wave of infection, but in December a renewed slowdown in activity began.


Monetary policy is aimed at stabilizing economic activity—first and foremost in the area of price stability—but also with regard to the scope of economic activity, which is measured mainly by GDP. To make decisions, economic policy makers are required to estimate the current state of activity, as official GDP data are only published with quarterly frequency and with a lag of about a month and a half. To that end, Nowcasting models are implemented at the Bank of Israel, which are intended to forecast data reflecting activity that has already occurred but for which estimates regarding it have not yet been published.


In a joint paper by Tim Ginker of the Information and Statistics Department and Dr. Tanya Suhoy of the Research Department, a new model was developed that is intended to provide a clearer picture of the state of activity in real time. This is as part of the continued effort of examining new sources of data, research and economic developments in Israel and abroad, and the scrubbing and updating of the existing models in use at the Bank of Israel. The model’s outputs include: (1) an estimate of the quarterly GDP growth rate, which has not yet been published; the estimate is generated already in the middle of the quarter, approximately 3 months before the publication of the first estimate by the Central Bureau of Statistics, and is updated weekly based on new information. (2) The monthly GDP series, alongside the real-time estimate of it.


The model is based on approximately 30 data series with monthly frequency, such as foreign trade data (imports and exports) for Israel, revenue data based on VAT, tax revenue data and employment data. As these indicators as well are published with a lag that ranges from 1 to 2 months, they are brought forward to real time via rapid indicators, some of them long-time, such as the Business Tendency Survey by the Central Bureau of Statistics, and some of which are new and that were developed during the COVID-19 period, such as daily credit card expenditure.


The hybrid statistical model, which relies both on the research literature and on specific development as part of this research, filters, out of all the information inherent in these 30 series, estimates for GDP at a monthly frequency from which the quarterly Nowcast is also calculated.


Similar outputs were generated in the past within the framework of various models operated at the Bank of Israel.  Two research papers in 2011 developed Nowcast models, and since the 1990s the Bank of Israel has published an estimate of the state of monthly economic activity as part of the “Composite State of the Economy Index”. One of the innovations of the current model is that it is consistent with the two outputs—the monthly estimate and the quarterly Nowcast are generated by the same system. Examinations that were carried out as part of this research indicate that the new model improves the forecasting ability for quarterly GDP by about 10 percent compared with existing Nowcast models, in addition to the improvement in speed of the data being issued. To illustrate, the Composite State of the Economy Index is calculated approximately 3 weeks before the end of the month, while the new model can generate parallel estimates immediately upon month’s end.


An examination of the model’s outputs, as of January 2022, points to several interesting findings (Figure 1). In an overall view of the coronavirus crisis, it can be seen that the first lockdown (March–April 2020), which arrived without advance preparation, adversely impacted economic activity with a magnitude that was 3 times greater than the 2 lockdowns that followed. The moderation of the negative impact can be attributed to adjustments made by businesses that enabled them to work better under lockdown conditions. However, throughout the entire period the level of activity did not return to that which characterized it prior to the outbreak of COVID-19. Looking at the end of 2021 indicates that during the fourth wave of morbidity, there was a halt in the economic recovery that began with the end of the previous lockdown (that took place during the fourth wave). However, with the wearing down of the fourth wave, and the end of the Jewish holiday season, which was very condensed in September this year, there was a marked increase in activity of about 1 percent in each of October and November 2021. In contrast, real time estimates indicate that in December there was some deterioration in economic activity, perhaps against the background of the start of the fifth wave of morbidity. According to these estimates, GDP growth in the fourth quarter of 2021 is expected to total 1.2 percent, which is 5 percent in annual terms.


The ongoing functioning of the model is expected to be examined at the Bank of Israel in the coming months, after which the policy regarding the publication of its results will be decided.