MONETARY POLICY REPORT | SPRING 2022 (appendix 1)
(authors: Tatiana Keseliová, Karel Musil, Tomáš Šestořád, Stanislav Tvrz, Jan Žáček)
Assessing the fulfilment of past forecasts is an integral part of the CNB’s forecasting and analytical architecture. These quarterly analyses are an important element of forecasting practice and a prerequisite for constantly enhancing the projections. They provide feedback on the functioning of the core projection model, which is the basic and unifying element used in preparing CNB macroeconomic forecasts. The conclusions of these analyses are used to verify the model’s current settings and as a source of inspiration for its further development.[1] Analysing the fulfilment of past forecasts has been a regular part of the CNB’s internal processes since it started using structural models for forecasting. This is the first Monetary Policy Report[2] to present an extended assessment[3] of the fulfilment of past forecasts. The publication of the assessment further increases the CNB’s already high degree of monetary policy transparency[4] and demonstrates the accountability the CNB as an institution feels towards the public.
The CNB’s macroeconomic forecast serves as an important guide for the Bank Board when setting interest rates. It is based on the core forecasting model (the g3+ model since 2019). The baseline scenario of the forecast describes the most probable future evolution of the domestic economy and the domestic interest rate path consistent with this which ensures the achievement of the CNB’s inflation target at the monetary policy horizon.
In this appendix, we first assess the fulfilment of the forecasts[5] prepared in 2020, starting with a comparison of their then assumptions (the exogenous factors of the forecast) and the observed developments. We then compare the forecasted paths of the main domestic variables with the now observed data. In the last part, we provide a hypothetical model simulation. This shows what the forecast in Inflation Report IV/2020 would have approximately looked like if what was subsequently actually observed, i.e. the future evolution of all the assumptions entering the forecast, had been known at the time of its preparation. Generally speaking, this simulation illustrates the predictive ability of the g3+ core model.
Assessment of the fulfilment of the 2020 forecasts – assumptions
The main assumption of the CNB’s macroeconomic forecast for the domestic economy is the outlook for the foreign environment.[6] With the exception of the pre-pandemic forecast in IR I/2020, the assumptions about a decline in economic activity in the effective euro area in the first half of 2020 broadly materialised (see Chart 1). The immediate impacts of the pandemic and the related lockdowns were therefore correctly predicted. The return of economic life abroad to normal was in fact rather faster and more volatile than expected in the outlooks at the time. From the second half of 2020 onwards, foreign economic activity was largely higher than forecasted, due partly to the fiscal support measures adopted in most EU countries. The impact of the Covid waves, in the form of a significantly negative output gap, which captures the demand side of the economy, was gradually re-evaluated in the CNB’s 2020 forecasts in favour of greater impacts on potential output (i.e. the supply side of the economy) amid an only limited decline of the output gap into negative territory. This was because supply was hit by restrictions imposed on firms and society as a whole during the various waves of the pandemic, later joined by longer-term impacts of the pandemic in the form of disruption and overloading of global value chains (GVCs).
The forecasts therefore initially expected stronger and longer-lasting anti-inflationary pressures in the effective euro area due to a fall in demand related to the outbreak of the pandemic. However, demand fell only to a limited extent and shifted from services to miscellaneous manufactured articles. Conversely, the overloading and disruption of GVCs caused unexpectedly strong inflation pressures at the level of foreign industrial producer prices. The faster global economic recovery was also reflected in a rise in the Brent crude oil price, which, following a sharp decline in the first half of 2020, rose significantly higher than originally assumed. At the start of 2022, Russia’s invasion of Ukraine – which could not, of course, have been foreseen by the 2020 forecasts – also affected the already high price of oil. The forecasts under assessment also did not contain an assumption of an unprecedented surge in foreign producer price inflation as occurred at the end of 2021, driven largely by an extreme jump in natural gas and electricity prices. The outlook assumed for the 3M EURIBOR was roughly confirmed. In its unconventional component, the ECB’s monetary policy was only slightly looser than in the assumptions of the forecasts under assessment.
The observed growth in administered prices in the domestic economy during 2021 was well below the assumptions of the forecasts under assessment (see Chart 1), due mainly to an unexpected decline in electricity and gas prices for households. The deviation from the previous forecasts widened further at the end of 2021 due to a temporary waiver of VAT on electricity and gas. At the start of 2022, the previous unexpectedly strong growth in energy prices on international exchanges was reflected in administered prices, hence their growth far exceeded that assumed in the IR IV/2020 forecast. Growth in nominal government consumption and the fiscal impulse, another domestic assumption, was higher on average than in the forecasts under assessment, due to additional pandemic-related government spending, mainly on health care and the emergency services. Government measures in support of the economy were more pronounced and lasted longer than the forecasts had expected.[7]
Chart 1 – Selected forecast assumptions
Assessment of the fulfilment of the 2020 forecasts – main domestic variables
Headline inflation was roughly in line with the forecasts under assessment until early 2021. The exception was the IR II/2020 forecast (see Chart 2), as the anti-inflationary effects assumed in this forecast did not materialise to the expected extent and inflation rose into the upper half of the tolerance band around the target. The subsequent surge in inflation in the second half of 2021 and early 2022 in particular came as a considerable surprise. This was caused by unexpectedly rapidly rising price pressures in both the foreign and domestic economies. In a situation where the activities of entire sectors of the economy, including services, had been severely limited or suspended, government fiscal stabilisation measures – in the form of subsidies and compensation for households and firms – fostered an overheating of the labour market and further accumulation of forced savings (withheld demand). This, coupled with a reduction in personal income tax under a government tax package, was reflected in solid consumer appetite and a willingness of households to tolerate increased inflation last year. Headline inflation also gained momentum due to a tight property market, which passed through to core inflation via imputed rent.
Chart 2 – Forecasts of main domestic variables
The 3M PRIBOR market interest rate fell significantly with the outbreak of the pandemic. This was due mainly to a sharp reduction in the CNB’s key interest rates in spring 2020 in anticipation of a decline in domestic demand and subdued inflation. The subsequent increase in market interest rates in the course of 2021 was initially roughly in line with the IR IV/2020 forecast.[8] However, in the second half of last year, the need arose to further respond to the combination of strong domestic and unprecedentedly escalating imported inflation pressures, so interest rates increased significantly more than the forecasts under assessment had foreseen.
The koruna reacted sharply to the various waves of the pandemic and the related measures, thus deviating from the forecasts under assessment. The pre-pandemic IR I/2020 forecast departed furthest from the observed path, as it did not yet incorporate the global impacts of the Covid-19 pandemic. By contrast, the following IR II/2020 forecast was too pessimistic and predicted a further depreciation of the koruna. In reality, however, after weakening in connection with the onset of the spring and then autumn waves of the pandemic, the koruna appreciated in 2020. In 2021, the exchange rate of the koruna was markedly stronger than forecasted, mainly reflecting an improvement in financial market sentiment owing to the gradual vaccination of a significant proportion of the population in advanced countries. The strengthening of the koruna from the second half of last year onwards was also due initially to the CNB’s communications and subsequently to its rapid raising of domestic interest rates in response to exceptionally strong inflation pressures. The appreciation was slowed temporarily at the end of the summer of 2021 by the onset of another wave of the pandemic.
Domestic economic activity was significantly affected by the pandemic in the last two years. The impacts of the pandemic were fully incorporated into the forecasts from the IR II/2020 forecast onwards. Overall, a faster recovery in external demand in 2020 and a stronger recovery in domestic demand in 2021 led to higher GDP growth relative to the forecasts under assessment. The more pronounced recovery of the domestic economy was due to stronger effects of government fiscal stabilisation measures and the use of the forced savings accumulated during lockdowns. Moreover, nominal wage growth in market sectors in 2021 exceeded the assumptions of the IR III/2020 and IR IV/2020 forecasts, which had expected a more substantial cooling of the labour market due to the pandemic.[9]
Hypothetical factors-known IR IV/2020 forecast
The factors-known simulation is a hypothetical version of the IR IV/2020 macroeconomic forecast which, in addition to current knowledge of the paths of the exogenous variables (the assumptions of the forecast), takes into account the revisions since made to the historical data. The simulation is not a fully fledged forecast, as perfect knowledge of the actual evolution of the forecast assumptions is unrealistic, and additional expert adjustments would probably be made even in this case. Such adjustments are an integral part of every forecast. However, the simulation can still reveal how the core forecasting model reacts to changes in the assumptions. Like the IR IV/2020 forecast, the simulation was created using the g3+ core forecasting model.
We first compare the difference between the original forecast and the hypothetical factors-known simulation. We then assess the deviations of this simulation from the observed historical outcome. Overall, the factors-known simulation is close to the observed outcome, confirming that the g3+ modelling system is appropriately configured.[10]
Comparing the hypothetical factors-known simulation with the authentic IR IV/2020 forecast, headline inflation is slightly above the original forecast until mid-2021 (see Chart 3) and then gradually decreases towards the CNB’s 2% target. The persistence of inflation above the target until mid-2021 is mainly due to the ex-post evaluated inflationary effect of the first wave of the pandemic, whose impacts in the direction of a contractionary phase of the business cycle were more modest than originally assumed, amid faster cost growth. In the factors-known simulation, inflation returns to the target at the start of 2022 due to a rapid rise in interest rates since the end of 2020, because in the hypothetical forecast, monetary policy reacts in a forward-looking manner to the strong inflation pressures from the foreign economy. Actual external developments foster tighter domestic monetary conditions, mainly due to overloading of GVCs and through a record increase in energy prices on international exchanges. However, this foreign inflationary effect is partly offset by the ECB’s still accommodative foreign monetary policy. This exerts latent appreciation pressure on the koruna, which, other things being equal, would have anti-inflationary effects. As a result, in the hypothetical simulation, a further widening domestic interest rate differential vis-à-vis the euro area fosters greater appreciation of the koruna relative to the authentic forecast. The faster pace of appreciation driven by the significantly widening interest rate differential therefore contributes significantly to achieving the inflation target in the hypothetical simulation. The domestic economy initially grows faster on average in the factors-known simulation than in the autumn 2020 forecast. This reflects an initially smaller decline and a subsequently faster recovery in external demand. The effect of the tighter domestic monetary conditions in the simulation relative to the forecast under assessment prevails from the second half of 2021 onwards, so economic growth is conversely subsequently slower.
Comparing the hypothetical factors-known simulation with the historical outcome, observed inflation was below the simulation in late 2020 and early 2021. This is because the simulation does not capture the effects of the then lockdowns in retail and services in the domestic economy, which hindered the realisation of the strong latent consumer demand and led, among other things, to the postponement of the usual January repricing. When the lockdowns ended, households were willing to accept higher prices of goods and services due to their accumulated forced savings. This made it possible for businesses to put up their prices and thus make up for their previous lockdown-related losses by increasing their profit margins. The resulting growth in prices, combined with accommodative monetary policy and an only partial pandemic-related cooling of the labour market, led to significantly higher actual inflation from mid-2021 onwards than in the hypothetical forecast. At that time, the central bank had not yet observed strong inflation pressures from abroad, so real interest rates were below the interest rate path in the factors-known simulation until autumn 2021. Subsequently, however, the observed interest rates even exceeded the rate path in the simulation, because the central bank started to respond forcefully to the rising inflation amid renewed overheating of the domestic economy, the labour market and the property market and accumulating significant foreign inflation pressures. The koruna’s exchange rate initially evolved in line with the factors-known simulation. It later appreciated more slowly due to adverse sentiment on the financial markets associated with the spread of new variants of the coronavirus. After the impacts of the first wave of the pandemic subsided, GDP growth was higher than in the hypothetical simulation, because Czech industry benefited from the aforementioned shift of domestic and foreign consumer demand from services to goods. Subsequently, GDP growth is in line with the hypothetical simulation on average.
Chart 3 – Comparison of the authentic IR IV/2020 forecast and the factors-known simulation (hypothetical forecast) with the observed data
grey area in charts shows IR IV/2020 forecast horizon
[1] The forecast arises on the basis of debates involving many economists and monetary policymakers, whose views are incorporated into the forecast in the form of expert adjustments. The core model serves as a unifying framework ensuring the necessary macroeconomic consistency. However, the model still has to demonstrate a good predictive ability, and regular quality control is desirable.
[2]The CNB’s current main monetary policy publication is the quarterly Monetary Policy Report (MPR), which replaced the previous Inflation Report (IR) in 2021.
[3] This type of analysis – albeit narrower (focusing on one forecast only) – was a regular part of the Inflation Report. However, the analysis was focused solely on comparing the forecast with the observed data. It has now been considerably extended to include a factors-known simulation, and, from a broader perspective, to assess multiple forecasts at the same time.
[4] In 2022, the CNB was named the winner of the prestigious Central Banking Transparency Award for its open and easy-to-understand communication with stakeholders and for its relentless commitment to further increasing its transparency. This is the second time the CNB has won the award. It did so for the first time in 2015.
[5] The choice of forecasts included in this analysis is determined by the availability of observations, which must cover the monetary policy horizons of the forecasts under assessment. The last forecast which can be included in the analysis is the IR IV/2020 one.
[6] The specific indicators considered are industrial producer prices in the effective euro area (broken down into their core and energy components), foreign economic activity (the GDP trend and the output gap in the effective euro area), the USD/EUR cross rate, the Brent crude oil price and the 3M EURIBOR interest rate and its shadow component capturing the ECB’s unconventional monetary policy measures. The domestic assumptions include the outlook for administered prices and nominal government consumption along with its deflator and the fiscal impulse.
[7] On the revenue side, the “super-gross wage” was abolished (on 1 January 2021) as part of a tax package. As the personal income tax rate was kept unchanged, this led de facto to a decline in labour taxation.
[8] This caused uproar in the Czech and international analytical communities at the time of its publication, as it predicted that domestic interest rates would start to rise in around mid-2021. This seemed unthinkable to many during the tough autumn lockdown of the Czech economy.
[9] On the other hand, production and exports in the industrial sectors of the domestic economy, especially the automotive industry, were significantly reduced in 2021 due to disruptions to material and component supplies as a result of overloaded global supply chains. In addition to higher additions to inventories (inputs and unfinished products), this led to lower export growth in the second half of 2021 relative to the forecasts under assessment.
[10] However, this statement does not apply to the inflation forecast. In the factors-known simulation, monetary policy would indeed react to ensure price stability. Inflation would thus always return to the CNB’s 2% target in this simulation. In this sense, the simulation does not try to be a backward prediction of reality, which was influenced by the fact that the CNB was unaware of the factors at that time.