2458
Views
411
Downloads |
The interplay of supply and demand shocks: measuring potential output in the COVID-19 pandemic*
Ozana Nadoveza Jelić**
Ozana Nadoveza Jelić
Affiliation: Croatian National Bank, Zagreb, Croatia; University of Zagreb, Faculty of Economics and Business, Zagreb, Croatia
 0000-0002-3651-7795
Nina Pavić** 
Article | Year: 2021 | Pages: 459 - 493 | Volume: 45 | Issue: 4 Received: September 18, 2021 | Accepted: September 22, 2021 | Published online: December 6, 2021
|
FULL ARTICLE
FIGURES & DATA
REFERENCES
CROSSMARK POLICY
METRICS
LICENCING
PDF
Source: Authors’ calculations.
Source: Authors’ calculations.
Source: Authors’ calculations.
Note: The letter D denotes a demand shock and the letter S a supply shock. Source: Author’s calculations.
Note: The output gap is expressed as a share of potential GDP, and the capital utilization rate as a share of total production capacity of the manufacturing industry. Source: European Commission (2020); authors’ calculations.
|
Variable
|
Source
|
|
CPI
inflation forecasts
|
Unpublished
and published (Macroeconomic Developments and Outlook) CNB's forecasts
|
|
GDP
forecasts
|
Unpublished
and published (Macroeconomic Developments and Outlook) CNB's forecasts
|
|
Real
GDP
|
Eurostat
(2015=100)
|
|
CPI
inflation
|
Croatian
Bureau of Statistics
|
|
Unemployment
rate
|
Croatian
Bureau of Statistics (Labour Force Survey)
|
|
Variable
|
Source
|
|
Real
GDP
|
Croatian
Bureau of Statistics, CNB's forecasts
|
|
CPI
excl. food and energy
|
Croatian
Bureau of Statistics, CNB's forecasts
|
|
Compensation
per employee
|
Eurostat,
CNB's forecasts
|
|
Unemployment
rate
|
Croatian
Bureau of Statistics (Labour Force Survey), CNB's forecasts
|
|
Total
employment
|
Croatian
Bureau of Statistics (Labour Force Survey), CNB's forecasts
|
|
Average
hours worked
|
JOPDD,
CNB's forecasts
|
|
Long-term
unemployment rate
|
Eurostat,
CNB's forecasts
|
|
Capital
|
Author's
calculations using PIM method (perpetual inventory method)
|
|
Working
age population
|
Croatian
Bureau of Statistics – forecast of the working age population 15-74, CNB's
forecasts
|
|
|
CNB
|
European Commission
|
|
Total
factor productivity
|
In
the medium term, the average growth rate from the period before the COVID-19
crisis is 1% (excluding 2009 when the decline was extremely large and
amounted to about -8%)
|
In
the medium term, we do not know what the projection is based on, because the
average TFP rate in the whole period is around 1%, and the current long-term
projection is 0.4%.
|
|
Investments
|
In
the medium term, investment growth of 5.5% is assumed
|
Calculated
from the investment / output ratio, investment growth rates towards the end
of the projection period are negative
|
|
Working
age population
|
CBS
- projection of working age population 15-74 (medium fertility variant with
medium migration)
|
Eurostat
– projection of working age population 15-74
|
|
Total
employment
|
CBS
(LFS) and the CNB projection, which suggests an increase in the number of
employees in the projection period.
|
National
Accounts; in the projection period, they suggest a milder increase in the
number of employees than the CNB in 2021 and 2022, and a decrease in the
number of employees in 2023.
|
|
Unemployment
rate
|
CBS
(LFS) and the CNB projection, which suggests a decrease in the unemployment
rate in the projection period.
|
In
the projection period, the unemployment rate is assumed to fall in relation
to HNB projection
|
|
Average
hours worked
|
A
decrease in working hours (JOPDD) is assumed.
|
Source
unknown. In 2020, potential working hours are higher than in 2019. In the
long run, EC forecasts slight growth.
|
Sources: European Commission, authors’ calculations.
Sources: European Commission, authors’ calculations.
Note: The revision of the potential GDP growth rate in 2016 on the right chart follows from the revisions of data related to the labour market and is not a consequence of the problems related to the estimation of potential GDP in the context of the corona crisis. Sources: European Commission, authors’ calculations.
Sources: European Commission, authors’ calculations.
Figure 1Comparison of different methodological approaches to estimating potential GDP and output gap DISPLAY Figure
Figure 2Illustration of the impact of pandemic shock on potential GDP and the output gap given the different nature of the shock DISPLAY Figure
Figure 3Estimation of potential output before and after the GDP 2020 data release (HRK bn) DISPLAY Figure
Figure 4Production factor contributions to potential output growth (in percentage, percentage points) DISPLAY Figure
Figure 5Potential growth and output gap estimated using different calibrations of the effect of the COVID-19 crisis on the trend and cycle component of TFP DISPLAY Figure
Figure 6Output gap and capacity utilization rate in Croatia (percentage) DISPLAY Figure
Table A3.1Data sources DISPLAY Table
Table A4.1Data sources DISPLAY Table
Table A5.1Assumptions used to calculate potential GDP DISPLAY Table
Figure A5.1Comparison of contributions to potential GDP growth, EC – left, ours – right (in percentage, percentage points) DISPLAY Figure
Figure A5.2Comparison of the potential GDP estimate from December 2020 and December 2019, EC – left, our estimate – right (HRK bn) DISPLAY Figure
Figure A5.3Comparison of the potential growth estimate from December 2020 and December 2019, EC – left, our estimate – right (percentage) DISPLAY Figure
Figure A5.4Comparison of the output gap (percentage) DISPLAY Figure
1 For a detailed overview of the ideas and development of the methodology related to the separation of the trend and cycle of GDP, see for example Cerra, Fatas and Saxena (  2020).
* All views presented in this paper are the authors’ own and do not necessarily reflect the official position of the Croatian National Bank.
** The authors would like to thank two anonymous reviewers and colleagues from the Croatian National Bank for their useful suggestions. This article will also be published as Croatian National Bank Working paper.
2 Earlier thoughts on economic cycles and fluctuations did not address their practical identification in a modern sense (see in more detail in Beveridge and Nelson, 1981).
3 One of the main purposes of the book was to list the methods of measuring the cyclical behaviour of the economy developed by the NBER and their practical application in identifying the turning points of the business cycle.
4 Despite the fact that business cycle identification today uses methods that over time deviated from the approaches proposed by Burns and Mitchell ( 1946), the conceptual framework they developed can still be recognized today in various methods of identifying business cycle turning points. One of the first researches in which the turning points of the business cycle in Croatia were identified, using three different methods, was conducted by Krznar ( 2011a).
5 According to Beveridge and Nelson ( 1981), one of the popular methods identified cycles as output deviations from a deterministically determined trend where the trend was most often shown as (most commonly polynomial) function of time (see for example Fellner, 1956), which is a very strong assumption. Although in Friedman ( 1957) when dividing income into permanent and transient components, the permanent (trend) component was not deterministic, it implied "fairly strong initial assumptions about the stochastic properties of the permanent component" (Beveridge and Nelson, 1981). Two alternative cycle measurement methods were developed by Mintz ( 1969; 1972), one of which defined the trend using a centred 75-month moving average, and the other focused on the analysis of fluctuations in rate of change. On the technical side, the problem arises towards the end of the sample in which future observations of the series under observation are not available (for a more detailed critique, see Beveridge and Nelson ( 1981)).
6 Arčabić ( 2018) analyses theoretically and empirically the separation of trend and cycle components of GDP in order to identify the nature of shocks (supply or demand) in selected post-transition countries. The paper shows how demand shocks are dominant in explaining the business cycles of almost all post-transition countries, which is in strong contrast to the conclusions of the real business cycle theory.
7 See for example Blanchard and Quah ( 1989).
8 Although this is the dominant conceptual view of potential GDP, different estimation methods often start from different definitions of potential GDP (see Chapter 3).
9 Depending on the estimation method, this problem may be more or less pronounced. The problem is most pronounced with (often symmetrical) two-sided filters such as HP filters (Jovičić, 2017) in which variable shifts (e.g., GDP in the case of univariate filter estimation) are used forward and backward to estimate potential GDP (historical and future data) where no data on future trends are available at the end of the sample.
10 The consequences of these two problems for the conduct of economic policies are beyond the scope of this paper.
11 Also, these new data cause (sometimes significant) revisions of historical estimates of potential GDP, which is certainly not a desirable feature of potential GDP estimates from the perspective of theoretical assumptions and desirable properties of potential GDP estimates (see Chapter 4).
12 For a more detailed explanation and demonstration of these properties, see Chapter 3.
13 See Jovičić ( 2017) for a detailed overview of differences in estimates of the cyclically-adjusted budget balance depending on the method used to estimate potential GDP.
14 Some authors (see for example Heimberger, 2020) believe that the underestimation of the negative output gap has contributed to the deepening and prolongation of the consequences of the 2008/2009 global financial crisis due to excessive emphasis on the need for fiscal consolidation in countries with unfavourable fiscal position (indicated by the cyclically-adjusted budget balance).
15 Hysteresis is easiest to explain by the impact of crises on the labour market. As the unemployment rate rises during the crisis, part of the labour force becomes inactive and deprives human capital during inactivity. The longer the period of inactivity, the more difficult it is to return to the labour market, which leads to an increase in structural unemployment, and thus to a long-term loss of productive resources that reduce potential GDP. In addition, numerous rigidities in the labour market can also slow down adjustment towards equilibrium in this market. Due to slow adjustment, changes in the unemployment rate become persistent and may have long-term consequences for potential GDP.
16 In that part in which it does not distort the economic balance.
17 Appendices 1-4 give a more detailed description of all four methodological approaches for estimating potential GDP.
18 A similar analysis of different ways of estimating potential GDP was conducted by Jovićić ( 2017).
19 Although both institutions use the production function method, estimates of potential GDP differ. The differences arise from the way the trend and cycle components of production factors are estimated (primarily those related to labour), from using different indicators (data) for production factors of production and finally from using different projections of production factors in the long run (see Appendix 5).
20 Alichi et al. ( 2015) showed that, although real-time estimates of potential GDP are quite uncertain, this approach gives more adequate estimates of potential GDP compared to univariate statistical filters.
21 See Jovičić ( 2017) for a more detailed analysis of the sensitivity of different potential GDP estimates to the end-of-sample bias based on Croatian data.
22 Arčabić ( 2018) gives a detailed overview of theoretical concepts and methodological approaches for separating the cycle and trend component of GDP, in the context of supply and demand shocks. The analysis includes Croatia, and the author shows that in Croatia, fluctuations in GDP in the past were dominated by demand shocks.
23 The authors conducted the analysis for the US, UK, euro area and Japan.
24 The red and blue lines start to diverge from 2015, but their differences become significant after 2016.
25 The same thing happened with the estimate of potential GDP published by the European Commission when comparing estimates of potential GDP in autumn 2019 (see European Commission, 2019) and autumn 2020 (see European Commission, 2020). The EC's estimates are compared with the estimates in this paper and presented in Appendix 5.
26 In the crisis of 2008/2009 this transmission mechanism has been strong in European Union countries (see, for example, the ECB, 2012).
27 See for example Caballero and Hammour ( 1994).
28 As explained earlier, this is a consequence of the introduced government support measures in the labour market, while in the case of capital the effect is reduced due to investments in earthquake-affected areas.
29 Given that this is a residual category, the estimated TFP, among other things, contains errors in measuring production factors and the degree of utilization of existing capacity in the economy. However, we do not have a sufficiently long reference measure for the degree of capacity utilization in Croatia that we could use to estimate potential GDP. Also, this is why the greatest short-term effect of the pandemic shock can be attributed TFP, because measures to combat the spread of the pandemic have had the greatest impact on the degree of utilization of existing physical and human capacity in the economy.
30 The criteria are in line with the ECB's ( 2020) recommendations on the desirable features of estimating potential GDP.
31 See Nadoveza Jelić and Ravnik ( 2021).
32 The adjustment was performed using the PACMAN macroeconomic model and projections of headline and core inflation with respect to alternative output gap sizes in the Phillips curve. The ultimate goal was for the selected calibration to result in an output gap that would align the 2020 inflation with the HNB's official December 2020 inflation projection.
33 Data on TFP for 2009 was not included due to an extremely high drop of approximately 8%.
34 The differences from the CNB's estimate are explained in more detail in Annex 5.
35 This indication of the earlier beginning of the economic “overheating” would signal to policymakers the need to introduce restrictive measures quicker due to the potential inflationary pressures.
36 Nelson ( 2008) and Morley and Panovska ( 2020) argue that the correlation between the appropriately estimated output gap and the one-year-ahead real GDP growth rate should be negative. The intuition behind this argument is that as the economy returns to its long-term trend when the output gap is positive we should expect future GDP growth rates to be below average. Correlations between output gap measures presented in Figure 6 and the one-year-ahead real GDP growth rate were calculated to verify if these measures satisfy this intuition. The correlations obtained verified that all three measures of the output gap are negatively correlated with the future GDP growth rate.
37 That is, we propose and argue one basic calibration of the possible countless calibrations of the decomposition of GDP decline in 2020 to supply and demand shock.
39 Several studies question the existence of the Phillips curve in Croatia; see for example Krznar ( 2011b), Botrić ( 2012), Jovičić and Kunovac ( 2017). When estimating the potential GDP in Croatia using a simple multivariate filter based on Alichi et al. ( 2015), we also try it with the inactive mechanism of the Phillips curve. In this case, equation P3.5 is simply given by: πt= πt-1. However, since the results do not differ significantly in the two alternative model specifications, we present the results with the active Phillips curve. On the other hand, when estimating potential GDP using the unobserved components model the results differ significantly and, therefore, we presume an inactive mechanism of the Phillips curve.
|
