Time series of EU growth on year-on-year basis; time series of machinery and equipment adjusted for seasonal and working-day effects; 1992=100)
Source:Eurostat, CSO.
As discussed in detail earlier, provided that a certain sector of the Hungarian economy exposed to an asymmetric shock is hit by a specific negative external demand shock, then the given sector's exports as well as its value added from exports will fall over the short term.145 As a consequence of the decline in demand, deterioration in domestic market positions caused by intensifying competition in the domestic market and the fall in contractors' existing orders may lead to a reduction in the entire sector's output, irrespective of the direction of sales. The slowdown in the sector's performance may sooner or later feed through to other sectors as well, so total Hungarian gross domestic product may experience a larger drop than the decline in value added of the sector affected by the shock. Table F–1 presents these observations translated into numbers, based on 2000 current data.
5 4 3 2 1 0
%
Real growth of European Union (left scale)
Export sales of Hungarian and equipmernt (right scale) 3000 2500 2000 1500 1000 500 0
Jan-97 Jul-97 Jan-98 Jul-98 Jan-99 Jul-99 Jan-00 Jul-00 Jan-01 Jul-01
Table F-1
Effect on GDP of a fall in exports by specified industries146
Notes: * Using the value deriving from the empirical observation of the multiplier effect.
** Data released by CSO differ because of the different database.
Source: APEH's database of corporate tax returns for 2000.
For example, when exports of road vehicle manufacturers suffer a decline of 10%, and the cyclical position of the European Union remains unchanged, Hungarian GDP may fall by some 0.7% over the long term relative to its earlier path. Assuming that exchange rate policy were able to handle such an export demand shock effectively, then, following Hungary's joining the euro area, the decline in GDP caused by the negative external shocks hitting the industries exposed to asymmetric and specific shocks would be accounted for as the real economic costs of giving up the independent monetary policy. But even this, the largest economic sector, only contributes 2.3% to total GDP; and adjustment of the exchange rate would also boost demand for goods produced by the competitive sector left unaffected by the asymmetric shock. For this reason, we maintain that an independent exchange rate policy is an inappropriate tool to handle negative external demand shocks hitting the specified industries.
Industries Gross value added Gross value added Per cent fall in GDP as per exports as a per of the entire sector an effect of a 10% fall in
cent of GDP as per cent of GDP exports*
Manufacture of road vehicles 2.1 2.3 0.68
Manufacture of computers 0.4 0.5 0.14
Manufacture of lighting equipment 0.4 0.5 0.14
Manufacture of electronic components 0.3 0.4 0.13
Manufacture of electrical 0.3 0.3 0.08
equipment for engines and vehicles
Television, radio receivers etc. 0.2 0.3 0.09
Manufacturing firms, total 9.1 18.3**
146It deserves special mention that, within the specified industries, the proportion of firms producing in customs-free areas is very large. These firms are characterised by salient foreign trade volumes relative to gross value added. According to a 1999 survey, total GDP produced by firms located in customs-free zones was HUF 411 billion, while their exports exceeded HUF 2,400 billion, compared with imports amounting to HUF 2,070 billion.
VI.2. Identifying supply and demand shock using structural VAR estimates
The effect of a supply shock hitting an economy can be identified in two steps.
After choosing the appropriate time series, mostly a real and a related nominal variable, the systematic relationship between the two must be estimated in the first step. Assuming first order integrated variables, this produces the estimation of the vector autoregression model as follows:
where y and p denote the natural logarithm of the real quantity and the price relating to it, and ey and ep are the two random shocks affecting the two variables which may correlate with each other. Ai,…,Aqare the estimated coefficient matrixes.
In the second step, we decompose the random terms by imposing structural restrictions, and the shocks calculated this way can now be readily interpreted in economic terms. In our assumption, the actual data generating process is as follows:
where L is the lag operator, εd and εs are shocks affecting demand and supply respectively, and matrixes Bi represent the corresponding impulse response functions for output and the price level.
In case the stationarity assumptions are met, the estimated VAR model can be written in an infinite MA form similar to the structural model:
∆ y
t∆ p
t=
q∑
i=1
A
i∆ y
t–i∆ p
t–i+ e e
y t p t
∆ = ∑
∆
(1)
(2)
The relationship between the residues of the estimated model and the structural shocks can be written as e=Cε. In addition to the orthonormality of the structural shocks, another condition is required for the exact identification of C. As our purpose is to segregate demand and supply shocks, it is the most straightforward and most frequently used restriction in such cases that demand shocks do not alter the level of output over the long term.147
We performed the estimate using GDP and consumer price data in a quarterly breakdown, taking into account that GDP data for the accession countries examined are only available from 1992–93. Although the GDP deflator would be more suitable as a price variable, we chose the consumer prices, due to the better reliability of data.
The GDP time series used for the purposes of the estimate are at constant prices, as provided by OECD. The sources of consumer prices are the IFS database of IMF and the CSO in the case of Hungary. We seasonally adjusted all time series using the model-based approach of the TRAMO-SEATS software. There was a shift in German GDP in 1991 Q1 due to reunification, which we corrected by averaging the first difference between the logarithms of seasonally adjusted 1990 Q4 and 1991 Q2 for 1991 Q1.
We used the logarithmic differences of the seasonally adjusted and corrected levels in the estimate. As a first step, we estimated two-variable VAR models for the sample covering the period 1980 Q1–2000 Q4 for developed countries. In defining the number of lags, we relied on the Akaike Information Criterion. In the case of a few countries, the impulse response functions proved to be counterintuitive;
therefore, we reduced the number of lags according to the Schwarz Criterion in the
∆ = ∑
∆
147
In respect of the coefficient matrixes of the structural equation this means that , i.e. the sum of elements on the upper left is zero.
∑ =
(3)
case of these countries. Then, all impulse responses had the expected sign over the long term. The number of lags was 1–2 for each country, with the exceptions of Portugal and Luxembourg (6–6), Spain (4) and Ireland (3).