important input only for the production of consumer goods. Finally, industrial raw materials
are most important for the production of intermediate goods. It is striking how similar our
estimations are to these shares, both in terms of relative impact and size, which indicates that
our estimations seem to capture very well the pricing chain of the economy.
Table 6 Imported commodity input shares versus estimation results of 1% rise in
commodity prices
Imported inputs |
Estimated impact after 4 years | ||
Energy commodities___________ | |||
Production of: |
Producer prices: | ||
Energy |
0.26 |
Energy |
0.26 |
Intermediate goods |
0.01 |
Intermediate goods |
0.04 |
Consumer goods |
0.00 |
Consumer goods |
0.02 |
Consumption___________ |
0.01 |
HICPX____________ |
0.01 |
Food commodities | |||
Production of: |
Producer prices: | ||
Energy |
0.00 |
Energy |
- |
Intermediate goods |
0.00 |
Intermediate goods |
- |
Consumer goods |
0.02 |
Consumer goods |
0.02 |
Consumption___________ |
0.02 |
HICPX____________ |
0.02 |
Industrial raw material | |||
Production of: |
Producer prices: | ||
Energy |
0.00 |
Energy |
- |
Intermediate goods |
0.12 |
Intermediate goods |
0.11 |
Consumer goods |
0.01 |
Consumer goods |
0.02 |
Consumption___________ |
0.01 |
HICPX____________ |
0.01 |
Source: Eurostat and own calculations.
As the standard errors of our regressions only show the uncertainty around our point estimates
and not around the transmission through the pricing chain, we have also used bootstrapping in
order to obtain confidence bands around our impact multipliers. These confidence bands are
obtained in the following way: taking as an example the PPI consumer goods component, we
first estimate the equations for PPI energy and intermediate goods, which are situated at
earlier stages of the production chain, and compute their impact multipliers. Then, we
estimate the equation for PPI consumer goods, store the residuals of this equation and
compute the impact multiplier. We then re-order randomly the residuals of the PPI consumer
goods equation for each country in the panel, apply them to the fitted values and re-estimate
the equation with these bootstrapped data for PPI consumer goods in order to obtain a second
version of an impact multiplier for this component. After 10,000 replications of this
procedure, we take out the upper and lower 2.5% of the total of 10,000 impact multipliers for
this component (sorted by the impact after 16 quarters) and thereby obtain a confidence band
of 95%. The results for producer prices are shown in Chart 10, those for consumer prices in
Я ECB
Working Paper Series No 1104
November 2009