Input-Output Analysis in Current or Constant Prices: Does it Matter?
Erik Dietzenbacher & Umed TemurshoevFaculty of Economic and Business
University of Groningen
This project is funded by the European Commission, Research Directorate General as part of the 7th Framework Programme, Theme 8: Socio-Economic Sciences and Humanities.
Grant Agreement no: 225 281
Motivation
Availability physical IO tables (PIOTs, single unit of mass) triggered discussion
Treatment of waste(Hubacek & Giljum, 2003; Giljum & Hubacek, 2004; Giljum et al., 2004; Suh, 2004; Dietzenbacher, 2005; Dietzenbacher et al., 2009)
Substantial differences with MIOTs (monetary IO tables)Weisz & Duchin (2006)reason: sectoral prices are not uniform for all deliveries
Motivation
Example: 3-sector PIOT (million tons) and MIOT (billion DM) for Germany, 1990land appropriation (hectares)
Question:how much land used in each sector due to exports
Results (in 1000 hectares): PIOT MIOT
primary 5,822.4 6,339.3secondary 478.9 807.1tertiary 544.5 134.9total 6,845.8 7,281.3
Motivation
Example: 3-sector PIOT (million tons) and MIOT (billion DM) for Germany, 1990land appropriation (hectares)
Question:how much land used in each sector due to exports
Results (in 1000 hectares): PIOT MIOT %diff
primary 5,822.4 6,339.3 -8.2secondary 478.9 807.1 68.5tertiary 544.5 134.9 -75.2total 6,845.8 7,281.3 6.4
Research question
Central question in this paper:
To what extent do results differ betweenmodel based on IO table in current prices andmodel based on IO table in constant prices?
Note:Results will be exactly the same
if and only if each sector sells its goods and services for the same price
Single sectoral deflator does the job
Problem: deflators are cell-specific
Methodology
Exercise:
take a “new” final demand vector in current pricesdetermine the effects for:
sectoral gross outputs in constant pricesemployment
Three approaches
Methodology
Approach A:
deflation after gross output calculations in current prices
IO table in current prices → input coefficients in current pricesnew final demands in current prices → new gross outputs in current prices
→ use gross output deflators → new outputs in constant prices→ use labor coefficients (in current prices)
→ new sectoral employment
gross output deflators: total gross output sector j in constant pricesdivided by total gross output sector j in current prices
Methodology
Approach B:
gross output calculations in constant prices after deflation of final demands
new final demands in current prices → final demand deflators → new final demands in constant prices → IO tables in constant prices → input coefficients in constant prices
→ new gross outputs in constant prices→ use labor coefficients (in constant prices)
→ new sectoral employment
final demand deflators: final demands for good j in constant pricesdivided by final demands for good j in current prices
Methodology
Approach C:
using cell-specific deflators
delivery (i,j) in constant prices divided by delivery (i,j) in constant prices
final demand j in constant prices divided by final demand j in current prices
Methodology
new final demands in current prices → new gross outputs in current prices
make a new IO table in current pricesnew delivery (i, j) in current prices
= input coefficient (i, j) × new output j
deflate new IO table (incl new final demands) in current pricesusing cell-specific deflators
summation of each row → new gross outputs in constant prices→ use labor coefficients (in constant prices)
→ new sectoral employment
Methodology
Observe:
A (deflation of new outputs in current prices)requires gross output deflators only!
B (calculation of outputs in constant prices after deflation of final demands)requires full IO table in constant prices
C (cell-specific deflation)requires full IO table in constant prices (to determine the deflators)
Application
Denmark, 2000-2007
IO tables in current pricesIO tables in constant prices (base year 2000)
Employment data
130 sectors
“new” final demands =average of final demands over 2001-2007
Results, total gross outputs
2001 2002 2003 2004 2005 2006 2007 MeanEconomy-wide gross output, n = 130
(B-A)/A% 0.011 -0.033 -0.007 0.012 0.002 0.006 -0.009 -0.002(C-A)/A% 0.008 -0.022 -0.002 0.012 0.001 0.001 -0.010 -0.002(B-C)/C% 0.003 -0.011 -0.005 0.000 0.001 0.005 0.002 -0.001
Results, total gross outputs
2001 2002 2003 2004 2005 2006 2007 MeanEconomy-wide gross output, n = 130
(B-A)/A% 0.011 -0.033 -0.007 0.012 0.002 0.006 -0.009 -0.002(C-A)/A% 0.008 -0.022 -0.002 0.012 0.001 0.001 -0.010 -0.002(B-C)/C% 0.003 -0.011 -0.005 0.000 0.001 0.005 0.002 -0.001
%differences for total (economy-wide) gross outputs:very small0.001-0.002% on average
Results, total gross outputs
2001 2002 2003 2004 2005 2006 2007 MeanEconomy-wide gross output, n = 130
(B-A)/A% 0.011 -0.033 -0.007 0.012 0.002 0.006 -0.009 -0.002(C-A)/A% 0.008 -0.022 -0.002 0.012 0.001 0.001 -0.010 -0.002(B-C)/C% 0.003 -0.011 -0.005 0.000 0.001 0.005 0.002 -0.001
%differences for total (economy-wide) gross outputs:very small0.001-0.002% on average
further away from the base year 2000:one would expect larger differencesno clear pattern over time
Results, total gross outputs
2001 2002 2003 2004 2005 2006 2007 MeanEconomy-wide gross output, n = 130
(B-A)/A% 0.011 -0.033 -0.007 0.012 0.002 0.006 -0.009 -0.002(C-A)/A% 0.008 -0.022 -0.002 0.012 0.001 0.001 -0.010 -0.002(B-C)/C% 0.003 -0.011 -0.005 0.000 0.001 0.005 0.002 -0.001
“large” differences for B-A and C-A“small” differences for B-C
intuition: B & C both use information from full IO tabel in constant pricesA uses only gross output deflators
Results, aggregation
2001 2002 2003 2004 2005 2006 2007 Mean Economy-wide gross output, n = 130 (B-A)/A% 0.011 -0.033 -0.007 0.012 0.002 0.006 -0.009 -0.002(C-A)/A% 0.008 -0.022 -0.002 0.012 0.001 0.001 -0.010 -0.002(B-C)/C% 0.003 -0.011 -0.005 0.000 0.001 0.005 0.002 -0.001 Economy-wide gross output, n = 56 (B-A)/A% 0.028 0.026 0.042 0.033 0.001 -0.009 -0.005 0.017(C-A)/A% 0.022 0.023 0.038 0.028 0.000 -0.008 -0.010 0.013(B-C)/C% 0.006 0.004 0.004 0.006 0.001 -0.001 0.006 0.003
aggregation increases the %differencesthe maximum difference is 0.042%
Results, aggregation
2001 2002 2003 2004 2005 2006 2007 Mean Economy-wide gross output, n = 130 (B-A)/A% 0.011 -0.033 -0.007 0.012 0.002 0.006 -0.009 -0.002(C-A)/A% 0.008 -0.022 -0.002 0.012 0.001 0.001 -0.010 -0.002(B-C)/C% 0.003 -0.011 -0.005 0.000 0.001 0.005 0.002 -0.001 Economy-wide gross output, n = 56 (B-A)/A% 0.028 0.026 0.042 0.033 0.001 -0.009 -0.005 0.017(C-A)/A% 0.022 0.023 0.038 0.028 0.000 -0.008 -0.010 0.013(B-C)/C% 0.006 0.004 0.004 0.006 0.001 -0.001 0.006 0.003
aggregation increases the %differencesthe maximum difference is 0.042%
aggregation affects B-A and C-A more than it affects B-C (differences remain very small)
Results, employment
total economy-wide employment very similar conclusions:
- extremely small differences- no clear pattern over time- same distinction between B and C versus A
but: aggregation does NOT tend to increase the differences(averages over 7 years even decrease)
2001 2002 2003 2004 2005 2006 2007 Mean Economy-wide employment, n = 130 (B-A)/A% 0.010 -0.010 0.005 0.011 -0.004 -0.015 -0.027 -0.004(C-A)/A% 0.007 -0.005 0.005 0.006 -0.004 -0.012 -0.019 -0.003(B-C)/C% 0.003 -0.005 0.000 0.005 0.000 -0.003 -0.008 -0.001 Economy-wide employment, n = 56 (B-A)/A% 0.005 0.000 0.007 0.022 -0.005 -0.012 -0.007 0.001(C-A)/A% 0.002 -0.001 0.004 0.013 -0.004 -0.008 -0.005 0.000(B-C)/C% 0.003 0.002 0.003 0.009 -0.001 -0.005 -0.002 0.001
Results, sectoral level
differences at sectoral level are larger largest positive difference: 3.297% (B-A) and 3.280% (C-A) in 2007 manufacturing of office machinery and computers (0.06% of national output) largest negative differences: manufacturing and distribution of gas (approx 0.5% of national output)
2001 2002 2003 2004 2005 2006 2007 Mean (B-A)/A%, n = 130 Min -1.278 -2.562 -1.586 -0.862 -0.588 -1.215 -1.708 -1.400Max 0.472 0.307 0.348 0.635 0.520 1.747 3.297 1.046 (C-A)/A%, n = 130 Min -1.321 -2.542 -1.565 -0.850 -0.584 -1.090 -1.395 -1.335Max 0.463 0.303 0.346 0.632 0.490 1.679 3.280 1.028 (B-C)/C%, n = 130 Min -0.055 -0.255 -0.214 -0.279 -0.071 -0.270 -0.409 -0.222Max 0.070 0.063 0.069 0.110 0.123 0.575 0.868 0.268
Results, sectoral level
differences between B and C: for only 3 (out of 910 cases) differences larger than 0.4%
2001 2002 2003 2004 2005 2006 2007 Mean (B-A)/A%, n = 130 Min -1.278 -2.562 -1.586 -0.862 -0.588 -1.215 -1.708 -1.400Max 0.472 0.307 0.348 0.635 0.520 1.747 3.297 1.046 (C-A)/A%, n = 130 Min -1.321 -2.542 -1.565 -0.850 -0.584 -1.090 -1.395 -1.335Max 0.463 0.303 0.346 0.632 0.490 1.679 3.280 1.028 (B-C)/C%, n = 130 Min -0.055 -0.255 -0.214 -0.279 -0.071 -0.270 -0.409 -0.222Max 0.070 0.063 0.069 0.110 0.123 0.575 0.868 0.268
Results, sectoral level
Results, sectoral level
Conclusion at sectoral level:
differences may occasionally be > 1.0% but in these cases sectoral output < 1.0% of national output
Conclusions
A = model in current prices + gross output deflatorsB = model in constant pricesC = model in current prices + cell-specific deflators
B & C very close to each other, versus A
at sectoral level: few differences > 1%, for sectors with output < 1% of national output (results for employment are exactly the same)
at level of total economy-wide outputs and employment: differences are extremely small
Conclusions
“new” final demand vector: average of final demand vectors 2001-2007
what about “out of sample” vectors?
size doesn’t matter!! multiply new final demand vector with k then outcomes (outputs, employment) are multiplied with k %differences will remain the same
composition of the new final demand vector may matter
Conclusions
future extensions:
• further aggregation (28, 14, 7, 3 sectors) • are the results unique for Denmark 2001-2007?• what if constant price tables are not available every year
(2003: constant prices 2007: current prices, gross output + final demand deflators are B & C still so close to each other?)