Handbook of Regional Growth and

Development Theories

Edited by

Roberta Capello
Politecnico di Milano, Italy

and

Peter Nijkamp
VU University Amsterdam, the Netherlands

Edward Elgar
Cheltenham, UK • Northampton, MA, USA
© Roberta Capello and Peter Nijkamp 2009

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system or
transmitted in any form or by any means, electronic, mechanical or photocopying, recording, or
otherwise without the prior permission of the publisher.

Published by
Edward Elgar Publishing Limited
The Lypiatts
15 Lansdown Road
Cheltenham
Glos GL50 2JA
UK

Edward Elgar Publishing, Inc.

William Pratt House
9 Dewey Court
Northampton
Massachusetts 01060
USA

A catalogue record for this book

is available from the British Library

Library of Congress Control Number: 2008017423

PEFC/16-33-111
CATG-PEFC-052
www.pefc.org

ISBN 978 1 84720 506 3 (cased) (cased)

Printed and bound in Great Britain by MPG Books Ltd, Bodmin, Cornwall
21 Modern regional input–output and impact
analyses
Jan Oosterhaven and Karen R. Polenske

21.1 Introduction
Economic impact analysis has a long tradition in the input–output (IO) field. A search on
Google in May 2007 for impact analyses and IO models produced 1 090 000 records.
Many were undoubtedly duplicates or referred to only one or the other of these terms, but
we were nevertheless impressed with the proliferation of this technique. Of the various
applications of IO models, impact analysis is undoubtedly the most widely used. Many of
the early applications estimated economic impacts, but soon analysts were also studying
environmental, energy, transportation, land-use and other types of impacts, and these
have proliferated greatly beginning in the 1990s. With the recognition of the important
worldwide climate change effects, we anticipate that analysts will conduct even more
environmental- and energy-impact studies than before.
Underlying the regional analyses is the important basic theory of input–output and
socio-economic accounting. We begin by reviewing this basic theory in terms of some of
the significant methodological debates that occur. Although not all developments are
region-specific, we cover them because regional analysts are beginning to adopt these the-
oretical advancements in their work. For the applications, we restrict our review to
regional and multi-regional impact analyses and the development of computer program-
ming packages that help analysts to conduct such studies quickly.

21.2 Theory of demand-driven IO and SAM impact analysis

One of the most frequently heard criticisms of IO analysis concerns the assumption that
the input–output coefficients are constant. By making this assumption, Leontief was able
to use data over ten-year and longer periods. Early tests by Carter (1970) and Vaccara
(1969) using US national input–output tables for 1939–60 showed that this was not an
unrealistic assumption, and Carter (1970) and Strout (1967) found that good output esti-
mates could be achieved by making changes in the input coefficients for only selected
sectors, such as chemicals and energy, respectively,
Those of us working at the regional scale, due to lack of data, often had to assume that
the national and regional technologies were identical. This seemed less realistic, due to
regional product-mix and price differences, than to assume that the actual technology
remained constant over time. Rather than to develop new theoretical models to take
account of economic geography and spatial accessibility differences, early US regional
analysts initially used surveys to obtain the data required for the regional IO tables. By
the 1980s, as funding became limited, they devised non-survey methods to estimate
regional IO tables, based upon either national technologies or upon regional technologies
from a different region. Only since 1990 have new spatial economic theories begun to
surface for use with the regional economic impact models. As we show later, the current

423
424 Handbook of regional growth and development theories

regional socio-economic impact analysts are building models to account for changes in
both technological and interregional trade relations, thus bringing economic geography
theories into prominence.
In practice, non-survey, symmetric IO table (IOT) construction is closely connected
with IO model building. In theory, however, they relate to two different operations. Most
data construction assumptions use uniform distributions to fill in lacking data in absence
of real data. Even when analysts use non-linear gravity- or entropy-maximizing assump-
tions to construct interregional IOTs (Oosterhaven, 1981a, Appendix; Batten, 1983), their
aim is to come as close to the lacking data as possible. The same holds when they use
‘industry technology’ or ‘commodity technology’ assumptions to construct symmetric
industry-by-industry or commodity-by-commodity IOTs from supply and use tables
(Kop Jansen and ten Raa, 1990; ten Raa and Rueda-Cantuche, 2003).
These data construction assumptions differ from the theoretically based behavioral
assumptions of fixed intermediate input ratios aij and fixed primary input ratios ckj used
in IO modeling. Notwithstanding the information on changes in these ratios over time,
analysts still commonly assume them to be constant when modeling the future or when
estimating impacts of specific exogenous changes with the Leontief model. One main
justification is that, for closed economies, the analyst can theoretically derive fixed ratios
by assuming profit-maximizing firms that produce total output xj with a Walras–Leontief
production function under constant returns to scale:

xj  max (zij aij, with i  1, ..., I, vkj ckj, with k  1, ..., K), (21.1)

and that firms sell this output on markets with full competition. This combination of
assumptions assures that, irrespective of the relative prices of intermediate and primary
inputs, cost minimization with a given total output xj always leads to fixed input ratios
and to the well-known IO demand functions for intermediate and primary inputs:

zij  aij xj and vkj  ckj xj or in matrix notation: Z  A x̂ and V  C x̂

ˇ (21.2)

where x̂ denotes a diagonal matrix, and where the combined column sum of A and C
equals one, that is, i’ A + i’ C = i’. Note that equation (21.2) implies full economic com-
plementarity of all inputs.
Adding the definition of total demand and assuming that the supply of output always
follows total demand gives:

xi  j zij q fiq or in matrix notation: x = Z i Fi (21.3)

Solving equations (21.2)–(21.3) for exogenous final demand f = F i, leads to the well-
known solution of the IO model for any aggregate impact variable v, such as total regional
or national value-added, employment, energy use or CO2 emissions:

v = c’ x = c’ (1A)1f (21.4)

in which c’ indicates a row with value-added, employment, energy use, or CO2 emission
coefficients per unit of output. Note that equation (21.3) also implies that any exogenous
 Modern regional input–output and impact analyses 425

or endogenous change in demand is fully met by the appropriate supply of intermediate

and primary inputs. Analysts who use the standard IO model thus assume that there are
no supply restrictions. Therefore this IO model is best labeled as a demand-driven quan-
tity model.
This foundation of equation (21.4) in production theory becomes insufficient if we move
from a closed economy to an open economy. Consider the closed world economy as a
system of R open regional economies, then (21.4) mathematically also describes the inter-
regional IO model (Isard, 1951) as well as the multi-regional IO model (Chenery, 1953;
Moses, 1955; Polenske, 1980), but then the vectors c, x and f all have dimension IR and the
matrices I and A have dimension IR x IR. The most important difference with the closed
economy model is that the intermediate input coefficients now also get two dimensions:

ij  tij aij
rs %s
ij  ti % aij
interregional IO model: ars multi-regional IO model: ars rs %s (21.5)

with the % indicating the summation over the relevant index. Equation (21.5) explicitly
shows that the interregional and the multi-regional input coefficients ars ij represent the
product of real technical coefficients aij%s from production function (21.1), and cell-specific
trade coefficients trsij in the case of the interregional input–output (IRIO) model, or (row)
aggregate trade coefficients trs i% in the case of the MRIO model. In the case of the single-
region IO model, the intra-regional trade coefficients trr rr
ij and ti% are better known as,
respectively, cell-specific or aggregate, self-sufficiency ratios or regional purchase coeffi-
cients (RPCs; Stevens and Treyz, 1986).
The theoretical foundation for assuming the trade coefficients to be fixed is less con-
vincing than that for the technical coefficients. The analyst may assume that the output
of, for example, agriculture is a different product in each different region. The trade
coefficients will then get a technical character and will be fixed for the same reasons as the
technical coefficients. As each cell then relates to different goods, this assumption fits best
with the IRIO model. The analyst may also assume that the products of, again for
example, agriculture in different regions are close substitutes for each other. The trade
coefficients will then only be fixed for as long as the relative prices of agricultural output
from different regions remain unchanged. As relative prices will influence all trade
coefficients along a row of the IO table in the same manner, this assumption fits best with
the MRIO model.
In interregional impact studies, equation (21.4) is disaggregated by regions to allow for
the separate estimation of intra-regional impacts and interregional spillover effects
(Miller and Blair, 1985). When intra-regional impacts from the single-region model are
compared with those from the interregional model, the latter will be larger. This difference
is caused by interregional feedback effects. These link the imports of the home region to
the output levels of other regions, which are linked back to the intermediate exports of
the home region. As a consequence, exogenous final demand will be smaller than in the
single-region model, as the export of intermediates becomes endogenous. Of course,
when the smaller exogenous final demand is multiplied by the larger multipliers that
include the interregional feedbacks, the resulting endogenous employment and value-
added will be the same. Interregional feedbacks are found to be relatively large (5–15
percent) for regions within well-integrated large conurbations and smaller ( 5 percent)
for relatively isolated regions (Oosterhaven, 1981a).
426 Handbook of regional growth and development theories

The size of the feedbacks and this difference become larger when the standard IO model
is extended with a consumption function for labor incomes, as labor incomes earned in
the home region will spill over into other regions and feed back into the home region
through the additional mechanisms of interregional commuting and interregional shop-
ping (Madsen and Jensen-Butler, 2005). There is a whole family of demo-economic exten-
sions of the basic Type I model into Type II, III, and so on, regional IO models (Batey,
1985). However, the important distinction between increases in labor incomes accruing to
resident workers (intensive income growth), new labor incomes accruing to migrants and
unemployed (extensive income growth), and the loss of benefits of formally unemployed
(redistributive income growth) can only be modeled properly if levels of economic activ-
ity are explicitly distinguished from changes therein (Oosterhaven and Folmer, 1985).
With levels and changes in levels distinguished, the interregional Type III variant of
equation (21.4) becomes:

v  c’ x c’ x1  c’ (I  A  Qw W ĉw Qu U ĉu ) 1 f c’ x1 (21.6)

If Δv represents, for instance, the system-wide change in employment, Δc’ represents the
IR-row with decreases in employment coefficients due to nominal labor productivity
increases, and x–1 represents the output impact of the combined lagged endogenous and
exogenous variables. Furthermore, Qw and Qu represent the consumption expenditure on
products from sector i in region r, respectively, per working resident and per unemployed
resident in region s. The interregional diagonal blocks of W (with r wrs j  1) represent the
shares of the new jobs in sector j in region s directly or indirectly taken up by residents of
region r, and the comparable blocks of U (with r urs j 1) represent the shares of the new
jobs in sector j in region s directly or indirectly taken up by residents of region r that were
formally unemployed. Finally, the diagonal matrices ĉw and ĉu represent the per unit labor
incomes and the per unit lost unemployment benefits, respectively. Typically, W and U are
determined by means of an IO vacancy-chain sub-model. With the unemployment
benefits of the Netherlands, Type III impact multipliers c’(I – A – Q + Qu)–1 from equa-
tion (21.6) move between 35 percent and 60 percent of the difference between Type I mul-
tipliers c’(I – A)–1 and Type II multipliers c’(I – A – Q)–1 (van Dijk and Oosterhaven,
1986).1
The distinction between demo-economic models and social accounting models (SAMs)
is not large in the sense that a SAM may be interpreted as a demo-economic model with
a more elaborate disaggregation of the household sector. It is, however, more fundamen-
tal in that SAMs invariably start with defining the underlying accounting framework
explicitly, and then derive their impact multipliers more or less directly from it. Pyatt
(2001) uses this approach to show that sectorally and regionally disaggregated Keynesian
income multipliers (Miyazawa and Masegi, 1963; Miyazawa, 1976) can be viewed as
special cases of various SAM multipliers (Pyatt and Thorbecke, 1976; Pyatt and Round,
1979). The core difference is that Miyazawa multipliers relate the factor (capital versus
labor) generation of income by sector and region, directly to the spending of that income
on products from different sectors and regions. SAM multipliers are more general in that
they add the redistribution of income by different institutions in-between the generation
and the spending of income to the IO model. As a consequence, SAMs are better suited
to study the impact of policy instruments on the distribution of income and poverty.
 Modern regional input–output and impact analyses 427

The obvious next question is: how far should an analyst endogenize the various com-
ponents of final demand? Studying the regional impacts of plant close-downs with a
SAM, Cole (1989, 1997) advocates the fullest possible closure of the single-region model
to capture all possible short- and long-run impacts. Government expenditure is made
dependent on tax income, investment expenditure on the operating surplus, and regional
exports on regional imports. This led to a major debate with Jackson et al. (1997) about
zero exogenous demand and infinite multipliers, which was concluded by Oosterhaven
(2000). Analysts cannot endogenize interregional feedbacks consistently without speci-
fying the full interregional model, in which case the full closure of the rest of the model
indeed results in zero exogenous demand. As a consequence, such a SAM may no longer
be used to evaluate the impacts of changes in demand, because these have become
endogenous.
The danger of overestimating impacts also occurs when total value-added or employ-
ment of an existing firm or sector is multiplied by that firm’s or sector’s Type II normal-
ized value-added or employment multiplier, c’ (I – A – Q)–1 ĉ–1, to indicate its economic
importance for the economy at hand. Of course, this is a misuse of impact analysis for
public-relations purposes. Formally, an analyst may only multiply an IO multiplier by
exogenous final demand and never by endogenous value-added or employment. Imagine
that the average normalized employment multiplier equals 2 and that the analyst would
apply this way of estimating the impact to all sectors; then, the predicted size of the total
economy would be twice its actual size. One solution is to correct the calculated ‘gross
impact’ of a certain sector with the part of that sector that is endogenously dependent on
the rest of the economy in order to obtain the net impact of that sector (see Oosterhaven
et al., 2003, for energy distribution). A second solution is to define a net multiplier that
may be multiplied with a sector’s total employment or output, such that the weighted
average of all sectors’ net multipliers equals one (Oosterhaven and Stelder, 2002).2
More generally, the fullest possible closure of standard IO models exacerbates the the-
oretically already problematic one-sidedness of that model. Possible shortages on local
labor markets, price and wage reactions, and the pressure to develop new markets and new
products will, for example, reduce the demand-driven quantity impact of a plant close-
down. In such cases, endogenizing price effects and supply reactions may be more impor-
tant than a full closure of the demand side, that is, if analysts are interested in the best
estimate of the real impacts instead of maximum multipliers.

21.3 Theory of price and supply-side impact analysis

An analyst may use the less well-known dual of the Leontief model to estimate endoge-
nous output price impacts p of the exogenous primary input prices pv (Leontief, 1951).
The dual, however, cannot be used to model the effects of price changes on quantities, as
follows from the solution of the standard (Type I) price model (Schumann, 1968):

p’  pv’ C (I  A) 1 (21.7)

In equation (21.7), the K exogenous, capital, labor, and import price changes pv are
directly passed on via C into output price changes, and are then indirectly carried forward
further via the rows of A into equilibrium total output price changes. Therefore, equation
(21.7) is best characterized as a cost-push price model.
428 Handbook of regional growth and development theories

Leontief price and quantity model Ghoshian price and quantity model
pi• p•j
Demand Supply

P-model P-model
Supply Demand

Q-model Q-model

xi xj

Figure 21.1 The functioning of markets in Leontief and Ghoshian IO models

In the left-hand panel of Figure 21.1, we show the relationship between the demand-
driven quantity model and this cost-push price model for an individual IO market. The
demand curve (driven by the quantity model) shifts left and right along the perfectly
elastic supply curve (no shortages). Independently, the supply curve (driven by the price
model) shifts up and down along the perfectly inelastic demand curve. This IO price
model, inter alia, has been used to estimate the price effects of pollution abatement
(Giarratani, 1974) and the effects of energy price rises in a multi-regional input–output
(MRIO) model (Polenske, 1979).
When the interregional quantity model is extended with a consumption function for
labor incomes, its dual price model of course changes analogously. Wages become
endogenously determined by the prices of consumption goods, and the remaining exoge-
nous primary input prices get multiplied with larger Type II cost-push multipliers result-
ing in the same equilibrium prices for total output (see Oosterhaven, 1981b, for
interregional wage and price impacts of the oil price hike).
Also disregarding the price-induced impact of supply on demand, Giarratani (1976),
Davis and Salkin (1984) and Chen and Rose (1985) have used the supply-driven IO model,
developed by Ghosh (1958), as a direct way to model the impacts of natural resource
shortages on output. In every respect, this quantity interpretation of the Ghosh model
represents the pure opposite of the demand-driven quantity model, as follows from its
solution for endogenous final demand (Oosterhaven, 1996):

F  v̂ (I  B) 1 D (21.8)

In equation (21.8), exogenous primary supply v̂, without further need of intermediate
inputs, directly leads to equally large total inputs x̂ (the direct forward effect), which is dis-
tributed to purchasers according to IxI fixed intermediate output coefficients, bij = zij/xi,
and IxQ fixed final output coefficients, diq = fiq/xi. The intermediate outputs, without
further need of primary or intermediate inputs, lead to equally large total inputs (the first
 Modern regional input–output and impact analyses 429

round indirect forward effect), which are again distributed to intermediate and final pur-
chasers according to B and D, and so on.
Originally, Ghosh formulated his model not in terms of quantities, but in terms of
values. Dietzenbacher (1997) proved that a value interpretation can be formulated such
that it is equivalent with the Leontief cost-push price model. Quantities remain
unaffected, and analysts may evaluate final output price impacts of exogenous primary
input prices in terms of values instead of in terms of prices. With this interpretation, the
row sums of the Ghosh-inverse (I – B)–1 may still be used as a descriptive statistic, mea-
suring the size of a sector’s forward linkages, just as the column sums of the Leontief-
inverse (I  A) 1 are used to measure the size of the backward linkages.
After the criticism by Oosterhaven (1988), the quantity interpretation of the supply-
driven IO model is no longer used, as it theoretically allows cars to drive without gasoline
and factories to work without labor. Only by intelligently combining processing
coefficients (= inverses of the technical coefficients aij%r and ckj %r) with intermediate output
rr
or allocation coefficients (bij ), while adapting regional purchase coefficients (trri% ) to accom-
modate for import and export substitution (Oosterhaven, 1988), may the forward effects
idea of the supply-driven IO model still be used (see Cartwright et al., 1982, for a nuclear
disaster application, and Oosterhaven, 1981a, for a land-reclamation application).
Interestingly, the dual of the Ghosh quantity model (Davar, 1989) has not been used
for price impact studies yet, whereas this is clearly possible, as follows from its solution:

p  (I  B) 1 D pf (21.9)

In equation (21.9), the Q prices for each column with homogenous final inputs (pf) are
exogenous, whereas the I prices for each column with homogenous sectoral inputs (p) are
endogenous. Any increase in, for example, the unit price for exports pq leads to a direct
increase in the unit revenues of each sector i with diq pq. Under full competition this
increase is entirely passed on into the price of the single homogenous input of sector i. In
the next round, this price increase leads to an increase in the unit revenues for all sectors
j with bji pi, which is further passed on again via B, and so on. Naturally this model may
best be characterized as the demand-pull price model.
The right-hand panel of Figure 21.1 shows how the Ghoshian price and quantity model
‘interact’. In the quantity model, the supply curve shifts left and right along a perfectly
elastic demand curve, not causing any price reaction (consumers consume whatever is
supplied at the going price). Independently, in the price model, the demand curve shifts
up and down along the perfectly inelastic supply curve, not causing any quantity reaction.
According to Oosterhaven (1996) the demand-pull price model is less implausible than its
companion supply-driven quantity model. But it is clear from Figure 21.1 that the more
plausible Leontief price and quantity model also may be labeled as ‘special’.
In fact, both sets of IO models represent extreme cases of the general equilibrium
model. Clearly, implementing the latter model at the combined inter-sectoral and inter-
regional level is more complicated and far more data-demanding than the comparable IO
model. For this reason, most developments in impact analysis seek to modify the basic
Leontief model by introducing more flexible (for example translog) production functions
for capital, labor, and intermediate input (for example the KLEM production function,
that is, Kapital Labor Energy and Materials), and by introducing econometrically
430 Handbook of regional growth and development theories

estimated consumption, investment and export functions, while sticking to the Leontief
specification for the matrix of intermediate demand only.

21.4 Developments in IO data construction and model applications

Spatially, analysts have constructed both regional and multi-regional IO impact models.

Regional IO table construction and applications

At a regional scale, we have progressed relatively slowly from the initial impact studies,
which included Moore’s (1955) classic article on ‘Regional economic reaction paths’, to
the well-explained inter-industry model for Utah by Moore and Peterson (1955), in which
they used a supply-demand pool procedure to estimate the regional transactions, to
Hirsch’s (1959) St Louis model for which he collected data from company records and
built a detailed export sector. Miernyk’s work in the nine Colorado river basins (with
Udis), and in the state of West Virginia, certainly contributed greatly to the collection of
regional data through surveys in the United States. In his easy-to-read book The Elements
of Input–Output Analysis Miernyk (1965) gave many useful guidelines for survey design
and model building, including how to select ‘best-practice’ (most technologically
advanced) firms. He also started the use of the terms ‘Type I’ and ‘Type II’ multipliers.
The most extensive and longest-lasting collection of regional data through surveys is by
the Washington State researchers in a tradition dating back to 1963 and continuing to
today.3 That state table is used as the standard by many US regional analysts.
Since the 1970s, most IO tables for other US regions have been estimated using various
short-cut techniques, such as multiplying national technical coefficients by regional loca-
tion quotients, primarily because of the high costs involved in conducting surveys. These
non-survey techniques, however, produce relatively inaccurate regional tables (see
Polenske, 1997, for an evaluation). Moreover, by using location and other quotients, most
methods implicitly maximize intra-regional transactions and minimize regional imports
in one way or another. Consequently, all regional multipliers from such tables have a sys-
tematic upward bias, even when analysts claim that there is relatively little cross-hauling
(as West, 1990, p. 108, does for Australia). In more densely populated and diversified
economies, however, cross-hauling is the rule, even more so when commuting across
regional boundaries is important and Type II multipliers are concerned (Oosterhaven,
1981a).
Many US regional analysts use the Washington State table either to create input–
output tables for their own region or to test for differences in the national and state
coefficients. Analysts could reduce errors to less than 10 percent only by sectoral aggre-
gation or by using exogenous information on more than 60 percent of the non-zero cells
(Polenske, 1997). The latter observation stresses the conclusion of Lahr (1993) that hybrid
methods should be preferred to simple non-survey methods.
A hybrid Dutch regional IO table, for example, resulted from a survey into four
tourist-specific branches in the province of Drenthe. The survey regional import
coefficients of the tourist branches were applied to the national technology coefficients
for the other sectors. The complete IO table resulted by combining the survey and non-
survey columns into a single table. A comparison with the multipliers from a survey-
based bi-regional table showed differences in the indirect part of the tourist branches’
multipliers of only 5–10 percent, whereas the differences in the indirect parts of the
 Modern regional input–output and impact analyses 431

other multipliers ran from –50 percent to +65 percent, with some outliers beyond that
(Spijker, 1985). We believe that this shows that both Bourque (1990) and Beemiller
(1990) are right in their discussion in the International Regional Science Review;
Bourque in his rejection of RIMS’s non-survey alternative for the Washington State
IOT and Beemiller in his claim that combining direct information for the sectors of an
impact study with a non-survey IOT produces sufficiently accurate estimates for most
practical impact questions.
The Washington State IO model is widely used for a diverse set of economic impact
analyses. Bill Beyers, for example, used it to estimate the economic impacts of arts and
cultural organizations on the Washington economy; the impacts of building and operat-
ing the new Seattle symphony hall; the impacts of the Mariners Baseball Team; and those
of the Fred Hutchinson Cancer Research Center. As indicated above, the Washington
IOT is extremely well known among US regional planners, many of whom have used it to
calculate the accuracy of their own tables, and/or to estimate state tables based upon the
Washington State table.
For the United States, two options seem relevant. First, regional technologies are
sufficiently diverse that regional analysts should be arguing for funds to conduct surveys
to derive their tables based upon actual data for that region. Second, analysts should be
devising new methods of estimating regional tables from the available data. Neither effort
is occurring, with the exception of the MITER accounts being developed by the US
Department of Agriculture (USDA) and Massachusetts Institute of Technology (MIT)
staff, described later on. We note that the construction of tables through survey methods
requires at least $200 000 per state. If each of the 51 US regions were to construct such a
table, the total would be over $10 million. Thus, the rationale for assembly of a complete
set of state tables by a central government unit is obvious.
Elsewhere, developments have been only partially comparable. Dutch regional IO
analysis up to the 1970s may be summarized as running ‘from regional tables with only
limited information used for primarily descriptive purposes towards ideal interregional
tables used for analytical purposes, such as estimates of economic impacts, experiments
with programming models and building full forecasting models’ (Oosterhaven, 1981a,
p. 23). As opposed to the Dutch tables constructed in the 1970s and also different from
most tables constructed in the United States and in Australia (West, 1990), the Dutch bi-
regional tables of the 1980s are mainly based on surveys of export coefficients instead of
import coefficients. This change in trade survey strategy also led to the need of an explicit
domestic intermediate and final sales table (Boomsma and Oosterhaven, 1992).
This strategy change was the outcome of earlier experiences that showed that firms, as
a rule, are better informed about the destination of their output than about the origin of
their inputs. This is especially the case if there are many different inputs and/or if inputs
are purchased through wholesale or retail channels. Firms are only well informed about
the real origin if they deal with one or a few dominant purchases directly from the pro-
ducer of the inputs. On the sales side, however, firms may lack the necessary information
on the spatial destination of their sales only if they primarily sell through wholesale firms.
But even then, they appear to be better informed about their sales than about their
purchases through wholesale firms.
We will not deal with regional sector-specific models. Although many exist, the issues
are similar to those for regional models.4
432 Handbook of regional growth and development theories

Multi-regional and interregional IO tables and applications

Since the 1960s when Isard (1960) published his text on regional analysis, many IO
analysts have considered that interregional input–output (IRIO) or multi-regional
input–output (MRIO) tables are needed to derive and compare the direct and indirect
economic impacts across regions. As globalization of the markets increases, the need for
such tables by analysts increases. In the 1960s, probably as a result of interchanges with
the Japanese, who were gathering extensive IO data, Isard set forth an IRIO set of
accounts in which each IO coefficient was specified in terms of the region and sector in
which the input originated and the region and sector in which it terminated. The Japanese
used that interregional accounting framework to construct the first IRIO set of accounts
for nine regions and ten commodities for the 1960 and 1963 Japanese economy.
Leontief felt that such detail was not theoretically justified for the technological inputs,
because when an engineer makes a widget, the region in which the input was produced
should not matter. In his national–regional model, therefore, he did not specify the region
of origin of any input (Polenske, 1999). Both Isard and Leontief were correct, in their own
way. On the one hand, for determining the technology (production function) of a sector
in a particular region an engineer or economist does not need to know the region from
which the inputs come. The analyst only needs to know the production technology being
used in each region for each sector. Leontief’s national–regional and MRIO accounts
(Leontief and Strout, 1963; Polenske, 1970, 1980, 1995) therefore require considerably less
information, thus cost less and take less time to construct than Isard’s IRIO accounts. On
the other hand, from a transportation planner’s perspective, the regional origin of the
input is critical for transportation planning, as well as for tracing the supply chains of
commodities (Polenske and Hewings, 2004).
A major innovation with the latest MRIO accounts (Canning and Polenske, forthcom-
ing) is that the data will not only represent actual data from the census, as the first US
MRIO accounts did,5 but the data are being collected from electronic files at the state and
county levels, using algorithms that will make it relatively easy to construct accounts in
the future, thus greatly reducing the time required to construct these accounts. Special
attention is being paid to estimating the suppressed data from the census. Most impor-
tantly, the data will be freely available. This is an important consideration given that the
data in most other models are available only by paying vendors in conjunction with buying
a model.
We note that several US groups, such as IMPLAN (Impact Analysis for PLANning),
REDYN (Regional Dynamics), and NIEMO (National Interstate Economic Model) are
constructing or have constructed MRIO accounts, but as far as we can determine, they
are not making their data freely accessible. IMPLAN has been compared with REMI
(Regional Economic Models, Inc.) and with RIMS-II (Regional Input–output
Multipliers, version II) by a number of analysts, with the finding that after adjusting for
differences in the models, the multipliers are relatively similar. Richman and Schwer
(1993, p. 143) for example found ‘that apparent changes in the multipliers in each model
result from undocumented or poorly documented changes in the vendor default values of
the available options for calculating the multipliers, not from structural changes in the
models’.
According to the REDYN vendors, their model is ‘more flexible and versatile’ than
other commercially available models, but so far no independent comparative evaluation
 Modern regional input–output and impact analyses 433

is available. The REDYN model uses the North American Industrial Classification
System at the five-digit detail level (703 sectors), identifies the more than 180 commodi-
ties consumed and produced by these sectors, and provides forecasts for over 800 occu-
pations. They obtain the underlying data with the use of County Business Patterns from
the US Bureau of the Census, and the Regional Economic Information System from the
US Bureau of Economic Analysis. The trade flows are provided for five transportation
modes at the county level for 3100 counties. Former REMI staff are constructing
REDYN, with the intent to provide on-line capabilities with up-to-date information.
NIEMO is a combined supply–demand-driven IO model developed by analysts at the
University of Southern California’s National Center for Metropolitan Transportation
Research and the Center for Risk and Economic Analysis of Terrorist Events, as part of
their research at the Homeland Security Center for Excellence. They constructed the 47-
sector model for 52 regions (50 states, Washington, DC, and Rest-of-the-World). The the-
oretical nature of the iteratively solved model is not entirely clear. One of its main uses so
far has been to determine the economic impacts on these regions resulting from a hypo-
thetical terrorist attack on the three major US ports of Long Beach, Newark and Houston
(Park et al., 2007) They made use of IMPLAN for the technology part of the model and
developed interregional shipments for the 47 sectors, using the doubly constrained Fratar
model (Richardson et al., 2005).
The development of several competitive multi-regional IO models is a sign that such
analytical impact-assessment tools with real data are in great demand.
Japan is the main country with full interregional IO accounts, which are available for
1960 and every five years since, for 42 prefectures, as well as originally for nine regions,
for ten sectors (Abe, 1986). They have set the standard for such detailed assembly of data.
Currently, China has MRIO accounts for 1987 constructed over a five-year period in col-
laboration with the Japanese for seven regions and nine sectors (Ichimura and Wang,
2003; Okamoto and Ihara, 2005). Analysts are using these tables to examine many impor-
tant regional topics in the rapidly growing economy of China, including factors creating
the regional differentials in income. Although the types of studies completed so far are
typical economic impact assessments, the availability of improved and detailed current
regional data in the near future will provide important possibilities for new analyses on
energy, environment, transportation, foreign trade and other current topics.
Since 1995, several countries, such as Canada (Siddiqi and Salem, 1995), the
Netherlands (Eding et al., 1999) and Finland (Piispala, 2000), have embarked upon the
construction of multi-regional supply and use (commodity-by-industry) tables. Note that
not all of these rectangular accounting schemes have a straightforward one-to-one rela-
tion with the symmetric IRIO and MRIO models discussed in section 21.2 (Oosterhaven,
1984).

21.5 Modern impact analysis

Today, many analysts are further developing social accounting models (SAMs) and inter-
sectoral computable general equilibrium (CGE) models, and they extensively use such
models for economic impact analyses. Regional analysts are conducting numerous eco-
nomic impact analyses at multi-regional levels in the United States, most of which use one
of three computer programs available from commercial vendors: REMI, IMPLAN and
RIMS-II, with the matrix multipliers furnished by the US Bureau of Economic Analysis.
434 Handbook of regional growth and development theories

REMI is an eclectic multi-regional model that combines economic base, input–output,

computable general equilibrium, and econometric methods, and was first put forward for
Massachusetts by Treyz et al. (1980). ECOTEC REMI is the daughter model for the
United Kingdom, and REMI-NEI is the daughter model for the Netherlands. REMI is a
sophisticated regional model with numerous policy variables. Analysts use it to determine
environmental pollution impacts, regional impacts of transport infrastructure, including
airports, and studies of many large investment projects. Latest versions of REMI include
job-accessibility measures and spatial-agglomeration effects.
Mourouzi-Sivitanidou and Polenske (1988) conducted one of the earliest evaluations
and comparison of these and other impact-assessment models.6 One of the most exten-
sive and longest uses of REMI is in Los Angeles by the South Coast Air Quality
Management District (SCAQMD), which has used it since 1989 for the determination of
job impacts of its rules and regulations, and since 1994 for its tradeable permit Regional
Clean Air Management program. The SCAQMD uses it to study job impacts in a four-
county (Los Angeles, Orange, Riverside, San Bernardino) region in southern California.
Polenske et al. (1992) conducted an extensive 13-month evaluation of the use of REMI
by the SCAQMD, determining that as of 1992, the SCAQMD had state-of-the-art impact
assessments with the use of REMI. The team provided over 30 recommendations for pos-
sible improvements both to the model and to the other evaluation methods being used,
many of which have been implemented.
In addition, other readily available regional modeling packages for the United States
include IMPLAN and RIMS-II. Each of these packages costs money, with RIMS-II
being the least expensive and REMI the most expensive. REMI, however, is the only
regional economic model that analysts can use for forecasting over 10–20 years. IMPLAN
and RIMS-II have more sectors than REMI, but they can only be used for comparative-
static impact assessments.
IMPLAN was started in the 1970s in a combined effort by Lofting at the Berkley
Lawrence Livermore National Laboratory and Alward at the US Forestry Service. In
1993, the Minnesota IMPLAN Group was founded by Lindall and Olson, based upon
work at the University of Minnesota. At the forest service starting in the 1970s, Alward
and others developed Micro-IMPLAN, an input–output impact model specifically
designed to meet the US Department of Agriculture (USDA) Forest Service needs.
Results for four years (1990, 1985, 1982 and 1977) are reported by Shields et al. (1995).7
The major characteristics of IMPLAN Version I, which is no longer used, were: (1) indus-
try-by-industry accounting; and (2) supply–demand pooling technique for trade estima-
tions. IMPLAN Version II is the basis for all subsequent IMPLAN software, including
the current (2006) release of MicroIMPLAN Rel 91–F. Its principal characteristics are:
(1) rectangular accounting (that is, make-and-use matrices); and (2) regional purchase
coefficients (RPCs) for trade estimation.
IMPLAN differs from REMI in several ways. First and perhaps most importantly,
IMPLAN is a static IO impact model, based on the latest national IO table (as of 2008
for the year 2002). In contrast, REMI is a dynamic input–output, econometric forecast-
ing and simulation model, in which Bureau of Labor Statistics forecasts of the input
coefficients are used to obtain regional technology forecasts for up to 20 years. The 2008
version of IMPLAN contains 400+ sectors, while REMI has only 57. Second, IMPLAN
initially used the 1963 regional IO data assembled in a manner similar to the national IO
 Modern regional input–output and impact analyses 435

data for the 50 states and Washington, DC, by Polenske (1980). When these regional tech-
nologies became more and more dated, they switched to using location quotients (LQs)
to adjust the national IO coefficients, and currently they are using regional purchase
coefficients (RPCs) estimated from trade flows. Starting from the initial REMI model,
Treyz recognized the superiority of using RPCs over LQs.
Although adjustments using RPCs create more accurate estimates than LQs and are
relatively inexpensive, some analysts, including the current authors, believe that the best
method is to assemble US regional IO tables using the same data sources the Bureau of
Economic Analysis uses for the national data. That is the thrust of the MITERS research,
in which staff from the Massachusetts Institute of Technology and the Economic
Research Service of the US Department of Agriculture are constructing multi-regional
accounts at a county level and for approximately 200 sectors. They use GAMS (General
Algebraic Modeling System) to create input files from census and other data to help
estimate suppressed data.
Several analysts are designing new multi-regional program packages in attempts to
make the models suitable for use in modern impact assessments, such as for distribution
of agricultural goods, terrorist attacks, air transportation impact, and so on. Examples
include Lahr who redesigned Stevens’s PC-IO package, and the University of Southern
California (USC) staff who designed the NIEMO model discussed above. These multi-
regional models are used by numerous state and county groups, and academics.
An important extension of the direct, indirect and induced impact analysis is determi-
nation of so-called catalytic effects. For air transportation, Oxford Economic Forecasting
distinguishes between supply-side and demand-side catalytic effects, with the supply-side
effects indicating the performance of the economy and long-run productivity and livabil-
ity, and the demand-side effects including the use of air services to transport goods,
business travelers and tourists (Cooper and Smith, 2005, p. 16).
In Europe, the need to evaluate a series of large transport infrastructure projects led to
IO-type new economic geography (NEG) models (Venables, 1996). Because freight and
passenger transport cost reductions impact upon different sectors differently, analysts use
an interregional inter-industry approach. Different regions sell varieties of the output of
each sector on monopolistically competitive regional markets, linked by transport cost.
Sectoral CES-aggregates of these varieties are combined in Cobb–Douglas consumption
and production functions (in the latter case also with capital and labor). In these inter-
industry NEG models, transport cost reductions increase each region’s exports (demand)
as well as imports (supply). The net economic impact may well be negative for some
sectors in some regions, while causing clustering of sectors and agglomeration of eco-
nomic activity in other regions (see Venables and Gasiorek, 1996; Oosterhaven and
Knaap, 2003; Thissen, 2005, for seminal applications).

21.6 Conclusion
Regional analysts are incorporating new theoretical perspectives related to economic
geography into their socio-economic impact models. We have reviewed the basic Types I,
II and III input–output quantity and price models and summarized the state of the art in
regional and interregional impact analysis. From this, we conclude that the future will
continue to feature interregional inter-industry models, as the sector-specific and loca-
tion-specific nature of the employment, energy and emissions impacts of all kinds of
436 Handbook of regional growth and development theories

exogenous shocks and policy measures requires such modeling. Increasingly, however,
these models will feature non-linear production and consumption functions, and integrate
simultaneous price and quantity impacts in models with non-perfect competition.
Obviously, the collection and processing of regional technology and regional trade data
remain a time-consuming and expensive task. Given the need for regional and interre-
gional impact analyses, in the United States several different groups are constructing
multi-regional models. Because most of this work is unpublished at the present time
(2008), we have presented information gleaned from websites, personal conversations and
personal involvement. From this overview, we conclude that new spatial theories are
needed before analysts can make significant advances on applications and that there is no
real substitute to survey-based, inter-industry, interregional information and modeling.

Notes
1. The InterRegional Input–Output Software package IRIOS (Stelder et al., 2000), which is based on a flexible
generalization of impact equation (21.6), is freely downloadable at www.REGroningen.nl/irios.
2. Note that de Mesnard (2006) takes offense at the use of the word ‘multiplier’ in this case and proposes an
alternative net multiplier definition. See Oosterhaven (2007) for a reply and Dietzenbacher (2005) for an
independent evaluation.
3. The first Washington State table was assembled in 1967 by Tiebout, Bourque, Thomas, and other faculty at
the University of Washington, and faculty at Washington State University in Pullman, who assembled the
agricultural components. Washington State tables are available from surveys conducted for 1963, 1972,
1982, 1987, 1997, and 2002 (as of 2008, the 2002 one is just being completed). Two 1997 tables were con-
structed, one with industrial sectors defined by the Standard Industrial Classification (SIC), the other by
the North American Industrial Classification System (NAICS), with employment, income and output mul-
tipliers for each classification scheme for 38 and 62 sectors (Illman, 1996; State of Washington, 2004).
4. One example is the Port Economic Impact Kit (Klaers, Powers & Associates, 2001). It has a customized
national IO model for the port’s region. The direct economic impacts include those generated by the tran-
shipment of cargo as well as capital investments made by waterfront facilities. The indirect and induced
impacts are obtained from the purchases required by this direct expenditure and by the wages paid to the
labourers in direct and indirect activities. Estimates are also made of the property taxes and occupation taxes
paid by the various facilities.
5. The United States has 1963, 1972 and 1977 MRIO accounts for 51 regions (50 states and Washington, DC)
and 79 to 120 sectors (Polenske, 1980), and will have by early 2009 1997 and 2002 MRIO accounts for 3076
counties plus about 500 ports of entry and 162 sectors; all the latter data for all years are constructed from
census and other official US data, mostly using electronic files (Canning and Wang, 2005; Canning et al.,
2007).
6. The website: http://www.remi.com/support/articlescomplete.shtmllist) lists an additional 23 evaluations.
7. See for details on the early IMPLAN research: http://www.fpl.fs.fed.us/documnts/fplgtr/fplgtr 95.pdf.

References
Abe, K. (1986), ‘Input–output tables in Japan and application for interregional analysis’, presented at the 8th
International Input–Output Conference, Sapporo, Japan.
Batey, P.W.J. (1985), ‘Input–output models for regional demographic-economic analysis: some structural com-
parisons’, Environment and Planning A, 17, 77–93.
Batten, D.F. (1983), Spatial Analysis of Interacting Economies, Boston, MA: Kluwer Publishing.
Beemiller, R.M. (1990), ‘Improving accuracy by combining primary data with RIMS: comment on Bourque’,
International Regional Science Review, 13 (1–2), 99–101.
Boomsma, P. and J. Oosterhaven (1992), ‘A double-entry method for the construction of bi-regional
input–output tables’, Journal of Regional Science, 32 (3), 269–84.
Bourque, P.J. (1990), ‘Regional multipliers: WAIO vs. RIMS’, International Regional Science Review, 13 (1–2),
87–98.
Canning, P. and K. Polenske (forthcoming), US 1992 and 2002 Multiregional Input–Output Accounts, US
Department of Agriculture, Economic Research Service.
Canning, P. and Z. Wang (2005), ‘A flexible mathematical programming model to estimate interregional
input–output accounts’, Journal of Regional Science, 45 (3), 539–63.
 Modern regional input–output and impact analyses 437

Canning, P., A. Ismail, K.R. Polenske, S. Huang and A. Waters (2007), ‘US energy consumption and food secu-
rity’, presented 21 October at the Associated Collegiate Schools of Planning Conference, Milwaukee, WI.
Carter, A.P. (1970), Structural Change in the American Economy, Cambridge, MA: Harvard University Press.
Cartwright, J.V., R.M. Beemiller, E.A. Trott and J.M. Younger (1982), ‘Estimating the potential impacts of a
nuclear reactor accident’, Bureau of Economic Analysis, Washington, DC.
Chen, C.Y. and A. Rose (1985), ‘The joint stability of input–output production and allocation coefficients’,
Modeling and Simulation, 17, 251–5.
Chenery, H.B. (1953), ‘Regional analysis’, in H.B. Chenery, P.G. Clark and V.C. Vera (eds), The Structure and
Growth of the Italian Economy, Rome: US Mutual Security Agency, pp. 97–129.
Cole, S. (1989), ‘Expenditure lags in impact analysis’, Regional Studies, 23 (2), 105–16.
Cole, S. (1997), ‘Closure in Cole’s reformulated Leontief model: a response to R.W. Jackson, M. Madden and
H.A. Bowman’, Papers in Regional Science, 76 (1), 29–42.
Cooper, A. and P. Smith (2005), ‘The economic catalytic effects of air transport in Europe’, Report prepared by
Oxford Economic Forecasting, September.
Davar, E. (1989), ‘Input–output and general equilibrium’, Economic Systems Research, 1 (3), 331–44.
Davis, H.C. and E.L. Salkin (1984), ‘Alternative approaches to the estimation on economic impacts resulting
from supply constraints’, Annals of Regional Science, 18, 25–34.
de Mesnard, L. (2006), ‘A critical comment on Oosterhaven–Stelder net multipliers’, Annals of Regional Science,
41 (2), 249–71.
Dietzenbacher, E. (1997), ‘In vindication of the Ghosh model: a reinterpretation as a price model’, Journal of
Regional Science, 37 (4), 629–51.
Dietzenbacher, E. (2005), ‘More on multipliers’, Journal of Regional Science, 45, 421–6.
Eding, G.J., J. Oosterhaven, B. de Vet and H. Nijmeijer (1999), ‘Constructing regional supply and use tables:
Dutch experiences’, in G.J.D. Hewings, M. Sonis, M. Madden and Y. Kimura (eds) Understanding and
Interpreting Economic Structure, Berlin: Springer Verlag, 237–63.
Ghosh, A. (1958), ‘Input–output approach in an allocation system’, Economica, 25, 58–64.
Giarratani, F. (1974), ‘The effect on relative prices of air pollution abatement: a regional input–output simula-
tion’, Modeling and Simulations, 5, 165–70.
Giarratani, F. (1976), ‘Application of an interindustry supply model to energy issues’, Environment and Planning
A, 8, 447–54.
Hirsch, W.Z. (1959), ‘Interindustry relations of a metropolitan area’, Review of Economics and Statistics, 41,
360–69.
Ichimura, S. and H. Wang (2003), Interregional Input–Output Analysis of the Chinese Economy, Singapore:
World Scientific Publishing Co.
Illman, D.L. (1996), ‘A century of excellence in science and technology at the University of Washington’,
University of Washington, www.washington.edu/research/pathbreakers/index.html#1965.
Isard, W. (1951), ‘Interregional and regional input–output analysis: a model of the space economy’, Review of
Economics and Statistics, 33, 318–28.
Isard, W. (1960), Methods of Regional Analysis: An Introduction to Regional Science, Cambridge, MA: MIT Press.
Jackson, R.W., M. Madden and H.A. Bowman (1997), ‘Closure in Cole’s reformulated Leontief model’, Papers
in Regional Science, 76 (1), 21–8.
Klaers, Powers & Associates (2001), ‘Economic impact of 2001 shipping season’, Duluth Seaway Port Authority,
Port of Duluth-Superior, www.duluthport.com/dsweconomicimpact2001.html.
Kop Jansen, P.S.M. and T. ten Raa (1990), ‘The choice of model in the construction of input–output coefficients
matrices’, International Economic Review, 31, 213–27.
Lahr, M. (1993), ‘A review of the literature supporting the hybrid approach to constructing regional
input–output models’, Economic Systems Research, 5 (3), 277–93.
Leontief, W.W. (1951), The Structure of the American Economy: 1919–1939, New York: Oxford University Press.
Leontief, W.W. and A. Strout (1963), ‘Multiregional input–output analysis’, in T. Bama (ed.), Structural
Interdependence and Economic Development, New York: St Martin’s Press, pp. 119–50.
Madsen, B. and C. Jensen-Butler (2005), ‘Spatial accounting methods and the construction of spatial social
accounting matrices’, Economic Systems Research, 17 (2), 187–210.
Miernyk, W.H. (1965), The Elements of Input–Output Analysis, New York: Random House.
Miller, R.E. and P.D. Blair (1985), Input–Output Analysis: Foundations and Extensions, Englewood Cliffs, NJ:
Prentice Hall.
Miyazawa, K. (1976), Input–Output Analysis and the Structure of the Income Distribution, Berlin: Springer.
Miyazawa, K. and S. Masegi (1963), ‘Interindustry analysis and the structure of income distribution’,
Metroeconomica, 15, 89–103.
Moore, F.T. (1955), ‘Regional economic reaction paths’, American Economic Review, 45 (2), 133–48.
Moore, F.T. and J.M. Peterson (1955), ‘Regional analysis: an interindustry model of Utah’, Review of Economics
and Statistics, 37, 368–83.
438 Handbook of regional growth and development theories

Moses, L.N. (1955), ‘The stability of interregional trading pattern and input–output analysis’, American
Economic Review, 45 (5), 803–32.
Mourouzi-Sivitanidou, R. and K.R. Polenske (1988), ‘Assessing regional economic impacts with microcom-
puters’, in R.E. Klosterman (ed.), A Planners Review of PC Software and Technology, Planning Advisory
Service, Report Nos. 414/415, Chicago, IL: American Planning Association, pp. 101–6.
Okamoto, N. and T. Ihara (2005), ‘Spatial structure and regional development in China: interregional
input–output approach’, Institute of Developing Economies, Japan External Trade Organization, IDE
Development Perspective Series No. 5, Japan.
Oosterhaven, J. (1981a), Interregional Input–Output Analysis and Dutch Regional Policy Problems, Aldershot:
Gower Publishing.
Oosterhaven, J. (1981b), ‘Export stagnation and import price inflation in an interregional input–output model’,
in W. Buhr and P. Friedrich (eds), Regional Development under Stagnation, Baden-Baden: Nomos-Verlag,
pp. 124–48.
Oosterhaven, J. (1984), ‘A family of square and rectangular interregional input–output tables and models’,
Regional Science and Urban Economics, 14, 565–82.
Oosterhaven, J. (1988), ‘On the plausibility of the supply-driven input–output model’, Journal of Regional
Science, 28 (2), 203–17.
Oosterhaven, J. (1996), ‘Leontief versus Ghoshian price and quantity models’, Southern Economic Journal, 62
(3), 750–59.
Oosterhaven, J. (2000), ‘Lessons from the debate on Cole’s model closure’, Papers in Regional Science, 79 (2),
233–42.
Oosterhaven, J. (2007), ‘The net multiplier is a new key sector indicator: reply to De Mesnard’s comment’, Annals
of Regional Science, 41 (2), 249–71.
Oosterhaven, J. and H. Folmer (1985), ‘An interregional labour market model incorporating vacancy chains and
social security’, Papers of the Regional Science Association, 58, 141–55.
Oosterhaven, J. and T. Knaap (2003), ‘Spatial economic impacts of transport infrastructure investments’, in A.
Pearman, P. Mackie and J. Nellthorp (eds), Transport Projects, Programmes and Policies: Evaluation Needs
and Capabilities, Aldershot: Ashgate, pp. 87–105.
Oosterhaven, J. and T.M. Stelder (2002), ‘Net multipliers avoid exaggerating impacts: with a bi-regional illus-
tration for the Dutch transportation sector’, Journal of Regional Science, 42 (3), 533–43.
Oosterhaven, J., E.C. van der Knijff and G.J. Eding (2003), ‘Estimating interregional economic impacts: an eval-
uation of nonsurvey, semisurvey, and full-survey methods’, Environment and Planning A, 35 (1), 5–18.
Park, J.Y., P. Gordon, J.E. Moore II and H.W. Richardson (2007), ‘Simulating the state-by-state effects of ter-
rorist attacks on three major US ports: applying NIEMO’, in H.W. Richardson, P. Gordon and J.E. Moore
II (eds), The Economic Costs and Consequences of Terrorism, Cheltenham, UK and Northampton MA, USA:
Edward Elgar, pp. 208–34.
Piispala, J. (2000), ‘On regionalising input–output tables: experiences from compiling regional supply and use
tables in Finland’, presented at the 13th International Input–Output Conference, 21–25 August, Macerata,
Italy.
Polenske, K.R. (1970), ‘An empirical test of interregional input–output models: estimate of 1963 Japanese pro-
duction’, American Economic Review, 60, 76–82.
Polenske, K.R. (1980), The US Multiregional Input–Output Accounts and Model, Lexington, MA: Lexington
Books.
Polenske, K.R. (1995), ‘Leontief’s spatial economic analyses’, Structural Change and Economic Dynamics, 6,
309–18.
Polenske, K.R. (1997), ‘Current uses of the RAS technique: a critical review’, in A. Simonovits and A.E. Steenge
(eds), Prices, Growth, and Cycles, London: Macmillan Press, pp. 58–88.
Polenske, K.R. (1999), ‘Wassily W. Leontief, 1905–1999’, Economic Systems Research, 11, 341–8.
Polenske, K.R. and G.J.D. Hewings (2004), ‘Trade and spatial economic interdependence’, Papers in Regional
Science, 83 (1), 269–89.
Polenske, K.R., K. Robinson, Y.H. Hong, X. Lin, J. Moore and B. Stedman (1992), ‘Evaluation of the South
Coast Air Quality Management District’s methods for assessing socioeconomic impacts of district rules and
regulations’, prepared for the South Coast Air Quality Management District, Diamond Bar, CA.
Pyatt, G. (2001), ‘Some early multiplier models and the relationship between income distribution and produc-
tion structure’, Economic Systems Research, 13 (2), 139–63.
Pyatt, G. and J.I. Round (1979), ‘Accounting and fixed price multipliers in a social accounting matrix frame-
work’, Economic Journal, 89, 850–73.
Pyatt, G. and E. Thorbecke (1976), Planning Techniques for a Better Future, Geneva: International Labour
Office.
Richardson, H.W., P. Gordon and J.E. Moore II (eds) (2005), The Economic Impacts of Terrorist Attacks,
Cheltenham, UK, and Northampton, MA, USA: Edward Elgar.
 Modern regional input–output and impact analyses 439

Richman, D.S. and R.K. Schwer (1993), ‘A systematic comparison of the REMI and IMPLAN models: the case
of southern Nevada’, Review of Regional Studies, 23 (2), 143–161.
Schumann, J. (1968), Input–Output Analyses, Berlin: Springer Verlag.
Shields, D.J., S.A. Winter, G.S. Alward and K.L. Hartung (1995), ‘Energy and minerals industries in national,
regional, and state economies’, US Department of Agriculture, Forest Products Laboratory, General
Technical Report, FPL–GTR–95 Washington, DC, www.fpl.fs.fed.us/documnts/fplgtr/fplgtr 95.pdf.
Siddiqi, Y.M. and M. Salem (1995), ‘Regionalization of commodity-by-industry input–output accounts: the
Canadian case’, Ottawa: Statistics Canada.
Spijker, J. (1985), ‘Een combinatie van directe en indirecte methoden’, in J. Oosterhaven and B.B.A. Drewes
(eds), Constructie en actualisering van regionale en interregionale input-output tabellen, Arnhem: Economisch-
Technologische Instituten, pp. 171–8.
State of Washington (2004), ‘The 1997 Washington Input–Output Table’, Office of Financial Management,
State of Washington, www.ofm.wa.gov/economy/io.
Stelder, T.M., J. Oosterhaven and G.J. Eding (2000), ‘Interregional input–output software’, IRIOS 1.0 Manual,
University of Groningen, www.REGroningen.nl.
Stevens, B.H. and G.I. Treyz (1986), ‘A multiregional model forecast for the United States through 1995’,
American Economic Review, 76 (2) (May), 304–7.
Strout, A.M. (1967), ‘Technological change and the United States energy consumption, 1939–1954’, PhD
Dissertation, University of Chicago.
ten Raa, T. and J.M. Rueda-Cantuche (2003), ‘The construction of input–output coefficients matrices in an
axiomatic context: some further considerations’, Economic Systems Research, 15 (4), 441–55.
Thissen, M. (2005), ‘RAEM: regional applied general equilibrium model for the Netherlands’, in F. van Oort,
M. Thissen and L. van Wissen (eds), A Survey of Spatial Economic Planning Models in the Netherlands,
Rotterdam: Ruimtelijk Planbureau/NAi-uitgevers, pp. 64–86.
Treyz, G.I., A.F. Friedlaender and B.H. Stevens (1980), ‘The employment sector of a regional policy simulation
model’, Review of Economics and Statistics, 62 (1), 63–73.
Vaccara, B.N. (1969), ‘Changes over time in input–output coefficients for the United States’, in A.P. Carter and
A. Brody (eds), Applications of Input–Output Analysis, Amsterdam: North-Holland, pp. 238–60.
van Dijk, J. and J. Oosterhaven (1986), ‘Regional impacts of migrants’ expenditures: an input–output/vacancy-
chain approach’, in P.W.J. Batey and M. Madden (eds), Integrated Analysis of Regional Systems, London:
Pion, pp. 122–47.
Venables, A.J. (1996), ‘Equilibrium locations of vertically linked industries’, International Economic Review, 37
(2), 341–59.
Venables, A.J. and M. Gasiorek (1996), ‘Evaluating regional infrastructure: a computable equilibrium
approach’, mimeo, London School of Economics.
West, G.R. (1990), ‘Regional trade estimation: a hybrid approach’, International Regional Science Review, 13
(1–2), 103–18.