Geography and Intra-National Home Bias: U.S.
Domestic Trade in 1949 and 20071
Nick Crafts
Department of Economics, University of Warwick
n.crafts@warwick.ac.uk
and
Alexander Klein
School of Economics, University of Kent
a.klein-474@kent.ac.uk
Preliminary, Please, Do Not Cite
April 2012
Abstract
This paper examines home bias in the U.S. domestic trade in 1949 and 2007. We use a unique
data set of 1949 carload waybill statistics produced by the Interstate Commerce Commission,
and 2007 Commodity Flow Survey data. The results show that home bias was considerably
smaller in 1949 than in 2007 and that for several commodities, home bias in 1949 was even
negative. We argue that the difference between the geographical distribution of the
manufacturing activities in 1949 and that of 2007 is an important factor explaining the
differences in the magnitudes of home bias estimates in those years.
Keywords: intra-national home bias, spatial clustering, manufacturing belt, gravity equation
1
We are grateful to Tim Guinnane and Mark Thomas for their help with the U.S. Carload Waybill data, Natalie
Chen and Dennis Novy for generously sharing their data with us, and Anwita Basu and Mariela Dal Borgo for
their excellent research assistance.
1
Introduction
The effect of borders on trade has received a considerable attention in recent years. Since the
seminal paper by McCallum (1995), which showed that after controlling for numerous
explanatory factors trade between the Canadian provinces was about 22 times higher than
their trade with U.S. states, scholars have paid a close attention to the robustness as well as
explanation of border effect. One surprising result that has emerged from that literature is that
home bias does not feature only in international trade: domestic trade also exhibits a rather
substantial border effect (Wolf 2000, Nitsch 2000, Hillberry 2000, Chen 2004, Millimet and
Osang 2007, Coughlin and Novy 2011). For example, Wolf (2000) shows that U.S. intra-state
trade is considerably larger than the U.S. inter-state trade, and Coughlin and Novy (2011)
find that the U.S. intra-national border effect is actually higher than the U.S. international
border effect. Those results are indeed surprising since the U.S. domestic trade takes place in
a situation when the usual trade barriers such as tariffs, quotas, exchange rate variability do
not exist. This suggests that other explanations for home-bias need to be looked for.
Earlier studies investigating home-bias in domestic trade focused on the estimation and the
robustness of that finding; later studies have begun to investigate its possible causes. Several
explanations have emerged. Hillberry (2000), when examining the U.S. domestic trade, finds
that the spatial clustering of firms negatively affects the magnitude of the intra-national home
bias. Hillberry and Hummels (2003) show that much of the U.S. intra-national home bias can
be explained by highly localized wholesaling activity. In a follow up paper, Hillberry and
Hummels (2008) argue that the home bias effect is ‘non-linear’, meaning that while it holds
at 3-digit zip-code data, it disappears when using 5-digit zip codes. Furthermore, they find
that most establishments ship only to geographically proximate customers. Chen (2004)
investigates trade among the EU countries and shows that technical barriers to trade, product
2
specific information costs related to homogeneous and differentiated products, and spatial
clustering can significantly explain the intra-national home bias effect: technical barriers to
trade as well as the differentiated products increase the bias while industry concentration
reduces it.
It appears that the spatial concentration of economic activities might be an important factor
explaining home bias in domestic trade. This paper looks closely at that relationship.
Specifically, it examines home bias of U.S. domestic trade in 1949 and 2007 respectively by
estimating gravity regression at aggregate as well as commodity level. Doing so, it makes two
contributions. First, it examines the variation of home bias across commodities and time.
Second, analysing home bias in 1949 provides a unique opportunity to examine it at the times
when the geographical distribution of industries dramatically differed from that of 2007. If
the spatial distribution of industrial activities indeed matters, different patterns of industrial
location should have a different impact on the intra-national home-bias.
It is a well-known fact that the spatial distribution of economic activities across U.S. regions
underwent significant changes in the last century. The industrialization of the U.S. economy
in the second half of the nineteenth century brought about a divergence in the regional
specialization. In manufacturing, regions became highly specialized and by the turn of the
twentieth century, most of the manufacturing employment was concentrated in New England,
Middle Atlantic and East North Central regions, later labelled as the Manufacturing Belt
(Fritz 1943, Perloff 1990, Meyer 1983, 1989, Kim and Margo 2004, Holmes and Stevens
2004). This pattern was sustained until the 1940s, after which the degree of regional
specialization has been on decline (Kim 1995). Indeed, while in 1947, a little more than
seventy per cent of manufacturing employment was concentrated in the Manufacturing Belt,
it was only forty per cent in 1999 (Holmes and Stevens 2004). This dramatic decline of the
importance of the Manufacturing Belt went hand-in-hand with a rise of the importance of the
3
southern states such as Tennessee, Arkansas, Mississippi, and Texas, and the emergence of
the Sun Belt (Glaeser and Tobio 2008, Glaeser, Ponzetto, Tobio 2011). Overall, we can say
that the spatial distribution of economic activities evolved from the one of concentration in
the north-east at the turn of the twentieth century to the one of dispersion toward the south by
the end of the twentieth century. This patter was not, however, uniform across all industries.
As documented by Kim (1995), even though most of the industries exhibited that pattern,
there were those which became more regionally specialized such as tobacco or apparel, then
those which became more regionally dispersed, such as food, paper, or printing and
publishing, and finally those which changed their regional specialization pattern very little
such as furniture or primary metal.
This paper examines the implication of the changes in the geographical distribution of
manufacturing activities in the second half of the twentieth century for the U.S. domestic
trade of manufactures. It shows that the home of bias in 1949 was considerably smaller than
in 2007 (for some commodities even negative) and finds that it can be explained by looking
at the spatial distribution of industries. Specifically, the paper finds that the U.S. inter-state
trade in 1949 was more prevalent than in 2007 and that it was very likely connected to the
existence of the Manufacturing Belt. Once the Manufacturing belt dissolves and the industrial
activities move to the south, giving rise to the Sun Belt, intra-state trade gains on importance,
causing home-bias to increase. Domestic trade-flows are analysed with 2007 Commodity
Flow Survey data and a unique data source of railway trade-flows in 1949, compiled by the
Interstate Commerce Commission; spatial distribution of industries is captured by a version
of Ellison and Glaeser (1997) index due to Maurel and Sedillot (1999).
The paper proceeds as follows. Section 2 derives a gravity regression equation. Section 3
discusses the data sources, section 4 presents the regression results, section 5 discusses them,
and the last section concludes.
4
Section 2
This section presents a theoretical and empirical framework for estimating the home bias
effect. We follow an approach that is common in the home-bias literature: a gravity
regression. To derive the gravity regression equation, we use the widely adopted framework
due Anderson and Wincoop (2003).2 Let’s denote Xijk as the value of shipments of
commodity k at destination prices from origin i to destination j. Let tijk be the trade cost of
shipment of commodity k from i to j, Ejk denote expenditure on commodity k at destination j,
Yik the sales of commodity k at destination prices from i to all destinations. The resulting
gravity equation model is the following:
(
( )
)
(
)
∑ (
)
( )
(
)
∑ (
)
( )
The term Πik is called outward multilateral resistance, Pjk inward multilateral resistance, and
δk is the elasticity of substitution between varieties of commodity k. If we have the data in
physical quantities (for example metric tons), we need to adjust the equation (1) as follows:
( )
(
2
( )
)
The exposition follows Anderson and Yotov (2010).
5
where Zijk is the volume of export in physical quantities, pik is the f.o.b. price of commodity k
at the origin and tijk is again the trade cost of shipment of commodity k.3 Expressing Zijk from
equation (4) and adding multiplicative error term εijk yields
(
( )
)
To complete the derivation of the gravity regression equation, we need to specify tijk.
Standard approach in the gravity literature is to relate the trade costs to a set of observables
such as distance, common language, and the presence of contiguous borders.4 Here we
specify the trade costs as follows:
( (
)
)
( )
where distance is the bilateral distance between trading partners, ownstate, capturing intrastate trade, is a dummy variable equal to one when i=j, and adjacent is a dummy variable
equal to one if i and j have a common border. Then, substituting (7) into (6) we get
(
)
( (
)
)
( )
The estimation of equation (11) presents several challenges. First, we need to take into
account unobserved multilateral resistance terms. We use exporter/importer fixed-effects
approach, as applied by number of authors, e.g. Hummels 1999, Rose and van Wincoop
2001, Hillberry and Hummels 2003, Coughlin and Novy 2011. Second, we need to address
high number of zero bilateral trade flows and heteroskedasticity. As was argued by Santos
Silva and Tenreyro 2006, standard log-linearized gravity equation is incompatible with zero
3
Here we follow Wolf (2009).
4
For a discussion, see Anderson and Wincoop (2004).
6
trade flow data and not accounting for heteroskedasticity leads to inconsistent estimates. To
address those issues, they propose Poisson pseudo-maximum likelihood (henceforth PPML)
estimation technique to estimate the gravity regression with the dependent variable in levels
rather than logs.
A few papers investigating intra-state home bias also control for the location of U.S. states
(e.g. Wolf 2000, Anderson et al. 2003, Miliment and Osag 2007). We do so as well and
following Wolf 2000, we use a remoteness measure to control for the location of states i and j
relative to all other states. Specifically, remoteness for exports from state i to state j is defined
as the GDP weighted average distance between state i and all other states but j:
( )
∑
Equation (8) then becomes
(
)
( (
)
)
(
)
We estimate equation 10 with PPML for 1949 and 2007 respectively where Zijk are physical
quantities shipped within and between the U.S. states.
Section 3
This paper uses two data sets: 1949 U.S. Interstate Commerce Commission carload waybill
statistics and 2007 U.S. Commodity Flow Survey (CFS). The carload waybill data were
collected by the Interstate Commerce Commission (ICC). They constitute a random sample
of all shipment on Class I railroads between the origin and the destination state. The ICC
collected data on the quantities shipped as well as the number of shipments for five
commodity groups: products of agriculture, products of forest, animals, products of mines,
7
and manufactures and miscellaneous products. We have obtained the commodity level data
for the category of manufactures and miscellaneous products and the data set includes thirty
nine products from every SIC 2 category.5 The CFS is collected by the Census Bureau on
behalf of the U.S. Department of Transportation and it’s a survey of shipments from origin to
destination of manufacturing, mining, wholesale trade and selected retail establishments. The
shipments were collected for eight single-modes and five multiple-modes of transportation.6
The survey excludes shipments in services, crude petroleum and natural gas extraction, farm,
forestry, fishery, construction, government, and most of the retail sector. We have obtained
the data for forty one commodity classes, but we have excluded the agricultural products and
animals to be comparable with 1949 carload waybill data.7 The CFL records the value of
shipment as well as its weight in tons.
Other data used in this paper includes the U.S. states’ GDPs, total personal incomes, intraand inter-state distances and geographical concentration indices. The GDP of U.S. states in
2007 as well as their total personal income in 1949 are from the Bureau of Economic
Analysis (BEA). The distance between the U.S. states is calculated using standard great circle
5
The commodities include: food products, candy and confectionery, cigarettes, cloth and fabric, wood ware,
wooden containers, wood pulp, furniture, furniture parts, newsprint paper, paper articles, chemicals, acids,
fertilizers, matches, gasoline, crude rubber, cellulose articles, plastics, tires and tubes and rubber, boots and
shoes, cement Portland, common bricks, lime, aluminium bar, copper ingot, copper and brass and bronze, iron
and steel, manufactured iron and steel, agricultural implements, agricultural implements parts, machinery,
machinery parts, electrical equipment, refrigerators, vehicles, passenger automobiles, vehicle parts, airplanes,
hardware.
6
Single modes include for-hire truck, private truck, rail, shallow draft, the Great Lakes, deep draft, air (includes
air and truck), and pipelines; multiple-modes include parcel, truck and rail, truck and water, rail and water, and
other.
7
The commodity classes include: live animals and live fish, cereal grains, other agricultural products, animal
feed and products, meat and fish, grains and alcohol and tobacco products, other foodstuff, alcoholic beverages,
tobacco products, calcareous monumental or building stone, natural sands, gravel and crushed stones, nonmetallic minerals, metallic ores and concentrates, nonagglomerated bituminous coal, gasoline and aviation fuel,
fuel oils, coal and petroleum products, basic chemicals, pharmaceutical products, fertilizers, chemical products
and preparations, plastic and rubber, logs and other wood in the rough, wood products, pulp and newsprint
paper, paper and paperboard articles, printed products, textiles and leather, non-metallic mineral products, base
meals, articles of base metal, machinery, electronic and electrical equipment, motorized and other vehicles,
transportation equipment, precision instruments and apparatus, furniture, miscellaneous manufactured products,
waste and scrap, mixed freight.
8
distance formula. As for intra-state distance, we use several measures: a distance between the
two largest cities in a state, a distance between the two largest cities in a state weighted by
their population as suggested by Wolf (2000), a measure suggested by Nitsch (2000) which
uses land area and, for 2007 only, a measure suggested by Hillberry and Hummels (2003)
which is based on the actual shipping distances calculated from the data on individual
establishments. We do so because previous research has shown that the magnitude of the
home-bias estimates can be influenced by the way the intra-state distance is measured
(Hillberry and Hummels 2003). To account for the geographical distribution of
manufacturing activities, we calculate indices of geographical concentration in 1939 from
1939 U.S. Census of Manufactures.
Before we proceed with the regression analysis, it is illustrative to present a few descriptive
statistics of the U.S. domestic trade. Table 1 shows the domestic trade by the modes of
transportation in 1948 and 2007 respectively. As we see in Panel A, the most prevalent
transportation mode in 2007 is trucking which accounts for more than seventy percent of the
value and the weight of shipments respectively and for more than forty percent of ton-miles.
Railway transport is a distant second most important mode based on the weight of shipment,
a close second based on ton-miles, and third based on the value of shipment. The reason that
railways seem to be almost as important as trucking in ton-miles but not in the value or
weight of shipment is that railways transport heavy goods over long-distances. The
distribution of domestic trade by the modes of transportation in 1948 is presented in Panel B.
Since we lack the data on the value and weight of shipment, Panel B contains information on
ton-miles only. Nevertheless, a comparison of ton-miles is still revealing and a picture that
emerges from that comparison is quite clear: railways were by far the most important mode
of transportation in 1948, accounting for more than sixty percent of all ton-miles while
9
trucking was at a distant fourth place with less than nine percent. Inland water ways and
transportation on the Great Lakes was the second most important mode.
Maps 1 and 2 show the inter-state U.S trade by the place of origin in 1949 and 2007
respectively expressed as a percentage from the total U.S. inter-state trade. We see that there
are notable differences: in 1949, most of the inter-state trade originated in the north-east
while by 2007 the origins spread toward the south-east and south-west. Indeed, in 1949, more
than 56% of the interstate trade originated in the Manufacturing Belt states while that share
dropped to about 35% by 2007.8 On the other hand, the south-east and south-west gained on
importance: while in 1949 only about 23% of inter-state trade originated in those regions, by
2007 that share raised to almost 50%.9 Tables 2 and 3 present two specific examples which
illustrate that shift. Table 2 shows the main U.S. states exporting motorized vehicles to other
U.S. states in 1949 and 2007 respectively. We see that in 1949, the main exporting states
were almost exclusively in the Manufacturing Belt, with an exception of the state of
California. That picture has changed by 2007: though the states of then becoming Rust Belt
are still among the main exporters of vehicles to other U.S. states, south-east states now
account for more than 20% of interstate vehicle trade. A similar picture arises from Table 3
which shows the U.S. states exporting machineries. Even though in 1949 the origin of
machinery export was more geographically spread than that of vehicles, most of the interstate trade still originated in the Manufacturing Belt. 2007, however, shows again a
significant shift toward the south.
The intra-state U.S. trade also shows interesting changes over time. This is visible in Table 4
which shows the summary statistics of the share of state’s intra- and inter-state trade on
8
The Manufacturing Belt states include Illinois, Indiana, Massachusetts, Michigan, Minnesota, New York, New
Jersey, Ohio, Pennsylvania, and Wisconsin.
9
South-east and south-west regions include the following states: Alabama, Arkansas, California, Florida,
Georgia, Kentucky, Louisiana, Mississippi, North Carolina, South Carolina, Tennessee, Texas, Virginia, West
Virginia.
10
state’s total domestic trade respectively. We see that, on average, the intra-state trade was
more prevalent in 2007 than in 1949. Similar picture emerges from Figures 1 and 2 which
present kernel distributions of the ratio of state’s intra-state trade on its total domestic trade in
1949 and 2007 respectively. We see that while the upper tails of the distributions look
similar, the lower tails differ: there are more states with low shares of intra-state trade on
their total domestic trade in 1949 than in 2007. This, again, indicates that by 2007 intra-state
trade was more prevalent than in 1949.
Our data sets, as already hinted by Tables 2 and 3, allow us to compare the domestic trade
patterns at the commodity level as well. A drawback is that the list of commodities in 1949 is
not exactly the same as the list in 2007. Nevertheless, we can still compare the intra-state
trade patters for the commodities which are highlighted as bold in Table 5 which presents the
share of intra-state trade on the total domestic trade for every commodity category in 1949
and 2007 respectively. The commodity categories are ordered by their share of intra-state
trade on the total domestic trade. The first look at that table reveals relatively high shares of
intra-state trade in 2007 compared to 1949, confirming the picture emerging from Figures 1
and 2. At the commodity level, we see that gasoline has the highest share of intra-state trade
in both years while the transportation equipment and motorized vehicles among the lowest.
Before we proceed further, one limitation of 1949 data warrants a discussion. As was
mentioned at the beginning of this section, 1949 trade data are based on the railway trade
only. Even though, as seen in Table 1, railways were the dominant mode of freight transport
at that time and the interstate highway system had not been built yet, trucking was a growing
industry (Barger 1951, Meyer et al. 1960). It served mostly local markets and was delivering
predominantly highly perishable goods such as livestock, poultry and dairy products. Our
data set, even though it contains the shipments of manufacturing and miscellaneous goods
11
only, does not capture shipments over short distances. We deal with this issue in the next
section.
Section 4
The gravity regression equation (10) is estimated at the aggregate as well as commodity level
for 1949 and 2007 respectively. To make the results comparable, the dependant variable is
the weight of shipments. We first present the results of the estimation at the aggregate level,
then at commodity level.
Empirical Results: Aggregate Level
Tables 6 and 7 report the PPML results for 1949 and 2007 respectively. 10 Since several
studies indicated that the estimates of the home-bias coefficients are influenced by the intrastate distance measure (Hillberry and Hummels 2003, Chen 2004, Coughlin and Novy 2011)
we estimated the equation (11) with four internal distance measures in 2007 and three in
1949. Furthermore, Table 7 shows the estimation results for 2007 in two parts. The first part,
presented in Panel A, provides the estimation results for all sectors for which the CFS
collected the shipment data; the second part, presented in Panel B, provides the estimates for
the manufacturing sector only. We do so to make the results comparable with 1949 ones
which could have been obtained for the manufacturing sector only.
Table 6 shows that the estimates of the distance variables are always negative and statistically
significant, thought the magnitude differs across different intra-state distance measure. The
estimated coefficients on the home bias and adjacent variables are positive and significant
10
We estimated the regression with and without the clustered standard errors. Even though the clustered
standard errors are a bit larger than non-clustered, the statistical significance of the estimated coefficients
remains unchanged. The presented results are without clustered standard errors.
12
across all specifications. The home-bias estimates show an interesting feature: while home
bias estimate in column I implies that the intra-state U.S. trade is not more prevalent than
inter-state trade (the magnitude of the estimated coefficient is 0.90), other intra-state distance
measures imply otherwise (their magnitudes are 1.29 and 1.29). The results in Table 7 show
that, overall, the signs and the statistical significance of the estimated coefficients are as
expected and similar to Table 6: the distance variable being negative and mostly significant,
and home bias and adjacent variables positive and always significant. In two cases, though,
the estimated distance coefficients of the distance variables are insignificant in Panel A: Wolf
internal distance measure and Hillberry and Hummels (2003) measure. The magnitudes of the
home-bias coefficients are all greater than one, implying that the intra-state trade in 2007 was
more prevalent than inter-state trade.
For the purpose of comparison, we estimated the gravity equation (10) with the weight of
shipment as the dependant variable. Our data sets, however, offer other measures of interand intra-state trade as well: number of shipments in 1949 and the value of shipments in
2007. To see whether the results in Tables 6 and 7 are robust to those measures, we reestimated the gravity equation (10) with those other measures and present the results in
Tables A2 and A3 in Appendix. We see that in both tables, the magnitude of the home-bias
estimates is lower than in Tables 6 and 7. Indeed, in 1949, the home-bias ranges from 0.67 to
1.06 as opposed to 0.90-1.29, and in 2007 from 1.28 to 1.78 as opposed to 2.01-2.51 for the
whole economy and from 1.43-1.90 as opposed to 1.67-2.28.
The magnitude of the home-bias coefficients deserves a closer attention. Table 7 shows that
the estimates of the home-bias coefficients range between 1.67 and 2.51. This means that the
intra-state U.S. trade is between 1.67 and 2.51 times higher than the interstate U.S. trade. The
estimated coefficients are, in general, higher in Panel A than Panel B, though the magnitudes
differ across specifications. The reason for that is very likely the commodity composition:
13
Panel A provides the estimates all commodities reported by CFS and which include hightransport costs commodities such as gravel and crushed stone, Panel B’s estimates are for the
manufacturing sector only. Another notable difference is between the magnitude of the homebias estimates in 1949 and 2007 (here we compare Table 6 with Panel B in Table 7). Indeed,
the size of the home-bias coefficient in 1949 is between 0.90 and 1.29 while it is between
1.67 and 2.28 in 2007. This indicates that in 2007, intra-state U.S. trade was much more
prevalent than in 1949. These results are in tune with the discussion in Section 3 in which we
have seen an increasing tendency toward intra-state trade since 1949 (Figures 1 and 2, Table
4).
The previous section has pointed out that the commodity flow data does not include
shipments over short distances. As a consequence, the estimates of home-bias might be
biased downwards. Since we show that the home-bias in 1949 is smaller than in 2007, we
need to establish whether it is not a result of a possible downward biasness of 1949 estimate.
To do this, we take advantage of the Interstate Commerce Commission study which estimates
the freight that railways lost to other modes of freight transportation in the period 1929-38.11
Specifically, the study calculates for all commodities an index of ‘potential tons’ that
railways would have carried and compares it with actual tons carried by railways. That
comparison yields a ratio of actual to potential tons in 1937 of 84.9 percent, which means that
railways lost 15.1 percent of their freight to other means of transport. We use that information
to estimate the home-bias in 1949 under the assumption that all of the lost freight was intrastate trade only. Specifically, we add 15.1 percent of 1949 railway freight to our carload
waybill data set such that the extra 15.1 percent are traded only within the U.S. states and not
across them. Then we estimate equation (10). This extreme assumption favours intra- over
11
It is a report by the Interstate Commerce Commission ‘Fluctuations in Railway freight Traffic Compared with
Production’, Statement 3951 from November 1939.
14
inter-state trade. Indeed, it is an extreme assumption because some of the lost shipments were
made across the U.S. states. Doing so we deliberately err on the side of home-biasness to see
how much the absence of shipments by trucks, which ship goods mostly locally, affects our
1949 home-bias estimate. If the estimated home-bias under that extreme assumption is still
lower than in 2007, then our arguments hold. Table 8 shows three different home-bias
estimates for each of the intra-state distance measure: lower bound 1949 is the estimate from
Table 6, 2007 estimate is from Table 7, Panel B, and upper bound 1949 is the estimate with
extra 15.1 percent of intra-state shipments. We see that even under the very extreme
assumptions made above 1949 upper bound estimates are still considerably smaller than 2007
ones.
Empirical Results: Commodity Level
Tables 9 and 10 present the results of estimating equation (10) at the commodity level for
1949 and 2007 respectively. For the clarity purposes, only the estimates of home-bias
coefficients are reported and, in 1949, the estimates are grouped according to their sign and
statistical significance.12 We start discussing the results for 2007 first before turning to those
of 1949. Table 10 shows that the estimates of the home-bias variable in 2007 are positive in
all and statistically significant in all but three cases. The magnitude of the estimates varies
across different intra-state distance measures, while the statistical significance is mostly
unchanged.13 Table A1 in Appendix ranks the magnitude of the estimated coefficients from
the smallest to the largest. Overall, the rank is relatively stable across intra-state distance
measure, though there are some exceptions. Specifically, ‘Metallic ores and concentrates’,
‘Logs and other wood in the rough’, and ‘Calcareous monumental or building stone’
12
The full set of results is available from the authors upon request.
13
An exception is ‘Calcareous monumental or building stone’ industry when the estimate is significant only in
one out of four cases.
15
industries show rather large changes of the magnitude of the home-bias estimates across
different intra-state distance measures.
Table 9 shows a very different picture. Unlike the estimates for 2007, the home bias estimates
for 1949 show a considerable variation in magnitude, statistical significance, as well as sign.
The sign of the estimated coefficients is the most distinctive difference between 1949 and
2007 estimates: there are many commodity groups with negative home-bias effect and several
of them are highly statistically significant. We also see that there is quite a bit of variation
across different intra-state distance measures, though we can identify commodity groups
which estimates do not change their statistical significance and sign. Those are highlighted as
bold in Table 9.
Commodities with the statistically significant and negative home-bias estimates across all
three different measures of the intra-state distance include ‘Copper Ingot’, ‘Copper, Brass,
Bronze’, ‘Automobiles’, ‘Vehicle parts’, ‘Hardware’, and ‘Airplanes’. Other commodities
which home-bias is negative for at least one of the intra-state distance measures are ‘Paper
Articles’, ‘Cigarettes’, ‘Agricultural Implements’, and ‘Agricultural Implements Parts’.
Commodities with the statistically significant but positive home-bias estimates across all
three intra-state distance measures include ‘Fertilizers’, ‘Gasoline’, ‘Boots and Shoes’,
‘Bricks’, and ‘Refrigerators’. Other commodities include ‘Cloth’, ‘Newsprint Papers’,
‘Acids’, ‘Rubber’, ‘Cement’, ‘Wood Pulp’, ‘Wooden Containers’.
Section 5
We have seen that the magnitude of the home bias differs substantially between 1949 and
2007 and especially the home-bias at the commodity-level exhibits considerable differences.
16
The literature has suggested several explanations of that variation, as was discussed in
Section 1. Here we pay a close attention to one of them: spatial concentration of industries.
Before we do that, let us first review the arguments and empirical results linking it with intranational home-bias.
The idea goes back to 1997 working paper version of Wolf (2000) paper. Wolf provides a
hypothesis that the ‘spatial comparative advantage’ is a possible explanation of the domestic
home-bias. He suggests that if spatial clusters occur within sectors, home-bias might be
observed because intra-sector trade of intermediate products might take places in the clusters
within states. Hillberry (2000) provides a comprehensive examination of the causes of homebias. Using commodity level data, he investigates the differences in legal and regulatory
environment, multinational activity, information flow, government purchases, past
transportation networks and geographical location of industries. His result shows that only
geographical location can significantly explain the variation of home-bias across
commodities. Specifically, using Ellison and Glaeser (1997) geographical concentration
index (henceforth EG), he finds a negative relationship between that index and the home-bias
estimates. This means that industries which are tied to specific locations exhibit low estimates
of home-bias while those which are not show large border effect. A study by Hillberry and
Hummels (2003) also alludes to the role of geography in explaining home-bias of intranational trade. Using commodity-flow survey data they show that the home-bias estimate
drops after excluding wholesale shipments which tend to be more localized.14 In another
study using the commodity-flow survey data (Hillberry and Hummels 2008) they find that the
location of intermediate demand explains the geographical pattern of U.S. trade. Chen (2004),
using trade flow data of 78 industries for the EU countries, explores the role of the spatial
14
The study also controls for multilateral resistance in the gravity regression as suggested by Anderson and van
Wincoop (2003).
17
concentration as well. She also uses EG index and estimates a gravity regression with an
interaction term between the home-bias dummy and EG index.15 Her results show a negative
relationship between the geographical concentration and the magnitude of home-bias which
lend support to the argument of Wolf (1997), and the findings of Hillberry (2000).
We also examine the relationship between the spatial distribution of industries and home-bias
by using a version of EG index developed by Maurel and Sedillot (1999) calculated for 1939.
There is a difference between Hillberry (1999) and Chen (2004) analysis and ours though.
They, by analysing home bias at the commodity level, attempt to understand how spatial
concentration affects the magnitude of home-bias estimates at a point in time. We, by
analysing home bias at the commodity level, try to understand why 1949 home-bias estimate
is considerably lower than that of 2007 and how it is linked to spatial concentration of
industries in 1949.
Let us first look at the commodity-level home bias estimates more closely. Table 8 and 9
clearly show that the estimates for 2007 are on average higher than those of 1949. In addition,
there are many negative home-bias estimates in 1949 unlike in 2007 when all estimates are
positive. The commodity level home bias estimates thus indicate why the magnitudes of the
home-bias estimates in 1949 and 2007 are in the aggregate regressions so different from each
other: low values of home-bias estimates and especially negative ones pull down the overall
home-bias estimates in 1949 relative to 2007. Therefore, looking at the causes of the
negativity of the home-bias estimates in 1949 might help us to understand why the home-bias
estimates are so low in 1949 relative to 2007, hence why the intra-state trade was not as so
prevalent in 1949 as in 2007.
15
Chen (2004) also explores the role of technical barriers to trade, and product-specific information costs.
18
Negative sign of the home-bias estimates implies a strong preference of a U.S. state for
trading with other U.S. states than with itself.16 This suggests that the commodities which
exhibit negative home-bias are heavily exported from a few U.S. states to other U.S. states.
Our earlier discussion of 1949 trade flows in Section 3 has already hinted that direction.
Indeed, we have seen in Tables 2 and 3 that the inter-state trade of motorized vehicles and
machineries originates mostly in the Manufacturing Belt. Therefore, a question arises: is there
any link between the sign of the home-bias estimate and spatial concentration of industries in
1949? To answer it, we begin with Table 9 which shows the origins of export of commodities
with negative and statistically significant home-bias estimates and Table 10 of commodities
with positive and statistically significant home-bias estimates. We see in Table 9 that, in
general, commodities with negative home-bias are geographically clustered in the northeastern part of the U.S. – even the inter-state trade of ‘Vehicle Parts’ which, at a first sight,
looks geographically dispersed, is concentrated in the Manufacturing Belt with a share of
almost 90%. On the other hand, Table 10 shows that commodities with positive home-bias
estimates are geographically dispersed.
To establish a relationship between the sign of the home-bias estimates and spatial
concentration of industries precisely, we calculate EG index for each of the commodity with
positive and negative home-bias estimates. We use EG index as amended by Maurel and
Sedillot (1999). Their version is more suitable for our purposes since it does not require
plant-level employment data that are not available for pre-1950 U.S Censuses, but only the
number of plants in each industry which, on the other hand, can be obtained from 1939 U.S.
Census of Manufactures. Table 11 presents EG index for the commodity groups in our 1949
carload-waybill data. We see that we can successfully match all but three commodities with
1939 industries. Now we can see whether geographical concentration is related to the sign of
16
Chen (2004) also finds that some commodities in EU trade exhibit negative home-bias.
19
the home-bias. Table 12 presents the geographical concentration index for every commodity
group with statistically significant positive and negative estimates respectively. The picture
emerging from that table is clear: industries with the negative and significant home-bias
estimates are, on average, more geographically concentrated than those with the significant
but positive home-bias estimates. One tail t-test, which is included in that table, supports that
picture as well. This confirms our conjecture: low magnitude of home-bias estimates in Table
6 is driven by the commodities with negative home-bias estimates which are geographically
highly concentrated.
So far we have established that the low values of home-bias are due to the high levels of
inter-state trade of commodities produced by highly concentrated industries. The next natural
question is: where are those industries concentrated and is there anything specific about that
concentration? By looking at Table 9 we see that the origin of the inter-state trade of those
commodities is in the Manufacturing Belt. Maps 3, 4, and 5 illustrate that as well: we see that
the industries of transportation, machinery, and instruments respectively are highly
concentrated in the Manufacturing Belt area. This and the fact that inter- rather than intrastate trade is more prevalent among industries in the Manufacturing Belt suggests that a
relatively small area of the U.S. with a massive concentration of production ships its products
into the rest of the country. Map 6 provides the first indication that this might be the case:
even though large portion of shipments ends in the Manufacturing Belt, overall the
destinations of the inter-state exports are scattered across the whole United States. Indeed,
while more than 56% of inter-state export originates in the Manufacturing Belt (see also Map
1), only about 41% of U.S. domestic trade ends there. Table 15 offers further evidence that
the Manufacturing Belt served the rest of the country with its products. We see that typical
Manufacturing Belt products such as passenger cars or manufactured iron and steel are
shipped more outside of the belt states than to the belt states themselves. Even the products
20
such as electrical equipment which is shipped more or less equally between outside and
inside of the belt support the argument that the Manufacturing Belt states served the rest of
the country when we realized that there were only handful of them.
How does the situation in 1949 compares with that of 2007? We have seen earlier that by
2007, the Manufacturing Belt dissolves and the production of manufactures moves towards
the south. This has an effect on the domestic trade: the Manufacturing Belt states are no
longer the dominant suppliers of goods such as passenger cars or manufactured iron and steel;
southern states become increasingly the producers of once typical manufacturing belt
products. Indeed, as was noted earlier, while the export from the Manufacturing Belt states is
about 56% of all U.S. inter-state trade in 1949, it drops to about 35% by 2007. This implies
that, for example, Michigan does not have to supply most of the passenger cars to the rest of
the United States; the states such as Kentucky or Georgia produce them as well. The spread
of manufacturing production is consequential for the relationship between intra- and interstate trade. As we have observed earlier, there is an increase in the intra- as opposed to interstate trade between 1949 and 2007, as we already showed earlier in Table 4 and Figures 1 and
2. This is then reflected in the sign and magnitude of the estimated home-bias coefficients at
the commodity level in 2007: unlike in 1949, neither of them is negative nor less than one.
Conclusion
The spatial distribution of manufacturing in the United States has changed dramatically
between the mid of the twentieth and the first decade of the twenty-first century. While the
former period was dominated by a spatially clustered production in the north-east and midwest corners of the U.S. known as the Manufacturing Belt, the latter period exhibits more
evenly spread of manufacturing activities along the south-east corridor. We argue that this
21
change affects U.S. domestic trade, specifically a relationship between intra- vs. inter-state
trade. We show the manufacturing inter-state trade was more prevalent in 1949 than in 2007,
as reflected in the magnitude of the home-bias estimates, and that it has to do with the
concentration of manufacturing activities in the Manufacturing Belt region. We use a unique
commodity-level data set of railway shipments in 1949 collected by the Interstate Commerce
Commission and show that the prevalence of inter-state trade in 1949 is related to the
commodities which are highly geographically concentrated and produced mostly in the
Manufacturing Belt region.
22
References
Anderson, J. E. and E. van Wincoop (2003). "Gravity with Gravitas: A Solution to the Border
Puzzle." The American Economic Review 93(1): 170-192.
Anderson, J. E. and E. van Wincoop (2004). "Trade Costs." Journal of Economic Literature
42(3): 691-751.
Anderson, J. E. and Y. V. Yotov (2010). "Specialization: Pro- and Anti-Globalizing, 19902002." NBER Working Paper No. 16301.
Barger, H. (1951). The Transportation Industries 1889-1946. A Study of Output,
Employment, and Productivity. New York, National Bureau of Economic Research, Inc.
Chen, N. (2004). "Intra-National versus International Trade in the European Union: Why Do
National Borders Matter? ." Journal of International Economics 63: 93-118.
Ellison, G. and E. L. Glaeser (1997). "Geographic Concentration in U.S. Manufacturing
Industries: A Dartboard Approach." Journal of Political Economy 105(5): 889-927.
Fritz, W. G. (1943). Natural Resources - Minerals. Industrial Location and National
Resources. Washington D.C., United States Government Printing Office.
Glaeser, E. L., G. A. M. Ponzetto, et al. (2011). "Cities, Skills, and Regional Change."
Mimeo, Harvard University.
Glaeser, E. L. and K. Tobio (2008). "The Rise of the Sunbelt." Southern Economic Journal
74(3): 610-643.
Hillberry, R. (2000). "Explaining the "Border Effect": What Can We Learn from
Disaggregated Commodity Flow Data?" Mimeo.
Hillberry, R. and D. Hummels (2003). "Intra-national Home Bias: Some Explanations." The
Review of Economics and Statistics 85(4): 1089-1092.
Hillberry, R. and D. Hummels (2008). "Trade Responses to Geographic Frictions: A
Decomposition Using Micro-Data." European Economic Review 52: 527-550.
Hillberry, R. H. (2002). "Aggregation Bias, Composition Change, and the Border Effect."
The Canadian Journal of Economics 35(3): 517-530.
Holmes, T. J. and J. J. Stevens (2004). "Spatial Distribution of Economic Activities in North
America." Handbook of Regional and Urban Economics Volume 4(Cities and Geography):
2798-2843.
Kim, S. (1995). "Expansion of markets and the geographic distribution of economic
activities: trends in U. S. regional manufacturing structure, 1860-1987." The Quarterly
Journal of Economics 110: 881-908.
23
Kim, S. and R. A. Margo (2004). Historical Perspectives on U.S. Economic Geography.
Handbook of Regional and Urban Economics: Cities and Geography. J. V. Henderson and J.F. Thisee. Amsterdam, Elsevier North-Holland: 2981–3019.
Maurel, F. and B. Sédillot (1999). "A Measure of the Geographic Concentration in French
Manufacturing Industries." Regional Science and Urban Economics 29(5): 575–604.
McCallum, J. T. (1995). "National Borders Matter: Canada-U.S. Regional Trade Patterns."
The American Economic Review 85: 615-623.
Meyer, D. R. (1983). "Emergence of the American Manufacturing Belt: An Interpretation."
Journal of Historical Geography 9: 145-174.
Meyer, D. R. (1989). "Midwestern Industrialization and the American Manufacturing Belt in
the Nineteenth Century." Journal of Economic History 49: 921-937.
Meyer, J. R., M. J. Peck, et al. (1960). The Economics of Competition in the Transportation
Industries. Cambridge, Massachusetts, Harvard University Press.
Millimet, D. L. and T. Osang (2007). "Do State Borders Matter for U.S. Intranational Trade?
The Role of History and Internal Migration." The Canadian Journal of Economics 40(1): 93126.
Nitsch, V. (2000). "National Borders and International Trade: Evidence from the European
Union." Canadian Journal of Economics 33: 1091-1105.
Perloff, H. S., E. S. Dunn, et al. (1960). Regions, Resources, and Economic Growth. Lincoln,
University of Nebraska Press.
Santos Silva, J. M. C. and S. Tenreyro (2006). "The Log of Gravity." The Review of
Economics and Statistics 88(4): 641-658.
Wolf, H. C. (1997). "Patterns of Intra- and Inter-State Trade." NBER Working Paper 5939.
Wolf, H. C. (2000). "Intra-national Home Bias in Trade." The Review of Economics and
Statistics 82(4): 555-563.
Wolf, N. (2009). "Was Germany Ever United? Evidence from Intra- and International Trade,
1885-1933." The Journal of Economic History 69(3): 846-881.
24
Table 1: Distribution of Shipments by Transportation Modes in the United States: 1948, 2007.
Values
Transportation Mode
Value of
shipment ($mil)
Weight of
shipment (000
tons)
% from total
Ton miles
Value of
shipment
Weight of
shipment
Ton miles
Panel A: 2007
Air (incl truck and air)
124,159
1,120
1,370
1.14
0.01
0.05
For-hire truck
4,891,695
3,994,568
993,599
44.88
34.89
36.77
Private truck
3,370,550
4,610,793
265,909
30.93
40.27
9.84
Truck and rail
124,282
120,296
100,219
1.14
1.05
3.71
Truck and water
21,500
58,146
28,195
0.20
0.51
1.04
1,520,533
32,002
25,584
13.95
0.28
0.95
Parcel, U.S.P.S., courier
Great Lakes
239
13,833
4,290
0.00
0.12
0.16
Water
88,930
305,669
108,817
0.82
2.67
4.03
Pipeline
348,073
543,169
3.19
4.74
9,521
21,956
7,019
0.09
0.19
Deep draft
0.26
Shallow draft
76,955
265,011
96,205
0.71
2.31
3.56
Rail
315,788
1,468,575
1,066,065
2.90
12.83
39.45
6,627
10,898,852
13,261
11,448,399
4,808
2,702,080
0.06
0.12
0.18
100.00
100.00
100.00
Rail and water
All modes
Panel B: 1948
Railway
647,267
64.39
Highways
Inland waterways and Great Lakes
87,640
150,530
8.72
14.97
Pipe lines
119,597
11.90
Airways
All modes
223
1,005,257
0.02
100.00
Notes: 1948 figures refer to intercity freight traffic.
Sources: Interstate Commerce Commission 1950, Commodity Flow Survey 2007.
25
Table 2: Origins of Inter-State U.S. Trade of Motorized Vehicles in 1949 and 2007 (% of U.S. Domestic Trade).
2007
Motorized and Other
Vehicles
1949
Vehicle, not Motor
Vehicle parts
Alabama
2.42
California
5.00
Illinois
California
7.21
Colorado
1.67
Indiana
Florida
1.77
Illinois
21.67
Michigan
Georgia
3.32
Indiana
0.83
New Jersey
Illinois
4.18
Massachusetts
4.17
New York
Indiana
4.37
Michigan
4.17
Ohio
Kentucky
5.55
Minnesota
2.50
Pennsylvania
Michigan
11.45
Missouri
6.67
Wisconsin
Missouri
2.25
New York
7.50
New Jersey
1.68
Ohio
32.50
New York
2.02
Pennsylvania
10.00
North Carolina
1.69
Wisconsin
3.33
Ohio
5.79
Pennsylvania
1.50
South Carolina
1.46
Tennessee
1.59
Texas
6.30
Wisconsin
2.01
Note: U.S. States with the shares less than 1% are excluded.
Source: Interstate Commerce Commission 1950, Commodity Flow Survey 2007.
26
Automobiles
2.72
7.00
53.74
1.16
4.37
18.21
6.10
4.22
California
Illinois
Indiana
Michigan
Ohio
Pennsylvania
23.34
2.46
8.81
61.43
3.89
0.07
Table 3: Origins of Inter-State U.S. Trade of Machinery in 1949 and 2007 (% of U.S.
Domestic Trade).
2007
1949
Machinery
Machines
Arizona
1.02
Alabama
California
5.40
California
Connecticut
1.07
Connecticut
Florida
2.07
Georgia
Georgia
2.19
Illinois
Illinois
4.80
Indiana
Indiana
2.63
Iowa
Iowa
1.86
Maine
Kentucky
1.85
Massachusetts
Massachusetts
1.05
Michigan
Michigan
4.67
Minnesota
Minnesota
1.32
New Jersey
Missouri
1.32
New York
New Jersey
1.38
Ohio
New York
2.60
Pennsylvania
North Carolina
2.18
Rhode Island
Ohio
5.77
Texas
Oklahoma
1.22
Vermont
Pennsylvania
2.58
Wisconsin
South Carolina
1.59
Tennessee
3.09
Texas
6.76
Virginia
1.24
Washington
1.21
Wisconsin
3.41
Note: U.S. States with the shares less than 1% are excluded.
Source: Interstate Commerce Commission 1950, Commodity Flow Survey 2007.
27
1.15
2.22
1.60
1.78
14.92
3.64
3.11
1.60
5.51
4.80
1.24
2.58
7.90
15.10
9.24
1.15
1.07
1.51
9.41
Table 4: U.S. Intra- and Inter-State Trade in 1949, 2007: Summary Statistics.
N
Mean
Std. Dev
Min
Max
1949
intra-state (%)
inter-state (%)
48
48
26.3
73.6
19.6
19.6
2007
86.7
13.3
intra-state (%)
48
31.9
17.3
8.7
88.6
inter-state (%)
48
68.1
17.3
11.4
91.3
Note: intra- and inter-state trade is percentage from state's total domestic trade
Sources: Interstate Commerce Commission 1950, Commodity Flow Survey 2007.
28
Table 5: Intrastate Trade by Commodities in 1949 and 2007 (% from Total Domestic Trade).
1949
2007
Cigarettes
Machines, parts
Lime
Manufactured Iron Steel
Candy, Confectionary
Airplanes
Hardware
Copper Ingot
Agriculture Impl, Parts
Furniture parts
Agriculture Implements
Vehicle parts
Fridge
Copper, Brass, Bronze
Furniture
Machines
Automobiles, Passengers
Vehicle, not Motor
Cotton Cloth
Electrical Equipment
Rubber Crude
Paper Articles
Woodware
Tires, Tubes, Rubber
Chemicals
Food products
Boots, Shoes
Newsprint papers
Wood Pulp
Containers Wooden
Acids
Aluminium Bars
Fertilizers
Cement Portland
Bricks, Common
Iron and Steel
Gasoline
1.0
Transportation Equip
1.1
Precision Instruments
2.4
Textile and leather
2.5
Electronic and electrical equip
3.3
Motorized vehicle
3.6
Pharmaceuticals
4.5
Pulp, newsprint paper
5.8
Machinery
6.0
Printed products
6.3
Furniture
7.5
Plastic and Rubber
8.5
Base metal
8.5
Basic chemicals
8.5
Paper and paperboard
8.8
Fertilizers
9.3
Wood Products
9.6
Non-metallic Minerals
9.8
Log and wood rough
10.1 Coal and petroleum
10.3 Tobacco
10.8 Fuel oils
13.3 Gasoline and aviation fuel
14.9
15.4
15.9
16.4
21.1
25.8
30.6
31.3
34.0
35.0
44.2
48.5
54.3
54.8
54.9
Source: The Interstate Commerce Commission 1950; 2007 Commodity Flow Survey.
29
18.87
19.98
20.49
23.81
23.98
24.68
25.16
26.36
26.78
26.87
26.90
27.90
29.98
30.80
33.75
34.57
37.62
37.72
38.86
43.30
45.75
46.51
Table 6: Gravity Equation with Intrastate Home Bias, U.S.1949.
Panel A: Manufacturing
Intra-state distance measures
ln_distance
home_bias
ln_remoteij2007_gsp
ln_remoteji2007_gsp
adjacent
Constant
Nitsch
Wolf
Largest Cities
(I)
(II)
(III)
-0.57***
-0.13*
-0.42***
[0.11]
[0.08]
[0.09]
0.90***
1.29***
1.27***
[0.21]
[0.23]
[0.19]
-1.1
-0.45
-1.48
[1.20]
[1.20]
[1.33]
2.30***
2.43***
2.25***
[0.63]
[0.62]
[0.65]
0.72***
1.25***
0.89***
[0.18]
[0.15]
[0.17]
-8.13
-20.58
-3.76
[15.82]
[15.62]
[16.91]
N
92864
92864
92864
Export/Import FE
Yes
Yes
Yes
Source: 1949 Carload Waybill Data, ICC.
Notes: the dependent variable is the weight of shipment (in tons).
30
Table 7: Gravity Equation with Intrastate Home Bias, U.S. 2007.
Panel A: Whole Economy
Nitsch
(I)
Intra-state distance measures
Largest
Wolf
Actual Distance
Cities
(II)
(III)
(IV)
ln_distance
-0.79***
[0.13]
[0.06]
[0.05]
[0.07]
home_bias
2.01***
2.48***
2.51***
2.43***
[0.12]
[0.09]
[0.06]
[0.12]
ln_remoteij2007_gsp
1.79
4.15
3.94
4.15
[4.43]
[4.63]
[4.53]
[4.63]
ln_remoteji2007_gsp
43.62***
92.11***
88.97***
92.11***
[10.54]
[7.09]
[7.22]
[7.09]
adjacent
0.58***
0.78***
0.79***
0.75***
[0.08]
[0.07]
[0.07]
[0.08]
Constant
51.24***
89.25***
86.69***
89.25***
[7.98]
[5.31]
[5.49]
[5.31]
2304
2304
2304
2304
Yes
Yes
Yes
N
Export/Import FE
Yes
-0.08
-0.11**
-0.1
Panel B: Manufacturing Sector
Nitsch
(I)
ln_distance
home_bias
ln_remoteij2007_gsp
ln_remoteji2007_gsp
adjacent
Constant
Intra-state distance measures
Largest
Wolf
Actual Distance
Cities
(II)
(III)
(IV)
-0.98***
-0.09*
-0.21***
-0.16***
[0.12]
[0.05]
[0.06]
[0.04]
1.67***
2.28***
2.25***
1.79***
[0.11]
[0.11]
[0.09]
[0.09]
24.85***
35.16***
32.49***
70.40***
[4.91]
[5.92]
[5.71]
[5.56]
13.17
67.32***
59.43***
10.10***
[8.83]
[6.89]
[6.60]
[3.86]
0.40***
0.67***
0.66***
0.51***
[0.10]
[0.09]
[0.10]
[0.07]
28.32***
72.12***
65.35***
31.97***
[6.80]
[5.24]
[5.08]
[2.79]
N
49137
49137
49137
49137
Export/Import FE
Yes
Yes
Yes
Yes
Source: The Commodity Flow Survey 2007.
Notes: the dependent variable is weight of shipment in short-tons (2000 pounds).
31
Table 8: Estimates of Home Bias in 1949 and 2007.
Lower Bound 1949
Upper Bound 1949
2007
Intra-State Distance: Nitsch Formula
home_bias
0.90***
1.44***
1.67***
[0.21]
[0.18]
[0.11]
Intra-State Distance: Wolf Formula
home_bias
1.29***
1.86***
2.28***
[0.23]
[0.18]
[0.11]
Intra-State Distance: Largest Cities
home_bias
1.27***
1.81***
2.25***
[0.19]
[0.14]
[0.09]
Source:
Lower bound 1949: Table 6; 2007: Table 7, Panel B
Upper Bound 1949: 1949 Carload Waybill Data, ICC; Fluctuations in Railway Traffic
Compared with Production', ICC Statement 3951, 1939
32
Table 9: Home Bias by Commodities, U.S. 1949.
Statistically Significant
Statistically Insignificant
Positive Effect
Estimates &
Negative Effect
Stat Signif
Estimates &
Estimates &
Positive Effect
Stat Signif
Stat Signif
Intra-State Distance Measured with Nitsch Distance Formula
Negative Effect
Estimates &
Stat Signif
Fertilizers
0.87***
Paper articles
-0.44***
Cloth and Fabrics
Gasoline
0.96***
Chemicals
-0.50**
Woodware
1.96
Food Products
-0.12
0.11
Candy, Confectionary
-0.66
Boots, Shoes
2.43**
Copper Ingot
-3.44***
Bricks Common
1.82***
Copper, Brass, Bronze
-1.45**
Wooden containers
0.01
Manuf Iron and Steel
-0.35
Wood pulp
0.53
Furniture
-0.52
Refrigerators
0.72**
Automobiles
Average
1.36
Vehicle parts
-4.40***
Acids
0.18
Iron and Steel
-1.00
-1.20***
Crude rubber
0.24
Tires, Tubes, Rubber
-0.34
Hardware
-3.42***
Cement Portland
0.08
Agricul Impl.
-0.35
Airplanes
-4.66***
Lime
0.81
Vehicle not Motors
-0.68
Cigarettes
-1.92*
Machines
0.09
Newsprint paper
-0.18
Average
-2.38
Electrical Equipment
0.13
Agric Impl. Parts
-0.77
Cotton Cloth
0.91
Machinery parts
-0.8
Home Bias Estimate for the Whole Economy: 0.90***
Average
Average
Intra-State Distance Measured by the Distance between the Largest Cities
Cloth and Fabrics
2.35**
Copper Ingot
-3.71***
Food Products
0.28
Cigarettes
-1.03
Wooden containers
0.60***
Copper, Brass, Bronze
-1.12**
Candy, Confectionary
0.01
Furniture
-0.13
Wood pulp
0.78**
Automobiles
-2.47***
Cotton Cloth
1.09
Manuf Iron and Steel
-0.31
Newsprint paper
2.16***
Vehicle parts
-0.77***
Woodware
0.7
Agric Impl. Parts
-0.7
Acids
0.79***
Airplanes
-5.50***
Paper articles
0.2
Agricul Impl.
-0.24
Fertilizers
1.74***
Hardware
-2.95***
Chemicals
0.1
Machinery parts
-0.85
Gasoline
2.26***
Average
-2.75
Tires, Tubes, Rubber
0.22
Vehicle not Motors
-0.03
Crude rubber
0.81***
Lime
0.94
Average
Boots, Shoes
2.59**
Electrical Equipment
0.33
Cement Portland
1.91***
Average
Bricks Common
2.36***
Machines
0.32*
Iron and Steel
1.59***
Refrigerators
0.58**
Average
1.49
Home Bias Estimate for the Whole Economy: 1.29***
33
Table 9: continued.
Statistically Significant
Positive Effect
Estimates &
Stat Signif
Negative Effect
Statistically Insignificant
Estimates &
Stat Signif
Positive Effect
Estimates &
Stat Signif
Negative Effect
Estimates &
Stat Signif
Intra-State Distance Measured by Wolf's Formula
Cloth and Fabrics
2.48*
Cigarettes
-2.01**
Cotton Cloth
1.04
Food Products
-0.24
Acids
0.92***
Paper articles
-0.62**
Wooden containers
0.05
Candy, Confectionary
-0.75
Fertilizers
1.63***
Copper Ingot
-4.42***
Newsprint paper
0.27
Woodware
-0.08
0.09
Wood pulp
-0.39
Gasoline
2.05***
Copper, Brass, Bronze
-2.03**
Crude rubber
Boots, Shoes
3.45***
Agricul Impl.
-0.63*
Lime
0.82
Furniture
-0.3
Cement Portland
1.13***
Agric Impl. Parts
-2.14*
Iron and Steel
1.04
Chemicals
-0.15
Bricks Common
2.10***
Automobiles
-4.09***
Machines
0.13
Manuf Iron and Steel
-0.6
Refrigerators
0.50*
Vehicle parts
-1.28***
Electrical Equipment
0.43
Machinery parts
-0.79
Average
1.78
Airplanes
-4.95***
Average
Vehicle not Motors
-0.88
Hardware
-3.53***
Tires, Tubes, Rubber
-0.5
Average
-2.57
Average
Home Bias Estimate for the Whole Economy: 1.27***
Note: Estimates are from the pseudo-Poisson ML regression with import and export dummies, adjacent dummy and remoteness controls.
Source: see text
34
Table 10: Home Bias by Commodities: Estimated Coefficients and Standard Errors, US 2007.
Internal Distance Measure, without GSP
Industry
SIC
20
Grains, alcohol, and tobacco products
20
Other prepared foodstuffs and fats and oils
20
Alcoholic beverages
21
Tobacco products
32
Calcareous monumental or building stone
32
Natural sands
32
Gravel and crushed stone
32
Nonmetallic minerals nec
33
Metallic ores and concentrates
12
Nonagglomerated bituminous coal
29
Gasoline and aviation turbine fuel
29
Fuel oils
29
Coal and petroleum products, nec
28
Basic chemicals
28
Pharmaceutical products
28
Fertilizers
28
Chemical products and preparations, nec
Plastics and rubber
30
24
Logs and other wood in the rough
24
Wood products
26
Pulp, newsprint, paper, and paperboard
26
Paper or paperboard articles
26
Printed products
35
Largest
City
Wolf
Actual
Nitsch
1.56***
1.62***
1.62***
0.68***
[0.11]
1.92***
[0.07]
3.41***
[0.17]
2.50***
[0.31]
2.01***
[0.48]
3.06***
[0.27]
4.45***
[0.27]
2.86***
[0.43]
3.57***
[0.54]
3.26***
[0.55]
4.10***
[0.33]
3.76***
[0.36]
3.41***
[0.19]
1.65***
[0.19]
1.69***
[0.21]
2.73***
[0.27]
1.66***
[0.13]
1.42***
[0.09]
2.06***
[0.51]
1.88***
[0.09]
1.15***
[0.09]
1.56***
[0.09]
1.98***
[0.11]
[0.12]
1.99***
[0.09]
3.14***
[0.22]
2.34***
[0.38]
1.04
[0.64]
2.97***
[0.33]
4.47***
[0.30]
2.45***
[0.46]
2.36***
[0.76]
3.18***
[0.55]
4.49***
[0.39]
4.25***
[0.30]
3.13***
[0.21]
1.46***
[0.18]
1.63***
[0.25]
2.66***
[0.35]
1.62***
[0.15]
1.30***
[0.09]
3.05***
[0.44]
1.81***
[0.11]
1.08***
[0.09]
1.39***
[0.12]
1.78***
[0.15]
[0.15]
1.93***
[0.11]
3.05***
[0.27]
2.00***
[0.49]
0.38
[0.74]
2.70***
[0.47]
4.49***
[0.35]
1.98***
[0.54]
2.02*
[1.03]
3.70***
[0.63]
4.70***
[0.46]
4.39***
[0.38]
3.05***
[0.26]
1.22***
[0.23]
1.61***
[0.30]
2.82***
[0.43]
1.61***
[0.19]
1.21***
[0.11]
3.07***
[0.64]
1.71***
[0.14]
1.00***
[0.11]
1.23***
[0.14]
1.69***
[0.19]
[0.18]
1.14***
[0.16]
2.78***
[0.26]
3.45***
[0.63]
2.01
[0.00]
1.96***
[0.49]
3.94***
[0.39]
2.38***
[0.47]
4.70***
[1.39]
3.48***
[0.66]
3.07***
[0.41]
3.57***
[0.55]
2.38***
[0.30]
1.08***
[0.24]
1.75***
[0.36]
2.87***
[0.47]
1.58***
[0.17]
1.07***
[0.10]
1.60**
[0.78]
1.37***
[0.14]
0.92***
[0.13]
1.36***
[0.16]
1.59***
[0.19]
Table 10: continued.
Textiles, leather, and articles of textiles or leather
Nonmetallic mineral products
Base metal in prim. or semifin. forms & in finished
basic shapes
Articles of base metal
Machinery
Electronic & other electrical equip & components &
office equip
Motorized and other vehicles (including parts)
Transportation equipment, nec
Precision instruments and apparatus
Furniture, mattresses & mattress supports, lamps,
lighting
Miscellaneous manufactured products
22
32
33
33
35
36
37
37
38
25
39
Average of commodity estimates
1.43***
[0.13]
2.87***
[0.10]
1.05***
[0.14]
1.51***
[0.09]
1.75***
[0.10]
2.15***
[0.14]
1.62***
[0.13]
0.73
[0.45]
2.14***
[0.28]
1.66***
1.30***
[0.14]
2.66***
[0.12]
1.10***
[0.15]
1.38***
[0.11]
1.78***
[0.12]
2.11***
[0.16]
1.75***
[0.14]
1.27**
[0.55]
2.01***
[0.30]
1.50***
1.15***
[0.17]
2.54***
[0.16]
1.05***
[0.18]
1.26***
[0.14]
1.68***
[0.15]
2.09***
[0.21]
1.74***
[0.17]
1.32**
[0.63]
1.92***
[0.34]
1.32***
1.26***
[0.17]
2.72***
[0.12]
0.71***
[0.19]
1.20***
[0.13]
1.21***
[0.18]
1.42***
[0.20]
1.04***
[0.19]
1.18*
[0.66]
1.67***
[0.43]
1.14***
[0.11]
1.97***
[0.11]
[0.12]
1.79***
[0.13]
[0.14]
1.70***
[0.15]
[0.17]
1.52***
[0.19]
2.25
2.17
2.09
1.94
Note: The estimates are from the regression with import and export dummies, adjacent dummy and remoteness
controls.
Source: see text
36
Table 11: Inter-State Trade by Commodities, U.S. 1949 (% from Total Domestic Trade).
Vehicle parts
Copper, Brass, Bronze
%
Vehicle, not Motor
%
Automobiles,
Passengers
Hardware
%
%
Copper Ingot
%
Airplanes
%
%
California
0.32
Arizona
11.76
California
5.00
Alabama
7.14
California
23.34
Arizona
27.55
California
44.44
Georgia
0.09
California
1.04
Colorado
1.67
California
2.38
Illinois
2.46
Michigan
3.06
Kansas
29.63
Illinois
2.72
Connecticut
10.38
Illinois
21.67
Connecticut
16.67
Indiana
8.81
New Jersey
39.80
New York
14.81
Indiana
7.00
Illinois
7.96
Indiana
0.83
Illinois
14.29
Michigan
61.43
New York
10.20
Ohio
11.11
Iowa
0.03
Indiana
4.15
Massachusetts
4.17
Indiana
11.90
Ohio
3.89
Texas
7.14
Kentucky
0.03
Maryland
8.30
Michigan
4.17
Iowa
2.38
Pennsylvania
0.07
Utah
12.24
Maryland
0.03
Massachusetts
1.38
Minnesota
2.50
Massachusetts
7.14
Massachusetts
0.14
Michigan
5.88
Missouri
6.67
Michigan
2.38
Michigan
53.74
Minnesota
3.46
New York
7.50
New York
19.05
Minnesota
0.17
Missouri
0.69
Ohio
32.50
Ohio
4.76
Missouri
0.43
New Jersey
13.15
Pennsylvania
10.00
Pennsylvania
11.90
New Jersey
1.16
New York
9.69
Wisconsin
3.33
New York
4.37
Ohio
2.08
Ohio
18.21
Pennsylvania
4.50
Pennsylvania
6.10
Rhode Island
1.73
Tennessee
0.46
Texas
10.03
West Virginia
0.78
Washington
2.77
Wisconsin
4.22
Wisconsin
1.04
Total
100.00
100.00
Total
100.00
Total
100.00
Total
100.00
Total
100.00
Total
100.00
Source: Interstate Commerce Commission 1950.
37
Total
Table 12: Inter-State Trade by Commodities U.S. 1949 (% from Total Domestic Trade).
Gasoline
Fertilizers
%
Bricks
%
%
Alabama
2.31
Alabama
4.17
Alabama
12.39
Arkansas
2.31
Arkansas
3.28
Colorado
0.88
California
5.07
California
1.41
Georgia
23.89
Florida
2.36
Colorado
0.21
Ilinois
11.50
Georgia
0.30
Connecticut
0.36
Iowa
4.42
Ilinois
6.06
Florida
0.83
Kansas
5.31
Indiana
3.64
Georgia
2.03
Kentucky
0.88
Iowa
0.98
Ilinois
9.49
Louisiana
1.77
Kansas
2.41
Indiana
5.79
Minnesota
3.54
Kentucky
0.34
Iowa
2.29
Mississippi
1.77
Louisiana
5.37
Kansas
2.19
North Carolina
10.62
Maine
1.23
Kentucky
0.36
Ohio
0.88
Maryland
1.03
Louisiana
2.92
Oklahoma
0.88
Massachusetts
0.49
Maine
0.26
Pennsylvania
6.19
Michigan
0.44
Maryland
3.91
South Carolina
0.88
Minnesota
4.38
Massachusetts
1.09
Tennessee
6.19
Mississippi
0.10
Michigan
0.42
Texas
3.54
Missouri
1.03
Minnesota
0.31
Virginia
1.77
Montana
1.23
Mississippi
0.21
Washington
0.88
Nebraska
0.05
Missouri
2.92
West Virginia
1.77
New Jersey
0.39
Nebraska
0.99
New Mexico
0.10
Nevada
0.31
New York
1.43
New Jersey
3.08
North Carolina
0.20
New Mexico
15.17
North Dakota
0.64
New York
2.09
Ohio
1.38
North Carolina
0.89
Oklahoma
10.88
Ohio
9.12
Oregon
1.33
Oklahoma
0.83
Pennsylvania
0.94
Oregon
0.10
Rhode Island
1.03
Pennsylvania
3.13
South Carolina
1.38
South Carolina
1.09
South Dakota
0.59
Tennessee
4.12
Tennessee
4.83
Texas
0.73
Texas
20.38
Utah
0.94
Utah
1.92
Virginia
12.67
Virginia
2.07
Washington
0.26
Washington
0.69
West Virginia
2.02
Wisconsin
1.28
Wyoming
5.37
Total
100.00
Total
100.00
Source: Interstate Commerce Commission 1950.
38
Total
100.00
Table 12: continued
Refrigerators
Boots, Shoes
%
%
California
0.54
Illinois
14.08
Illinois
2.39
Indiana
1.41
Indiana
19.09
Massachusetts
49.30
Iowa
0.33
Missouri
5.63
Kentucky
0.65
New Hampshire
2.82
Maryland
0.65
New York
12.68
Massachusetts
1.63
Rhode Island
2.82
Michigan
16.81
Tennessee
11.27
Minnesota
6.29
Missouri
0.43
New Jersey
0.98
Total
100.00
New York
3.36
Ohio
28.85
Pennsylvania
16.59
Tennessee
0.33
Texas
0.22
Wisconsin
0.87
Total
100.00
Source: Interstate Commerce Commission 1950.
39
Table 13: Ellison and Glaeser Index of Geographical Concentration:
Matching of 1939 Industries with 1949 Commodity Groups.
Commodity Group in ICC
1949 Trade Flow Data
Industries in 1939 Census of Manufactures
SIC 2
SIC 3
Tires, Rubber
30
301
Electrical Equp.
36
362
Plastics
28
282
Cigarettes
21
211
Airplanes
37
372
Vehicles Not Motor
37
379
Boots and Shoes
31
314
Chemicals
28
Cotton Cloth
22
221
Copper Ingot
33
333
Copper, Brass, Bronze
33
333
Gasoline
29
291
Hardware
34
345
Furniture
25
Furniture, Parts
25
Wood Pulp
26
261
Automobiles (Passengers)
37
371
Vehicle Parts
37
371
Wooden Containers
24
244
Machinery Parts
35
356
Machinery
35
356
Aluminium Bar
33
335
Cement
32
324
Paper Articles
27
275
Manuf. Iron and Steel
34
344
Refrigerators
36
363
Agricultural Impl.
35
352
Agricultural Impl. Parts
35
352
Iron and Steel
33
332
Candy and confectionary
20
206
Bricks
32
325
Food Products
20
Fertilizers
28
287
Acids
28
281
Lime
32
327
Newsprint papers
27
271
Woodware
24
Rubber Crude
Cellulose Articles
Source: U.S. Census of Manufactures, 1939.
EG
index
SIC Category
Tires and Inner Tubes
Electrical Industrial Apparatus
Plastics Materials & Synthetics
Cigarettes
Aircraft & Parts
Miscellaneous Transportation Equipment
Footwear, Except Rubber
Chemicals
Broadwoven Fabric Mills, Cotton
Primary Nonferrous Metals
Primary Nonferrous Metals
Petroleum Refining
Screw Machine Products, Bolts, Etc.
Furniture
Furniture
Pulp Mills
Motor Vehicles & Equipment
Motor Vehicles & Equipment
Wood Containers
General Industry Machinery
General Industry Machinery
Nonferrous Rolling & Drawing
Cement, Hydraulic
Commercial Printing
Fabricated Structural Metal Products
Household Appliances
Farm & Garden Machinery
Farm & Garden Machinery
Iron & Steel Foundries
Sugar & Confectionery Products
Structural Clay Products
Food Products
Agricultural Chemicals
Industrial Inorganic Chemicals
Concrete, Gypsum & Plaster Products
Newspapers
40
0.308
0.304
0.288
0.266
0.153
0.138
0.133
0.120
0.118
0.112
0.112
0.109
0.108
0.108
0.108
0.106
0.096
0.096
0.091
0.089
0.089
0.085
0.084
0.074
0.074
0.071
0.068
0.068
0.059
0.059
0.057
0.057
0.051
0.049
0.046
0.036
Table 14: Home Bias Estimates of Commodities and EG Index, U.S. 1949.
Positive Effect
EG Index
Negative Effect
EG Index
Intra-State Distance Measured with Nitsch Distance Formula
Fertilizers
0.051
Paper articles
0.074
Gasoline
0.109
Chemicals
0.120
Boots, Shoes
0.133
Copper Ingot
0.112
Bricks Common
0.057
Copper, Brass, Bronze
0.122
Refrigerators
0.071
Automobiles
0.096
Vehicle parts
0.096
Hardware
0.108
Airplanes
0.153
Cigarettes
0.266
Average
0.128
Average
0.084
One-Tail Test: Positive Home Bias > Negative Home Bias: t = 1.54*
Intra-State Distance Measured by the Distance between the Largest Cities
Cloth and Fabrics
0.118
Copper Ingot
0.112
Wooden containers
0.091
Copper, Brass, Bronze
0.122
Wood pulp
0.106
Automobiles
0.096
Newsprint paper
0.036
Vehicle parts
0.096
Acids
0.049
Airplanes
0.153
Fertilizers
0.051
Hardware
0.108
Gasoline
0.109
Average
0.115
Crude rubber
Boots, Shoes
0.133
Cement Portland
0.084
Bricks Common
0.057
Machines
0.089
Iron and Steel
0.059
Refrigerators
0.071
Average
0.074
One-Tail Test: Positive Home Bias > Negative Home Bias: t = 2.45**
Intra-State Distance Measured by Wolf's Formula
Cloth and Fabrics
0.118
Cigarettes
0.266
Acids
0.049
Paper articles
0.074
Fertilizers
0.051
Copper Ingot
0.112
Gasoline
0.109
Copper, Brass, Bronze
0.122
Boots, Shoes
0.133
Agricul Impl.
0.068
Cement Portland
0.084
Agric Impl. Parts
0.068
Bricks Common
0.057
Automobiles
0.096
Refrigerators
0.071
Vehicle parts
0.096
Airplanes
0.153
Hardware
0.108
Average
0.079
0.116
One-Tail Test: Positive Home Bias > Negative Home Bias: t = 1.39*
Note: *, ** denote statistical significance at 10% and 5% respectively.
Source: Tables 8, 11.
41
Table 15: Export from the Manufacturing Belt States by Commodities, U.S. 1949 (%)
Export from Manufacturing Belt States Into:
Manufacturing Belt
Food Products
Candy and Confectionary
Cloth and Fabric
Cotton Cloth
Woodware
Wooden Container
Woodpulp
Furniture
Furniture parts
Newspaper
Paper articles
Chemicals
Acids
Fertilizers
Matches
Gasoline
Rubber crude
Cellulose articles
Plastics
Tires,Tubes,Rubbers
Boots. Shoe findings
cement portland
Bricks common
Lime
Aluminium bar
Copper Ingot
Copper, brass, bronze
Iron and Steel
Manufactured iron&steel
Agriculture Implements
Agric imp. Parts
Machines
Electrical equipment
Refrigerators
Vehicle not motor
Automobiles (passengers)
Vehicle parts
Airplanes
Hardware
Source: Interstate Commerce Commission 1950.
58.0
39.8
50.0
53.3
88.9
70.8
87.5
55.6
40.0
80.5
66.3
67.1
91.9
69.6
43.5
84.3
87.2
63.6
66.7
69.3
67.2
82.0
90.9
8.2
83.3
62.5
68.4
89.8
39.4
43.0
52.4
53.7
54.2
49.7
56.1
35.1
52.6
42.9
54.8
42
Rest of U.S.
42.0
60.2
50.0
46.7
11.1
29.2
12.5
44.4
60.0
19.5
33.7
32.9
8.1
30.4
56.5
15.7
12.8
36.4
33.3
30.7
32.8
18.0
9.1
91.8
16.7
37.5
31.6
10.2
60.6
57.0
47.6
46.3
45.8
50.3
43.9
64.9
47.4
57.1
45.2
Figure 1: Kernel Distribution of Intra-State Trade, U.S. 1949
.01
0
.005
Density
.015
.02
Kernel density estimate
0
20
40
60
intra-state/total state trade
80
100
kernel = epanechnikov, bandwidth = 7.2347
Figure 2: Kernel Distribution of Intra-State Trade, U.S. 2007
.02
0
.01
Density
.03
.04
Kernel density estimate
0
20
40
60
intra-state/total state trade
kernel = epanechnikov, bandwidth = 4.9763
43
80
100
Map 1: U.S. Inter-State Trade by State of Origin, 1949
Map 2: U.S. Inter-State Trade by State of Origin, 2007
44
Map 3
Map 4
45
Map 5
Map 6
46
47
Appendix
Table A1: The Rank of Home-Bias Estimates by Commodities, 2007.
SIC
Largest
City
Wolf
Actual
Nitsch
Grains, alcohol, and tobacco products
20
27
23
22
34
Other prepared foodstuffs and fats and oils
20
19
16
15
27
Alcoholic beverages
20
5
5
6
8
Tobacco products
21
12
13
13
5
Calcareous monumental or building stone
32
16
34
34
12
Natural sands
32
8
8
9
13
Gravel and crushed stone
32
1
2
2
2
Nonmetallic minerals nec
32
10
11
14
10
Metallic ores and concentrates
33
4
12
12
1
Nonagglomerated bituminous coal
12
7
4
4
4
Gasoline and aviation turbine fuel
29
2
1
1
6
Fuel oils
29
3
3
3
3
Coal and petroleum products, nec
29
5
6
6
10
Basic chemicals
28
25
26
29
29
Pharmaceutical products
28
22
22
23
14
Fertilizers
28
11
9
8
7
Chemical products and preparations, nec
28
23
23
23
18
Plastics and rubber
30
31
29
30
30
Logs and other wood in the rough
24
15
7
5
16
Wood products
24
20
17
18
21
Pulp, newsprint, paper, and paperboard
26
32
33
33
32
Paper or paperboard articles
26
27
27
28
22
Printed products
Textiles, leather, and articles of textiles or
leather
Nonmetallic mineral products
Base metal in prim. or semifin. forms & in
finished basic shapes
Articles of base metal
26
17
19
20
17
22
30
29
31
23
32
9
9
10
9
33
33
32
32
33
33
29
28
27
25
Machinery
Electronic & other electrical equip &
components
Motorized and other vehicles (including parts)
35
21
19
21
24
36
13
14
11
20
37
26
21
17
31
Transportation equipment, nec
37
34
31
25
26
Precision instruments and apparatus
38
14
15
16
15
Furniture, mattresses & mattress supports,
lamps, lighting
25
23
25
25
27
19
19
Industry
Miscellaneous manufactured products
39
18
18
Note: The estimates are from the regression with import and export dummies,
adjacent dummy and remoteness controls.
Source: Table 9
48
Table A2: Gravity Equation with Intrastate Home Bias, U.S.1949.
Panel A: Manufacturing
Intra-state distance measures
Nitsch
Wolf
Largest Cities
(I)
(II)
(III)
ln_distance
-0.61***
-0.15**
-0.46***
[0.09]
[0.07]
[0.08]
home_bias
0.67***
1.06***
1.04***
[0.18]
[0.21]
[0.17]
ln_remoteij2007_gsp
-0.6
0.03
-0.99
[1.00]
[0.99]
[1.15]
ln_remoteji2007_gsp
2.03***
2.19***
1.98***
[0.54]
[0.53]
[0.55]
adjacent
0.52***
1.08***
0.69***
[0.16]
[0.13]
[0.15]
Constant
-13.87
-26.78*
-9.51
[14.55]
[14.29]
[15.82]
N
92864
92864
92864
Export/Import FE
Yes
Yes
Yes
Source: 1949 Carload Waybill Data, ICC.
Notes: the dependent variable is the number of carloads.
49
Table A3: Gravity Equation with Intrastate Home Bias, U.S. 2007.
Panel A: Whole Economy
Intra-state distance measures
Nitsch
Wolf
Largest Cities
Actual Distance
(I)
(II)
(III)
(IV)
-0.80***
-0.15***
-0.19***
-0.19***
[0.09]
[0.03]
[0.05]
[0.04]
home_bias
1.28***
1.71***
1.78***
1.61***
[0.08]
[0.08]
[0.06]
[0.09]
ln_remoteij2007_gsp
40.01***
51.73***
49.31***
51.73***
[4.15]
[4.01]
[4.15]
ln_remoteji2007_gsp
[3.43]
16.30***
15.83***
14.23***
15.83***
[5.20]
[3.56]
[3.69]
[3.56]
0.44***
0.69***
0.69***
0.62***
[0.08]
[0.07]
[0.07]
[0.08]
9.25**
35.86***
34.12***
35.86***
[4.06]
[2.80]
[2.88]
[2.80]
2304
2304
2304
2304
Yes
Yes
Yes
ln_distance
adjacent
Constant
N
Export/Import FE
Yes
Panel B: Manufacturing Sector
Intra-state distance measures
Nitsch
Wolf
Largest Cities
Actual Distance
(I)
(II)
(III)
(IV)
ln_distance
-0.76***
-0.13***
-0.19***
-0.16***
[0.10]
[0.03]
[0.04]
[0.04]
home_bias
1.43***
1.86***
1.90***
1.79***
[0.08]
[0.08]
[0.07]
[0.09]
ln_remoteij2007_gsp
55.30***
70.40***
66.67***
70.40***
[5.56]
[5.38]
[5.56]
ln_remoteji2007_gsp
[4.57]
20.55***
10.10***
7.34**
10.10***
[5.99]
[3.86]
[3.74]
[3.86]
0.34***
0.58***
0.57***
0.51***
[0.08]
[0.07]
[0.07]
[0.07]
5.69
31.97***
29.03***
31.97***
[4.47]
[2.79]
[2.70]
[2.79]
N
49137
49137
49137
Export/Import FE
Yes
Yes
Yes
Source: The Commodity Flow Survey
2007.
Notes: the dependent variable is $ value of shipment.
49137
adjacent
Constant
50
Yes