United States Naval Academy and... · Web viewIrwin, Douglas A. and Kevin H. O’Rourke. 2011....
Transcript of United States Naval Academy and... · Web viewIrwin, Douglas A. and Kevin H. O’Rourke. 2011....
EX TRIDENS MERCATUS* – SEA-POWER AND MARITIME TRADE IN THE AGE OF GLOBALIZATION
Darrell J. Glaser and Ahmed S. Rahman, Department of EconomicsUnited States Naval Academy
“Without commerce the navy would not be needed; without a navy commerce could not exist.” Commodore George M. Ransom, USN 1880.1
The late 19th century witnessed an unprecedented rise in international commerce (O'Rourke and
Williamson 2002). Economic historians are still trying to understand the precise nature of this wave of
globalization - was it due to transport technologies (Harley 1988), or the gold standard (Lopez-Cordova
and Meissner 2003), or shifts in the international system of trade (Irwin and O’Rourke 2011)? But the
rise of military power and its potential influence over global commerce remains under-explored,
particularly for this very crucial period for the histories of world trade and military expansions. How did
the rise of a few hegemonic powers, and the rapidly growing use of the tools necessary for the expansion
of power and influence, affect trade?
Our study uses archival naval data to assess the impact of sea power projection by the major
powers of the time on bilateral trade patterns from 1870 to the precipice of the Great War. Outright wars
can disrupt trade through a variety of channels, through embargos, or privateering activities, or the
fomenting of market uncertainty (Williamson 2011). Naval vessels, an important tool for international
war-making, can conceivably either strengthen or hinder such forces. While the trade-stimulating peace of
the pax Britannica prevailed during this time, naval powers still exerted great influence over trade
patterns, both in positive and negative ways.
Rahman (2010) establishes a general link between naval power and trade for the 18th, 19th and
early 20th centuries. Specifically, fighting war ships tended to lower world trade, even for neutral
countries. On the other hand, neutral or allied ships tended to be a boon to world trade. The paper shows
these naval effects to be both statistically significant and economically meaningful. However, the sea
power-projection measures used are quite aggregative. For each naval power, the author counts the stock
* “from sea-power commerce”
1 from “The Naval Policy of the United States,” United Service 2.1
of capital fighting ships to construct annual time series of potential power projection for each global
power. But these measures do not distinguish between active versus inactive or repairing vessels, nor do
they capture the different kinds of ships deployed or the location of deployed ships. To estimate the trade
impacts of active vessel deployment to a specific region, such distinctions are crucial.
This study provides a test of a particular aspect of the Kindleberger Hypothesis, which states that
hegemonic powers produce public goods that can generate positive spillovers such as peace and
commercial and financial security (Kindleberger 1973, 1981). An alternative view might be considered
the zero-sum game approach of neo-mercantilism, where one hegemon’s commercial security must come
at the expense of another’s (Bartlett 2011). To test these ideas we analyze naval power projection, a
critical tool to promote hegemonic influence, and its effects on world commerce. This approach allows us
to capture the effects of de facto measures of power projection, as opposed to effects from de jure
changes in international policy by hegemons.
Following Rahman (2010), we focus on a much neglected player in the international
infrastructure of commerce: sea powers. We exploit the unprecedented degree of detail concerning naval
activities to establish more precisely the links between the activities of particular naval powers and global
commerce. We construct power metrics using various naval vessels, which vary over country of origin,
region of deployment, and time. To these measures we link bilateral trade data (which vary by country-
pair and year), and a variety of controls important in explaining bilateral trade. We analyze not only how
a naval power's ships stationed in a particular region can affect trade between two regions, but also
differences between the effects on a naval power's own trade and the effects on other countries’ trade.
The distinction is important, as a navy’s effect on commerce may be considered a private good for the
naval power, but a (sometimes quite expensive!) public good for all others. That sea-power has been used
as a national defense strategy to protect one's own trade and commercial interests remains fairly
uncontroversial.2 But the public good nature of sea-power can create a host of potential international
externalities that may either help or harm the trade of other nations.
This work joins the group of papers analyzing the effects of military power and trade. One branch
of analysis considers the effects of international conflict on trade.3 Another branch analyzes the transport
2 See for example Lewis (1959), Crowhurst, (1977), Harding (1999).
3 Results from this body of work are mixed. Bergeijk (1994), Mansfield and Bronson (1997) and Glick and Taylor (2010) estimate gravity models and find that conflict lowers trade; Mansfield and Pevehouse (2000) and Penubarti and Ward (2000) also estimate gravity models but find no statistically significant effects of conflict on trade.
2
infrastructure of trade (Irwin and O'Rourke 2011), of which sea-going navies form an important
component. Our study combines aspects of these literatures to help us further understand the important
factors influencing the international wave of commerce of the latter 19th century.
Of course, capturing the causal effects of naval power projection on trade is complicated by the
fact that naval deployment is in part motivated by concerns over trade. During this period of relative
peace, navies were often considered “pioneers of commerce.”4 But how navies responded to trade flows
remains unclear - naval powers could endeavor to protect their own trade, but could also seek to disrupt
the trade of rival powers. To address endogeneity, we employ a two-stage strategy. First, we develop a
model of naval power projection. A country deploys naval capital to different regions around the world
for many motivations, including the fact that naval capital might have been deployed there by a rival
power. Thus, we develop a simultaneous equations model, where naval deployment to a certain region at
a certain time is jointly determined by all major naval powers. We identify the system using a number of
country-specific variables that we argue are orthogonal both to the naval deployment of a rival power in a
particular region, and to bilateral trade between any two particular regions. This “arms race” model
produces estimated measures of naval power deployed around the world.5
In the second stage, we incorporate these estimates in a gravity trade model. Following Glick and
Rose (2002) we construct a gravity model with panel data using country-pair fixed effects estimation to
control for any time-invariant country-pair characteristics. The naval power estimates created in the first
stage form our instruments to measure the spillover effects of power projection on commerce. Arguably
they can influence trade between two particular countries but are themselves not influenced by such trade.
Concentrating attention on the spillover effects of navies provides us another view of the causal effects of
military expenditures by hegemonic powers on international trade.
We use this framework to help us answer a number of questions. How have naval powers
influenced the ebb and flow of international commerce in history? More specifically, did the active use of
naval force help spur trade and commerce for that naval power? Further, we can test the “naval corollary”
to the Kindleberger Hypothesis - do we observe naval powers creating the kinds of public goods that
fostered greater trade among third parties?
4 Schufeldt, Robert W. 1878. The Relation of the Navy to the Commerce of the United States - A Letter Written by Request to Hon. Leopold Morse, M.C., Member of Naval Committee, House of Representatives. J.L. Ginck.
5 See Blalock (1985) for an in-depth discussion of use of simultaneous equations in modeling arms races.3
We first compile data on vessels, their stations and their characteristics from the navy registries of
three of the major powers of the time: England, France and the United States. These registry books,
housed in the National Archives and arranged in annual volumes, maintain lists of active naval vessels,
their present duty or station, and basic ship characteristics such as rate, number of guns, ship personnel,
and displacement (in tonnage).
To this we merge a number of other data series. Bilateral trade data are assembled from two main
sources: Barbieri (1996) and Mitchell (1992, 1993, 1998). The Barbieri (1996) dataset contains bilateral
trade for around sixty countries during the period 1870 – 1947. Data here typically measure bilateral trade
between countries a and b by summing imports into a from b and into b from a. We augment this with
data from Mitchell (1992,1993,1998) to fill in some of the gaps in Barbieri's coverage from 1870-1913.
Measures of conflict are compiled using data on militarized interstate disputes collected by the
Correlates of War Project (COW) at the University of Michigan. This dataset measures both the incidence
and intensity of hostility at the country level. We code our war variable with conflicts of hostility at
medium to high levels of intensity (these include blockades, occupations of territory, seizures, clashes,
raids, declarations of war, uses of weaponry, and interstate wars).
Finally, a number of other standard variables are added to estimate the gravity model; these
include real GDP, population, and various country-pair characteristics, such as contiguity, distance, and
mutual use of the gold standard. Real GDP and per capita GDP data come predominantly from Maddison
(1995,2001), supplemented where necessary by data from Mitchell (1992,1993,1998). Gold standard data
were cobbled together from Lopez-Cordova and Meissner (2003) and Meissner (2005). The CIA’s World
Factbook is used to provide a number of country-specific variables, including longitude and latitude, land
area, physically contiguous neighbors and common languages.6
The final merged dataset provides us with a method to gauge the global effects of military power
that evolves both spatially and longitudinally. Each country-pair year observation includes instrumented
measures of “naval power.” These are aggregative measures of vessels active in waters through which
commerce between two nations could conceivably flow. While studied and discussed extensively by
naval historians, this rich data on naval vessel deployment has hitherto never been codified, and thus has
never been used in cliometric studies.
6 We use the time-invariant measures only with the random effects version of the gravity model, the results of which we do not report here.
4
Our results provide a number of insights. With our first stage “arms-race” study, we see that the
English and French compete primarily with each other, matching each other's naval deployments. Neither
England nor France tend to react to American deployment, nor does the U.S. react to England or France.
Naval strategies appear to differ in other ways as well. The U.S. (and to a lesser extent England) tend to
deploy more naval resources to international “hot-spots” where conflicts erupt. The French Navy on the
other hand had no such proclivity. Our results are broadly consistent with the different naval strategies
among the global powers (England’s Bluewater School, France’s Jeune Ecole, and America’s Mahanian
doctrine).
Using our naval instruments generated from this exercise in the gravity model, we discover that
the English and American navies had something else in common - they were both promoters of global
commerce, bolstering not only their own trade, but also the trade of other countries. The French Navy on
the other hand promoted its own trade but tended to disrupt the trade of others. These results demonstrate
that whether the Kindleberger Hypothesis held during this time depended on the hegemon and its strategy
vis-a-vis the world.
BACKGROUND
In the second half of the nineteenth century tensions between states found a new expression through arms
races. The mid-nineteenth century naval race between Britain and France was in fact the first example of
an arms rivalry between modern societies. The race was particularly fierce in the latter 19th century,
despite (or perhaps due to) the extended peace of the Pax Britannica. The race effectively ended in 1912
when the exchange of the Grey-Cambon letters ushered a new era of naval cooperation between Britain
and France (Williamson 1969). Before that point however an enormous global naval infrastructure had
been erected. The England-France arms race had contributed more than any other factor to the emergence
of the modern battleship, “the most complicated machine of the nineteenth century” (Hobson 2002), and
the primary mode of naval power projection (Modelski and Thompson 1988).
This paper suggests that the naval arms race happened for many political and strategic reasons
unrelated to commerce protection, shaping our empirical study in the next section. Certainly navies were
used to protect and open market access. But peacetime navies were also used to police international
waters, improve the safety of navigation, enforce neutral rights, attempt to expand scientific knowledge,
and generally protect sovereign interests and “show the flag” (Bartlett 2011). New overseas bases were
sought not necessarily for their intrinsic economic value, but as potential stepping stones to more
5
important places. Indeed existential threats were at times manufactured to spur greater naval expenditure
(Glaeser 2009). And with industrialization, naval power increasingly became an offense weapon by which
hegemons could exert pressure on modern industrial nations (Kennedy 1991).
Throughout this period England led the race, keeping well ahead of its closest rivals by spending
around £5M annually through the end of the 19th century.7 Part of its goal was to persuade its primary
rival France that they could never win a naval competition with them. Not only did this not have the
expected effect on France, it also helped spur the 1898 Navy Law, committing Germany to building a
new navy to directly compete against the Royal Navy (Hobson 2002).
During this time British defense strategy was under the so-called Bluewater School, which
emphasized dominant command of the sea. The necessary precondition of such superiority was to place
its vessels near the ports of foreign regions. This was deemed the most cost-effective way for England to
defend her hegemonic position. To accomplish this legislation such as the Naval Defense Act and the
Spencer Program were initiated and effectively created a massive naval buildup.8
This was in sharp contrast to France’s Jeune Ecole, the tacit recognition that France could not
compete with England one for one in ship building and deployment (as we will empirically see in the next
section). Rather its navy would focus on England’s (and other rival powers’) apparent vulnerability of
international commerce dependence. A total command of the sea was not required for commerce warfare,
and the opportunities for such warfare expanded rapidly as trade around the world grew (Ropp 1971). Of
course this type of commerce-raiding can impose heavy costs, not just in the form of lost or captured
merchandise - trade can be diverted through less expeditious or efficient routes. And merchants and
sailors demanded higher compensation for ex ante threats to their voyages (Hilt 2006).
Finally, while the U.S. had a much smaller fleet, its strategy slowly evolved into one similar to
that of England's. U.S. naval officers during the 1870s and 80s were ambivalent toward overseas
expansion, but once the new steam naval buildup was launched in the late 1880s they became outspoken
proponents of territorial acquisition and global expansion (Bartlett 2011). At the initiation of its
modernization program the U.S. Navy ranked 12th among the powers in terms of size; by 1900 it had
advanced to third place (McBride 2000).7 Source: http://www.cityofart.net/bship/gunnery.html
8 The Naval Defense Act established the standard for England to maintain its number of battleships to be at least as many as those from the next two largest navies (which at that point were France and Russia). This greatly contributed to the arms race, though of course the original objective was to prevent the race in the first place. The Spencer Program was another expansion aimed to match foreign naval growth.
6
Much of this expansion can be attributed to Alfred Mahan, whose naval doctrine might be called
a “theory of mercantilistic imperialism” (Gough 1991). The naval expansion he championed was tightly
linked to his unwavering support for overseas territorial and economic expansion. It was Mahan who
introduced to the U.S. the idea of naval use during peacetime to generate economic benefits from
command of the sea, an idea long championed in England. One question this paper raises is whether
“mercantilistic” aptly describes the Navy, or if instead greater naval strength produced positive spillovers
to others. Did international commerce improve as active American vessels were increasingly used in
peacetime missions like defending the Monroe Doctrine (Brown 1881)?
ESTIMATING MEASURES OF NAVAL POWER PROJECTION – AN “ARMS RACE” MODEL
From the history of 19th century naval power, we see that naval expansion by hegemons likely had
profound trade impacts. But testing the actual impact of naval deployment is complicated since it was
itself motivated, at least in part, by concerns over commerce. We tackle the problem in two stages.
For the first stage, we develop a naval arms-race model by estimating a simultaneous equations
model (SEM). Specifically, let Zcst be a measure of naval power deployed by country c in region (station)
s in year t. These will be measures of gross ship tonnage of vessels deployed to region s, or alternatively
measures of total ship personnel. Our empirical specification for naval power c is
(1)
for c = {England, France, U.S.}, and l = {England, France, U.S.}. Thus naval deployment by a naval
power to region s will be a function of the deployments to the same region of the other L navies being
considered, and of M exogenous factors (x) which influence the deployment of vessels by the naval
power. As we are considering three navies, L here will be two.
These power projections are jointly determined. Since the structural errors εuk, εfrn, and εus will be
correlated with each other, we estimate this system using simultaneous equations.9 This requires an
IV/2SLS approach, where we initially regress country c's naval projection on all exogenous variables (xc),
produce predicted values for zc, and then estimate (1) using our predicted measures of the naval
deployments of rival naval powers. The resulting final estimates will then be used as exogenous variables
in our gravity trade model.
9 Durbin-Wu-Hausman tests suggest that this is an appropriate empirical approach.7
To estimate (1), we require measures of power projection, and measures of exogenous variables
which influence these projections. By “naval projection,” we mean the extent to which navies deploy
forces to specific regions. Because we have data on the ships stationed abroad for each naval power, we
are able to construct such metrics. We will alternate between measuring the total tonnage of these vessels
deployed to a region, and the total officer personnel assigned to these deployed vessels (scatter plots of
these measures for England and France are depicted in Figure 1).
Estimating the simultaneous system of equations also requires having exogenous variables that
influence naval power c's deployment of vessels and personnel to a region. The first variables we consider
vary only either over time (naval expenditures) or over region of deployment (distance from country c to
region s). Naval expenditures are aggregative spending on nearly all areas of naval operations and are
measured in 1913 British pounds.10 Use of expenditures in the first stage suggests that these expenditures
do not directly influence trade between any two countries. There is much historical evidence to support
this. Hagen (1991) demonstrates how the myriad of random international and domestic events influenced
naval budgets. Schulman (1995) goes further, showing for the U.S. case that domestic factors such as
political and cultural changes were the primary causes of new naval policies during this period. Sexton
(1976) carefully analyzes Congressional voting records, suggesting that changes in congressional
representation had profound influences in expenditures as naval appropriation bills passed or failed nearly
completely on party-line votes. Indeed few in Congress really understood the broad strategic implications
of their votes, let alone how naval expenditures would influence deployments to particular regions around
the world (Sprout and Sprout 1939).
We also use a naval expenditure-distance cross term to gauge how random naval budget
allocations interacted with deployment to particular regions. Along with these we use a number of other
explanatory variables that relate to conflict in particular regions. As each navy had rather different
strategies regarding its role in policing international waters, disruptions occurring in different places in
the world would potentially have different naval responses. Finally we also use information on the
number of allies that country c has located in region s, to gauge how allegiances might influences naval
deployments.
10 Alternatively we use aggregative total military expenditures, taken from the Correlates of War's National Materials Capabilities dataset (Singer et. al 1972). Arguably this provides an even cleaner variable to gauge the effects of military budgets on naval power deployments. Not surprisingly, the correlation is somewhat weaker, but the estimates produced from these measures, and the resulting effects on trade (results not reported), closely mirror those presented here.
8
These first-stage exercises accomplish a number of things. It provides a new quantitative
assessment of the first “arms-race” among hegemonic powers in modern history. It allows us, in an
empirically robust way, to observe who “competed” against whom. More critically to the overall
objective of the paper, it produces instruments for naval power projections that we use in the next section
to assess the causal effects of such projections on world trade.
We use original-source data for three naval powers. These include the number of ships deployed,
tonnage, and personnel aboard each vessel for England (1878-1914), France (1875-1913), and the United
States (1870-1911).11 For power projection measures, we alternatively use total ship tonnage
(displacement) and total number of officer personnel aboard vessels. We do not use total number of guns
on ships, or the horsepower of ships, because these measures are not available for all naval powers
considered, and because the degree of comparability across naval powers is limited.
Observations are at the region(station)-year level (s,t). Regions are the North Atlantic, South
Atlantic, North Pacific, South Pacific, North Sea (Europe), the Mediterranean, the Asiatic front, and the
Indian Ocean. In these specifications the home station is also included - these are active vessels stationed
at various home ports either patrolling local waters or preparing for international tours (our results are
quite similar to those presented here when we exclude home stations).
Table 1 summarizes our findings when measuring naval deployment in gross tonnage of ships.
We estimate deployment for England, France and the U.S. in separate specifications. The first variables
listed are considered “exogenous” (x’s) and influence the gross tonnage of deployed vessels to particular
regions. Note that data on naval expenditures evolve only over time. These involve funding for all naval
activities, including paying personnel, building and repairing vessels, and research and development.
Distance (measured as nautical miles between country c and the center of naval station s) evolves only
over station.
The first specification for each country are estimated by simple OLS. Invariably, aggregative
naval expenditures at home are positively related to naval deployment to station s but negatively related
to these expenditures interacted with distance from the naval power to the naval front. This makes sense -
expenditures will allow greater deployment of the fleet, but such deployment will take time to reach far
off stations. Naval money turned into vessels is first deployed in local waters, and over time are sent out
to distant seas.11 The gaps in coverage are for full years (certain annual volumes of navy registries were unavailable). For years that are covered, the data coverage is comprehensive and complete - we capture every active vessel, where it is deployed, and basic ship characteristics.
9
We also look at the effects of hostilities erupting in different regions. Estimates of hostility
effects are very imprecise. But they still provide interesting insights into the decisions of each naval
power to deploy naval resources. As mentioned earlier, both England and the United States had ambitions
for global naval reach (granted the U.S.’s ambitions sprung relatively late in the game), and part of that
objective meant that their navies would police international waters. In fact evidence suggests that both
England and the U.S. tended to deploy vessels to international “hot-spots.” France had no such policing
ambitions, and indeed the French Navy tended to avoid such areas. All these efforts tend to diminish over
the length of the conflict, as indicated by the hostility (days, lagged) term. It also tends to be the case that
naval powers deploy more naval resources to these hotspots if their allies are also engaged in the conflict.
The latter specification for each country come from the simultaneous equation model (SEM),
where the variables above are used as observed shifters of naval power that allow us to estimate the
deployment of vessels by multiple naval powers simultaneously. So looking at specification (2) in table 2,
for instance, we simultaneously estimate all nations’ deployment, using English, French and U.S. naval
expenditures, distances, expenditure-distance cross terms and hostility measures as exogenous variables
(shifters). The estimates are the measured effects of French and U.S. deployment on English deployment.
From these results it is clear that the English and French navies responded heavily to each other's
naval deployments. For each ton the French deployed, the English tended to respond about 160-170
percent in kind. The French on the other hand tended to send out roughly one-third to two-thirds of the
naval capital that England was able to deploy. Given the disparate naval capabilities of each nation, these
numbers make sense. Neither power however responds to the naval deployments of the United States. The
U.S. likewise appears to be unaffected by the deployments of England or France - the SEM model
appears to add little explanatory value for the U.S. case. These results confirm the general suggestions of
naval historians that the first “modern arms-race” was one dominated by England and France.
Finally, we also perform the same empirical exercises, but use total number of line officers
assigned to vessels as an alternative measure of naval power projection. Similar results emerge (results
not reported). We use these as alternative predicted measures of power projection in the next section.12
EFFECTS OF NAVAL POWER PROJECTION ON BILATERAL TRADE
12 Alternatively we estimate all these using random effects, as well as station fixed effects - results remain consistent in signs and magnitudes.
10
We now want to see how the naval arms-race among the super-powers of the late 19th century affected
global commerce. To that end we use the estimates from our previous exercise in a panel gravity model,
the workhorse of empirical trade literature (for what follows, we use results from specifications 2,4 and 6
in Table 1). That is, we use predicted measures of naval projection (alternatively using gross vessel
tonnage and total number of line officers) as separate explanatory variables in our gravity model. We thus
consider these measures as potential spillover effects from naval build-up on commerce.
Measures of naval power projection are given by the following dot product:
(2)
where navycst is a 1-by-N vector of naval power projection measures for naval power c at time t across N
different regions, and sij is a 1-by-N spatial vector, comprised of ones and zeros denoting the relevant
regions through which conceivably maritime trade between countries i and j would pass.13 In producing
our power measures, we use predicted values of navycst that we produce from the above exercises -
specifically we use estimated values of equations described by (1), which simultaneously estimate power
projection by England, France and the United States.14
We use estimated values of naval power, as opposed to directly observed values, for several
reasons. The first is simultaneity - trade between countries might motivate hegemons to send naval capital
to certain regions. We want to use variations in naval deployment not directly related to bilateral trade.15
The second reason is potential omitted variable bias - if naval deployment by one power is heavily
influenced by the deployment of a rival power (as appears to be the case for England and France), then
the measure for any single naval power will be correlated with the error term. By using naval power
13 Once again, there are nine possibilities - the home station, the North and South Atlantic, the North and South Pacific, the North Sea, the Mediterranean, the Asiatic front, and the Indian Ocean.
14 Note that the data used to estimate the system of equations described in (1) vary by station and year, while data used to estimate the gravity model vary by country-pair and year.
15 One might question the use of conflict measures in the first stage. It should be noted that these measures of conflict are for disruptions confined to particular spots, which will have limited trade disruption between any two countries. Vessels on the other hand can be attracted to such disruptions, and because they are highly mobile can have more influence on commercial flows through the region. Thus we suggest that regional conflict is likely to influence trade only through the indirect channel of attracting naval vessels.
11
estimates from the simultaneous equation model, we gauge the potential trade effects from each naval
power in isolation.
We have two alternative series of spatial vectors s’s. The first considers any region through which
trade between countries i and j would pass as fair game (takes on a value of 1). The second considers only
the regions where i and j are located - it thus ignores in-between regions, and so considers two regions at
most. Results below are only for the first case, but results for the second echo these quite closely (not
reported).
Further, we use two alternative estimates of naval power projection, for each estimate captures
somewhat different aspects of power. For example, estimates of tonnage capture the capital intensity of
deployment and by proxy the extent of naval technology, since the size and displacement of a vessel was
highly correlated to its technological sophistication (Bennett 1896, Glaser and Rahman 2012). Personnel
measures on the other hand capture the labor-intensive part of naval deployment.
All gravity models include bilateral-pair fixed effects and year effects. Results from these
exercises are displayed in tables 2-4. For each naval power we look at deployment effects on the
country’s own trade, and the effect on other countries’ (third party) trade.16
Table 2 show results for England. Perhaps not surprisingly, English naval deployment, when
measured either by tonnage or officer personnel, is a positive boost to English trade. Perhaps more
striking are the results from specifications 3 and 4, which suggest that English power projections created
positive spillovers for global commerce. This makes sense, England being the dominant force in the
world, essentially serving as the global police. The magnitudes of these effects are modest. For example,
a one percent increase in naval tonnage of deployed vessels generates a 0.04 percent increase in trade
volumes. But it bears repeating that these are spillover effects, the unintended consequences of the naval
arms race. Thus it appears that Kindleberger’s notion of the stability of hegemonic influence does have
some validity when we consider the Royal Navy.
In table 3 we turn our attention to France. It appears French naval power positively influences its
own commerce. This would be consistent with the historical record, as France was bent on protecting its
commercial interests. We consider the effects of the French navy on the trade of other countries (those
that do not involve France). Here the French navy appears to be a destroyer of commerce. As we
16 We should also mention that we include cross effects of war and naval power, with no meaningful effects (not reported). We also redo all exercises with bilateral random effects instead of fixed effects (not reported); results closely resemble those presented here.
12
mentioned earlier, this is consistent with a Jeune Ecole strategy of commerce raiding. It thus appears that
its navy, in protecting its own commercial interests, did so at the expense of the commerce of the world.
Finally Table 4 demonstrates the effects of U.S. naval power on commerce. The U.S. appears to
help both its own trade and produce positive spillovers. Moreover, given the U.S.’s rather dramatic shift
in policy towards more muscular international involvement and greater naval development after 1890, we
would expect naval capital to exhibit greater effects on third-party trade during the latter part of the
sample. This is indeed the case (results not shown). This would be consistent both with a Mahanian
strategy and Kindleberger's notion of America's rising hegemonic influence around the world.17
CONCLUSION
This paper has demonstrated the effects of naval power on bilateral trade flows during the latter 19th and
early 20th centuries. It has done so by first developing an arms-race model of naval power projection, and
using estimates of these as instruments in the trade model.
It is important for us to get a full picture of the events and forces that gave rise to the
unprecedented globalization of the late 19th century. We suggest in this paper that a worldwide naval
infrastructure helped to slowly create it, and of course with alarming rapidity then helped to decimate it.
Therein lies the complicated story of naval power's role in Kindleberger’s Hypothesis - the tools of global
hegemony can stabilize or disrupt commerce. It depends on how they are wielded.
A final analogy might prove helpful. A police force is detailed to an area with certain goals
(lower crime, increase security, etc.), but of course it would not have been detailed there at all were it not
for the rise of criminals bent on undermining those goals. Similarly, Kindleberger’s tacit suggestion that it
takes strong hegemonic support to promote global stability is a sound one. The rub is that it may take the
aggressive de-stabilizing effects of weaker powers (a.k.a. France) to illicit such support in the first place.
17 We also include all three naval power measures simultaneously for trade involving countries other than England, France or the U.S. (not reported). Similar trade patterns emerge.
13
REFERENCES
Barbieri, Katherine. 1996. Economic Interdependence and Militarized Interstate Conflict, 1870-1985. Ph.D. dissertation, SUNY Binghamton.
Bartlett, Lawrence Wood, 2011. “Not Merely for Defense: The Creation of the New American Navy, 1865-1914.” Ph.D. dissertation.
Bennett, Frank M. 1896. The Steam Navy of the United States - A History of the Growth of the Steam Vessel of War in the Navy, and of the Naval Engineer Corps. Pittsburgh, Pa: Warren Co. Publishers.
Bergeijk, Peter van. 1994. Economic Diplomacy, Trade, and Commercial Policy: Positive and Negative Sanctions in a New World Order. Vermont: Edward Elgar.
Blalock, H.M., ed. 1985. “Simultaneous Equations Techniques,” Causal Models in the Social Sciences, Piscataway, NJ: Transaction Publishers.
Brown, R.M.G. 1881. “The Commercial and Naval Policy of the United States.” United Service.
Crowhurst, Patrick. 1977. The Defense of British Trade, 1689–1815. Folkerstone: Dawson and Sons.
Glaser, Darrell J. and Ahmed S. Rahman. 2012. “Naval Engineering and Labor Specialization during the Industrial Revolution.” USNA working paper.
14
Glick, Reuven and Andrew Rose. 2002. “Does a Currency Union Affect Trade? The Time-Series Evidence.” European Economic Review 46(6): 1125-1151.
Glick, Reuven and Alan M. Taylor. 2010. “Collateral Damage: Trade Disruption and the Economic Impact of War.” Review of Economics and Statistics 92(1): 102-127.
Gough, Barry M. 1991. “The Influence of History on Mahan.” in Hattendorf: The Influence of History on Mahan.
Hagen, Kenneth. 1991. The People's Navy.
Harley, C. K. 1988. “Ocean Freight Rates and Productivity, 1740-1913: The Primacy of Mechanical Invention Reaffirmed.” Journal of Economic History 48(4): 851-76.
Harding, Richard. 1999. Seapower and Naval Warfare from 1650–1830. London, U.K.: University College London Press.
Hilt, Eric. 2006. “Incentives in Corporations: Evidence From The American Whaling Industry,” Journal of Law and Economics. 49(1), 197-227.
Hobson, Rolf. 2002. Imperialism at Sea: Naval Strategic Thought, the Ideology of Sea Power, and the Tirpitz Plan, 1875-1914. Boston and Leiden: Brill Academic Publishers.
Irwin, Douglas A. and Kevin H. O’Rourke. 2011. “Coping with Shocks and Shifts: The Multilateral Trading System in Historical Perspective.” NBER working paper no. 17598.
Kennedy, Paul. 1991. The Rise and Fall of British Naval Mastery. London.
Kindleberger, Charles P. 1973. The World in Depression. Berkeley: University of California Press.
Kindleberger, Charles P. 1981. “Dominance and Leadership in the International Economy: Exploitation, Public Goods, and Free Rides.” International Studies Quarterly 25(2): 242-254.
Lewis, Michael. 1959. History of the British Navy. Fair Lawn, NJ: Essential Books.
Lopez-Cordova, J. E. and C. M. Meissner. 2003. “Exchange Rate Regimes and International Trade: Evidence from the Classical Gold Standard Era.” American Economic Review 93(2): 344-53.
Maddison, A. 1995. Monitoring the World Economy, 1820-1992. OECD, Paris.
Maddison, A. 2001. The World Economy: A Millennial Perspective. OECD, Paris.
Mansfield, Edward D. and Rachel Branson. 1997. “Alliances, Preferential Trading Arrangements, and International Trade.” American Political Science Review 91(1): 94-107.“Mansfield, Edward D. and J.C. Pevehouse. 2000. Trade Blocs, Tade Flows, and International Conflict.” International Organization 54: 775-808.
15
McBride, William M. 2000. Technological Change and the United States Navy, 1865-1945. Baltimore & London: Johns Hopkins University Press.
Meissner, Christopher M. 2005. “A New World Order: Explaining the International Diffusion of the Gold Standard, 1870-1913,” Journal of International Economics 66: 385-406.
Mitchell, Brian R. 1992. International Historical Statistics: Europe, 1750-1988. New York: Stockton Press.
Mitchell, Brian R. 1993. International Historical Statistics: The Americas, 1750-1988. New York: Stockton Press.
Mitchell, Brian R. 1998. International Historical Statistics: Africa, Asia and Oceania, 1750-1988. New York: Stockton Press.
Modelski, George and William R. Thompson. 1988. Sea-power in Global Politics. London: Macmillan.
O’Rourke, Kevin H. and Jeffrey G. Williamson. 2002. “When Did Globalization Begin?” European Review of Economic History 6: 23-50.
Penubarti, Mohan and Michael Ward. 2000. “Commerce and Democracy.” Center for Statistics and the Social Sciences working paper no. 6, University of Washington.
Rahman, Ahmed S. 2010. “Fighting the Forces of Gravity – Sea-power and Maritime Trade between the 18th and 20th Centuries” Explorations in Economic History 47(1), 28-48.
Ropp, Theodore. 1971. “Continental Doctrines of Sea Power,” in Edward Burrell's Makers of Modern Strategy - Military Thought from Michiavelli to Hitler.
Schulman, Mark R. 1995. Navalism and the Emergence of American Sea Power, 1882-1893, Annapolis, MD: Naval Institute Press.
Sexton, Donald J. 1976. “Forging the Sword: Congress and the American Naval Renaissance, 1880-1900.” Ph.D. dissertation, University of Tennessee.
Singer, J. David, Stuart Bremer, and John Stuckey. 1972. “Capability Distribution, Uncertainty, and Major Power War, 1820-1965.” in Bruce Russett (ed) Peace, War, and Numbers, Beverly Hills: Sage, 19-48.
Sprout, Harold, and Margaret Sprout. 1939. The Rise of American Naval Power, 1776-1918. Princeton, N.J.
Williamson, Jeffrey G. 2011. Trade and Poverty: When the Third World Fell Behind. Cambridge, MA: MIT Press.
Williamson, Samuel R. 1969. The Politics of Grand Strategy - Britain and France Prepare for War, 1904-1914. London.
16
Figure 1 – England vs. France – Naval Power Projections
050
000
1000
0015
0000
Fren
ch s
hips
0 100000 200000 300000English ships
by station-yearTotal Tonnage of Vessels Assigned to Stations
17
010
020
030
0Fr
ench
nav
al o
ffice
rs
0 100 200 300 400English naval officers
by station-yearTotal Naval Officers Assigned to Stations
Table 1 – Estimating Aggregative Tonnage of Naval Vessels Deployed
u.k. u.k. frn frn u.s. u.s.Variable (1) (2) (3) (4) (5) (6) OLS SEM OLS SEM OLS SEMnaval expenditures 4484.6*** 2159.8*** 2962.4** -0.099 2792.2*** 3927.2*** (893.3) (447.9) (1428.7) (0.07) (870.6) (975.0)distance 10.7*** -0.06 3.79** 0.29 0.51 0.40 (3.26) (1.72) (1.86) (1.36) (0.87) (1.90)exp*distance -0.46*** -0.14*** -0.29 0.11 -0.28** -0.38*** (0.11) (0.05) (0.15) (0.12) (0.13) (0.12)hostility (indicator, lagged) 16938.0 7531.2 -502.6 -553.8 23967.2* 36045.2*** (15743.3) (9962.0) (13966.9) (6824.8) (12738.2) (10623.0)hostility (days, lagged) -222.95*** -116.2** 86.6 48.5 -40.8 -91.7 (73.1) (57.3) (74.6) (40.1) (64.8) (65.3)number of allied countries (lagged) 61718.4*** 10313.0 31041.8*** 6176.5 73764.6*** 53221.8***
18
(16140.1) (7727.9) (11120.7) (4864.6) (153.34.4) (21324.5)hostility (days by allies, lagged) 106.8*** 13.8 21.6 2.41 207.1*** 219.7* (27.3) (31.7) (27.1) (19.0) (59.1) (114.9)English navy deployment - - - 0.38*** - -0.20 (0.12)French navy deployment - 1.74*** - - - 0.18 (0.21) (0.28)U.S. navy deployment - -0.05 - -0.10 - - (0.14) (0.065) R-squared 0.47 0.62 0.26 0.58 0.33 0.27F-statistic 17.5*** - 4.15*** - 30.97*** -Chi-squared - 443.6*** - 171.4*** - 119.5***Observations 315 234 288 234 369 234
Notes: Dependent variable is total tonnage of naval vessels deployed in a region. Standard errors clustered by country reported in parentheses with *10%, **5% and ***1%.
Table 2 – Effects of English Naval Deployment on Trade
Variable (1) (2) (3) (4)
own trade
own trade 3rd-party trade 3rd-party trade
ln(YiYj) 0.61*** 0.61*** 0.48*** 0.46*** (0.10) (0.10) (0.09) (0.09)ln(yiyj) 0.20 0.19 -0.047 -0.02 (0.17) (0.17) (0.15) (0.16)War -0.25*** -0.25*** 0.005 0.01 (0.09) (0.09) (0.12) (0.12)Global War 0.10 0.10 0.01 0.007 (0.07) (0.07) (0.04) (0.04)Gold standard -0.05 -0.05 0.13*** 0.13*** (0.04) (0.04) (0.04) (0.04)
0.07** - 0.04*** -
19
(0.03) (0.01)
- 0.06** - 0.02** (0.03) (0.01)R-squared 0.45 0.45 0.21 0.21Observations 1012 1012 3813 3813Number of country-pairs 40 40 222 222
Notes: Dependent variable is logged bilateral trade. (1) and (2) uses trade between England and another country. (3) and (4) uses trade between two non-U.K. countries.
Final two variables are estimates of naval power projected by England generated from SEM model.
Yi denotes the GDP of country i; yi denotes the GDP per capita of country i.
All specifications include bilateral pair and time fixed effects.
Standard errors clustered by bilateral country pair reported in parentheses with *10%, **5% and ***1%.
Table 3 – Effects of French Naval Deployment on Trade
Variable (1) (2) (3) (4)
own trade
own trade 3rd-party trade 3rd-party trade
ln(YiYj) 0.19 0.21 -0.15 -0.17* (0.17) (0.17) (0.09) (0.09)ln(yiyj) 0.34 0.33 1.02*** 1.05*** (0.28) (0.28) (0.14) (0.14)War 0.003 0.014 -0.07*** -0.07*** (0.13) (0.13) (0.03) (0.03)Global War -0.09 -0.09 -0.006 -0.006 (0.08) (0.08) (0.08) (0.08)Gold standard -0.08 -0.09 0.17*** 0.16*** (0.06) (0.06) (0.03) (0.03)
0.15*** - -0.09*** -20
(0.05) (0.02)
- 0.10*** - -0.17*** (0.04) (0.02)R-squared 0.49 0.52 0.07 0.07Observations 673 673 4886 4886Number of country-pairs 27 27 235 235
Notes: Dependent variable is logged bilateral trade. (1) and (2) uses trade between France and another country. (3) and (4) uses trade between two non-France countries.
Final two variables are estimates of naval power projected by France generated from SEM model.
Yi denotes the GDP of country i; yi denotes the GDP per capita of country i.
All specifications include bilateral pair and time fixed effects.
Standard errors clustered by bilateral country pair reported in parentheses with *10%, **5% and ***1%.
Table 4 – Effects of U.S. Naval Deployment on Trade
Variable (1) (2) (3) (4)
own trade
own trade 3rd-party trade 3rd-party trade
ln(YiYj) 0.009 -0.17 0.71*** 0.71*** (0.18) (0.17) (0.11) (0.11)ln(yiyj) 1.73*** 1.86*** 0.02 -0.02 (0.31) (0.06) (0.17) (0.12)War 0.09 0.09 -0.20* -0.20* (0.13) (0.13) (0.12) (0.12)Global War -0.05 -0.06 0.09* 0.10* (0.09) (0.09) (0.05) (0.06)Gold standard -0.11* -0.11* 0.23*** 0.21*** (0.06) (0.06) (0.04) (0.04)
21
0.08*** - 0.01** -(0.014) (0.005)
- 0.08*** - 0.10*** (0.02) (0.02)R-squared 0.19 0.09 0.25 0.23Observations 855 855 2923 2923Number of country-pairs 42 42 279 279
Notes: Dependent variable is logged bilateral trade. (1) and (2) uses trade between the United States and another country. (3) and (4) uses trade between two non-U.S. countries.
Final two variables are estimates of naval power projected by the U.S. generated from SEM model.
Yi denotes the GDP of country i; yi denotes the GDP per capita of country i.
All specifications include bilateral pair and time fixed effects.
Standard errors clustered by bilateral country pair reported in parentheses with *10%, **5% and ***1%.
22