The Interwar Period International Trade in Arms: A New Dataset

Introduction

With the Russian invasion of Ukraine, international arms transfers have re-emerged as a key topic in international relations. Which countries do (not) send weapons to Ukraine has regularly made news reports. The conflict led Germany, a global Top-5 arms exporter, to revise its policy on shipping armaments to conflict zones. But even before, research on the arms trade has undergone something of a renaissance as numerous recent publications study its drivers as well as consequences from a multitude of theoretical and empirical perspectives (e.g. Grant 2018; Mehrl and Thurner 2020; Pamp et al. 2021; Spindel 2023; Thurner et al. 2019; Willardson and Johnson 2022). These contributions enhance our knowledge about the arms trade at a time when its policy relevance is becoming very clear once again.

At the same time, our systematic understanding of arms transfers remains restricted by data limitations. The temporal coverage of existing arms trade datasets is confined to the period since 1950 or, in the case of small arms and light weapons (SALW), only to a few selected years after the end of the Cold War. This affects our understanding of the role of arms transfers in the lead-up to historically important events such as World War II. But it also implies, arguably more importantly, that our empirical insights about the arms trade rest very much on the bipolar Cold War and unipolar post-Cold War international systems. These were dominated, respectively, by two and then a single Superpower, but preceded, and increasingly also followed (Posen 2011; Schweller and Pu 2011), by multipolar systems with more than two countries at the top. There is an extensive literature on the consequences of polarity and the balance of power in the international system, often rife with disagreement, but most scholars agree that these configurations are associated with differences in states’ behavior, alliance-making, and systemic stability (see e.g. Braumoeller 2012; Jervis 2009; Niou, Ordeshook, and Rose 1989; Wagner 1993; Waltz 1979). This suggests that what we know about arms transfers in uni- and bipolar systems will not necessarily generalize to the multipolar settings characteristic of the nineteenth century, the Interwar years or, possibly, contemporary and future international politics. From this perspective, studying the international relations of the period before the systemic shift induced by World War II can be understood as benefiting our knowledge not just of historical but also current and future events. So far, however, we lack the arms trade data necessary to engage in this research.

We introduce the Interwar Period International Trade in Arms (IPITA) dataset to address this gap. IPITA offers comprehensive, dyadic data on the arms trade in the Interwar years. It covers transfers of military aircraft, ships, armoured fighting vehicles, as well as small arms and light weapons including artillery, ammunition, and explosives between independent countries, as well as some colonies and non-state actors, for the period 1920-39. In the following, we discuss how these data were collected, provide some descriptive insights on arms transfers between the two World Wars, and apply the new data to study how weapons imports and exports affected the Interwar demand for military expenditures. These results join other early evidence in showing how the structure, drivers, and consequences of weapons transfers in the multipolar Interwar years differed from those observed in the bipolar Cold War period, but also the unipolar years after its termination, which dominate our understanding of the arms trade (Mehrl, Seussler, and Thurner 2022, 2023). Before introducing the IPITA data, we start by discussing what data on international arms transfers currently exists.

Existing Arms Trade Datasets

Without a doubt, the most prominent and commonly used dataset on international arms transfers is the SIPRI (2023) Arms Transfers Database. The SIPRI data contain dyadic information on major conventional weapons (MCW) transfers, beginning in the year 1950, fully rely on open source information, and are updated yearly. The data specify the sender and recipient, the delivery and, where possible, order date, the weapon type, the number of transferred weapons, as well as their Trend-Indicator Value (TIV) for a given transaction. The TIV is intended to measure the military or strategic value of a transfer. It is based on the production costs of a set of weapons for which this is known, TIVs for other weapons are then estimated based on their size, performance, technical details, and production year. The large majority of senders and recipients in the SIPRI data are governments, but the data also cover reported transfers involving non-state actors such as rebel groups. In contrast, they do not include information on transfers which occurred clandestinely or were made public by neither the sender, recipient, nor a third-party (see SIPRI 2023). SIPRI data can thus be used to investigate the drivers and consequences of MCW transfers, imports, and exports during the Cold War and post-Cold war periods. Along these lines, existing research uses these data to study how arms imports affect domestic conflict and state repression (Johnson and Willardson 2018; Mehrl and Thurner 2020; Moore 2012; Pamp et al., 2018b; Sullivan, Blanken, and Rice 2020; Suzuki 2007), in what ways weapons transfers influence interstate relations (Beardsley et al. 2020; Krause 2004), whether arms sales are related to defence spending (Blum 2019; Pamp et al., 2018a; Pamp and Thurner 2017), and under which conditions countries trade MCW with each other (e.g. Akerman and Seim 2014; Bove, Deiana, and Nisticò 2018; Comola 2012; Fritz, Thurner, and Kauermann 2021; Martínez-Zarzoso and Johannsen 2019; Thurner et al. 2019; Willardson and Johnson 2022).

However, the SIPRI data do not offer information on transfers of SALW. This information is available from the Norwegian Initiate on Small Arms Transfers (NISAT 2017). The NISAT data specify the sender and recipient, the delivery date, the weapon type, and the financial value of transferred weapons for a given transaction. These data are collected from customs data, most importantly the UN Comtrade Database, and span the period 1962-2017. However, the data are generally considered reliable only for the post-Cold War period and academic applications of it are accordingly limited to these years. The NISAT data thus provide for SALW what the SIPRI data do for MCW, albeit for a shorter time period and with transfer volumes being expressed in terms of financial instead of military value. That being said, the NISAT data have been used somewhat less commonly to systematically investigate and test the drivers and consequences of the small arms trade, with the only published work we are aware of being Baronchelli, Caruso, and Ricciuti (2022), Baronchelli and Caruso (2023),Lebacher, Thurner, and Kauermann (2021), Mehltretter (2022), and Mehrl and Thurner (2020).

Finally, information on arms transfers is also available from the World Military Expenditures and Arms Transfers (WMEAT) series, published first by the United States’ Arms Control and Disarmament Agency and then the Bureau of Arms Control, Verification and Compliance Releases within the State Department. WMEAT data are currently available for the period 1964-2019 and contain financial information on the value of transferred MCW as well SALW. However, the WMEAT data do not distinguish between MCW and SALW and also do not report weapon type-specific information. What is more, the data are not in dyadic format but instead aggregate each importer’s or exporter’s entire yearly transfer activity (see US Department of State 2021). As such, WMEAT data have mostly been used to study the demand for arms imports (Goodhart and Xenias 2012; Smith and Tasiran 2005, 2010).

Especially in combination, the SIPRI, NISAT, and WMEAT datasets provide rich information on the arms trade, allowing researchers to systematically investigate its structure and test both its drivers and outcomes. At the same time, these efforts are currently limited to the years after World War II as a result of data availability. SIPRI, NISAT, and WMEAT cover the Cold War and Post-Cold war periods, meaning that the very large majority of quantitative literature on the arms trade is concerned with these periods. Present datasets do not span previous years, implying that, with one exception, existing research on the Interwar arms trade is unable to engage in hypothesis testing, instead mainly relying on historical methods (e.g. Grant 2018; Harkavy 1975; Krause 1992). The sole exception is Eloranta (2002) who collects SALW transfer data for nine European medium powers and the period 1920-37 to investigate the drivers of both their imports and exports. Earlier, Sloutzki (1941) and Harkavy (1975), respectively, collected global data on SALW and MCW transfers in the 1930s but use them only for descriptive purposes. Unfortunately, the raw data used in these three publications pertain only to a limited set of countries or years, do not offer dyadic information, and, most importantly, are not publicly available. This means that even in terms of descriptive, quantitative assessments of the Interwar arms trade, the data tables presented by Sloutzki (1941) and Harkavy (1975) represent our best knowledge, with later work simply reprinting them (see Krause 1992; Laurance 1992).

There is thus a clear gap regarding systematic, global quantitative data that covers the transfer of SALW as well as MCW in the entire Interwar period at the dyadic level. The IPITA data goes a long way towards filling this gap. IPITA includes dyadic, yearly information on MCW and SALW transfers which can be further distinguished by more specific weapon types and covers the years 1920 until 1937 (SALW) or 1939 (MCW). These data thus facilitate the quantitative study of the drivers and outcomes of arms transfers in the years between the two World Wars.

The IPITA Data: Collection and Construction

The IPITA data consist of two parts, one covering the trade in SALW and artillery and the other MCW transfers. The SALW and artillery data are collected from a series of publications by the League of Nations, the Statistical Year-Books of the Trade in Arms and Ammunition, and cover the period 1920-1937. The MCW data are separate as military ships, aircraft, and armoured fighting vehicles were not covered by the yearbooks, instead had to be collected from a wide variety of primary and secondary sources, and cover the years 1920-1939. As a result, the data collection and construction process for the two parts of the IPITA data was different and we hence discuss them separately. We summarize this process for both parts below and offer more details in the detailed technical report in the supplementary material.

Descriptive Statistics

We now offer some descriptive information on the IPITA data, both as a first perspective on the Interwar arms trade and to hint at the types of analysis they may be used for. For this, we use the minimum reported transfer volumes and, for the SALW transfer-data, focus on information provided by exporters which we augment with importer-reported transfers in case of missingness.

In Figure 1, we differentiate between SALW and three types of MCW, aggregate over all transfers in a given year, and thus show how trade volumes in these four kinds of arms developed in the Interwar years. While ships were most actively traded in the very early 1920s, before the Washington Naval Treaty came into effect, the SALW trade peaked in the late 1920s, and both aircraft and, especially, AFV were traded mostly in the second half of the 1930s. Figure 1 also shows that while the temporal trends of aircraft and AFV transfers were similar as both kinds of arms were the newly emerging military technologies of the period, the former were much more commonly traded than the latter, with only AFV deliveries in the mid-1930s coming close to any of the yearly aircraft transfer totals.OPEN IN VIEWER

Next, Tables 1 and 2 respectively show the top ten exporters and importers for each of the four weapons categories, aggregated over the entire period. The first column of 1 shows the UK to be the top exporter of small arms and light weapons in the interwar period, followed by Belgium, Czechoslovakia, Germany, the US, and France. After a wide gap of 7,000,000 US$, Sweden then takes seventh place, followed by three other small, industrialized countries of western Europe. While the SALW trade was mainly in the hand of European industrialized countries, both big and small, major conventional weapons were mostly exported by the major powers of the time. Accordingly, France and the UK dominated the aircraft and AFV markets while for naval arms sales, the UK exported almost four times the total tonnage of second-placed Italy. It is also noticeable that while numerous countries exported broadly comparable volumes of SALW, the MCW trade was almost entirely in the hands of a few select exporters. Here, smaller countries specializing in a specific MCW also appear, e.g. Czechoslovakia’s role in the AFV market, Dutch exports of aircraft, and Japan, a borderline major power but otherwise not a relevant arms exporter, transferring non-negligible volumes of naval weapons.

OPEN IN VIEWER

Table 1.

Top 10 Arms Exporters, Aggregated Over all Observed Years.

	Exporter	SALW (1000 US$)	Exporter	Aircraft (Number)	Exporter	AFV (Number)	Exporter	Ships (Displacement)
1	UK	41,254	France	3136	France	571	UK	249,248
2	Belgium	33,063	UK	2570	UK	535	Italy	64,236
3	Czechoslovakia	31,147	USA	1390	USSR	441	Germany	20,694
4	Germany	27,862	USSR	1171	Italy	404	France	19,013
5	USA	24,237	Italy	1155	Czechoslovakia	222	USA	17,960
6	France	23,798	Germany	1138	Germany	52	Japan	13,920
7	Sweden	16,016	Netherlands	697	Sweden	52	Spain	9554
8	Denmark	13,932	Poland	199	Poland	36	Australia	7508
9	Spain	11,502	Czechoslovakia	161	USA	29	New Zealand	5400
10	Switzerland	5976	Canada	60	Netherlands	4	USSR	5202

OPEN IN VIEWER

Table 2.

Top 10 Arms Importers, Aggregated Over all Observed Years.

	Importer	SALW (1000 US$)	Importer	Aircraft (Number)	Importer	AFV (Number)	Importer	Ships (Displacement)
1	Australia	20,752	Poland	1337	China	397	Spain	87,262
2	Romania	16,490	China	1201	Poland	204	Australia	56,202
3	China	15,420	Spain	1010	Romania	149	Chile	45,872
4	Argentina	13,663	Romania	866	Spain	149	Argentina	33,304
5	Brazil	12,772	USSR	794	Turkey	138	Canada	19,198
6	Japan	12,445	Japan	442	Hungary	127	Siam	18,652
7	Iran	11,851	Yugoslavia	439	Belgium	120	Italy	16,227
8	Estonia	10,884	Belgium	380	Iran	120	Portugal	14,122
9	USA	9161	Turkey	378	Austria	92	Poland	14,105
10	France	8126	Brazil	327	Siam	84	UK	13,248

Table 2, in turn, presents the period’s top importers of arms. Its first column shows that Australia was the top importer of SALW in the Interwar period and, no doubt, a major customer for British weapons. It is followed by a mix of Eastern European, Asian, and South American countries. Among them, China is found to have imported the third most weapons, unsurprising given its near-constant state of civil war throughout the Interwar years. Accordingly, the Chinese government was also one of the top importers of aircraft and armoured fighting vehicles, with further arms going to some of the warlords fighting against it and each other. Spanish imports of arms rose dramatically once the civil war there erupted while Romania and Poland sought to build up armies in defence against the USSR and Germany, respectively. Spain also shows up as top importer of warships, though these imports mainly occurred before the civil war, followed by five non-European countries with long coastlines. Interestingly, Italy and the UK also show up as top importers of warships, with the latter re-obtaining vessels it had earlier supplied to its dominions in Oceania while the US and France exhibit non-negligible imports of SALW, potentially hinting at the more commercial nature of the trade in small arms.

Finally, the IPITA data can not only be investigated in terms of importers or exporters, but also from a network perspective. Recent studies on the Cold War and post-Cold War arms trade show that these transactions now form a global network with implications for system-embedded interstate politics (Beardsley et al. 2020; Thurner et al. 2019). As shown in Figure 2 for the year 1930, network structures emerging from dyadic transfers can also be observed in the Interwar arms trade. Panel 2a presents the MCW transfer network, where a directed tie exists between countries i and j if i transferred at least one military aircraft, AFV, or ship to j. It is visible that top MCW exporters in terms of trade volume, such as the UK, France, and the US, also had the most outgoing ties while China is at the core of the network as a main importer. There are numerous importer countries with only a single supplier. Interestingly, a tendency towards triadic closure is really only visible between the UK and US, with both exporting to Spain, Japan, and China while trading with each other. Instead, arms exporters more often shared recipients without trading with each other. This is a notable difference to the Cold War system (Thurner et al. 2019).OPEN IN VIEWER

Panel 2b shows the SALW trade network, a directed tie exists if i transferred weapons worth at least 100 US$ ⁷ to j. As compared to 2a, it becomes clear that this network has more participating states, transfer ties, and an overall higher density. The SALW market thus had more, and more active, participants than the trade in MCW. It is also visible that much fewer countries had a single supplier of SALW than MCW. But beyond these general statements, the network in 2b is already too dense for further visual inspection, necessitating the use of network analysis methods. These methods also allow inspecting multiple of the yearly networks at once.

As an example, Figure 3 plots how the centralization of both the SALW and the MCW trade networks developed over the period between the two World Wars. Because arms transfers are directed ties, we can distinguish between In- and Out-Centralization (see Wasserman and Faust 1994), which capture to what extent the network revolves around a central core of, respectively, importers and exporters. Figure 3 shows that both arms transfer networks were more centralized in terms of exporters than importers, i.e. few exporters transferred weapons to many importers, that both In- and Out-Centralization were more pronounced for the SALW than the MCW market, and finally that especially Out-Centralization underwent substantial changes over time, decreasing for SALW but increasing for MCW in the 1930s.OPEN IN VIEWER

An Application: International Arms Transfers and the Demand for Military Expenditures, 1920-37

Finally, we present an application of the IPITA data. A major benefit of these new arms trade data is that they allow us to re-investigate established findings on the drivers and consequences of arms transfers and to check whether they also hold when examining not the bipolar Cold War or the unipolar post-Cold War orders, but instead the multipolar Interwar system. This is especially relevant for arguments which include a specifically temporal component. For instance, Pamp and Thurner (2017) find that weapons imports had a positive effect on defence spending after the Cold War but not during it. They attribute this to the arms trade becoming more commercialized after 1989 whereas free transfers in the form of military assistance decreased. The Interwar arms transfer system is usually regarded as highly commercial (Harkavy 1975, 98), arguably even moreso than the post-Cold War system, meaning that arms imports should be associated with higher military expenditures there. But at the same time, the multipolarity and economic situation of the 1920s and 1930s meant that many transfers at the time involved barter deals and large credits, factors which Pamp and Thurner (2017) identify as potentially reducing such a positive effect. But even if theoretical arguments have no explicit scope conditions regarding time and space, temporal generalizability is an important, albeit underappreciated, facet of quantitative conflict research (Fordham 2020). Along these lines, Pamp and Thurner (2017) propose that arms exports reduce the defence spending of democracies because their exports usually go to allies, thus increase their security, and accordingly allow them to reduce their own investments in security. But as noted in later research by the same authors, this substitution effect may depend on alliance commitments being sufficiently loose to allow free-riding and hence not apply to the Cold War years (Pamp et al., 2018a). How a country’s defence spending is influenced by its arms imports and exports thus appears to depend quite substantially on the examined period. We take this as a motivation to examine the relationship between arms transfers and military expenditures in the years 1920-37, replicating Pamp and Thurner’s (2017) empirical models with the IPITA data.

While the original models focus on MCW only, we include variables for both MCW and SALW because IPITA includes data on both kinds of weapons ⁸ . We measure MCW imports and exports as the total number of MCW received and sent, respectively, in a given year while SALW transfers indicate the total financial value of the transferred weapons. We log-transform these variables and, following Pamp and Thurner (2017), lag the export measures by 1 year. Otherwise, our model specifications mirror those reported in their tables two (imports) and four (exports), albeit drawing on alternative data sources where necessary due to the different period of observation. We present the results of a total of eight panel models, four each for arms imports and exports, in Figure 4. Following Pamp and Thurner (2017), these models employ country- and year-fixed effects, control for GDP, population, intra- and interstate conflict, the extent of democracy, and neighbours’ military spending and conflict, and use heteroskedasticity and autocorrelation consistent (HAC) standard errors. They differ in how they capture time dependence. Importantly, the export models include only countries which exported a non-zero amount of weapons at least once. We offer additional detail on the specification of these models in the supplementary material.OPEN IN VIEWER

The left panel of Figure 4 indicates that in the Interwar years, SALW imports were associated with increased military expenditures as they exhibit a positive and statistically significant coefficient across all specifications. In contrast, the coefficient estimate of MCW imports is consistently closer to and statistically indistinguishable from zero. This divergence may be due to the SALW trade being more commercial than MCW transfers (see Lebacher, Thurner, and Kauermann 2021), though many studies consider the Interwar arms trade as a whole to have been very economically oriented (see e.g. Harkavy 1975; Laurance 1992). The lack of an association between MCW imports and military expenditures, which exists in the post-Cold war period (Pamp and Thurner 2017), may be seen as a sign that the Interwar trade in MCW was not as commercial as the literature assumes, though barter deals and large credits were arguably also more common for MCW than SALW transfers. The right panel of Figure 4 offers some, albeit weak evidence in line with the expectation that MCW exports reduce defence spending as their coefficient is statistically significant at the 90%-level in the first two models. This effect becomes insignificant once we instrument for the lagged dependent variable and SALW exports exhibit no statistically significant effects in any specification. Sending MCW may thus have allowed exporters to free ride on the security provided by allies in the Interwar years, though this result is sensitive to modelling choices and based on the small sample of countries which actually exported weapons at the time. In sum, these results demonstrate the applicability of the IPITA data to reassess existing empirical findings and to obtain new insights on the arms trade, indicating that the findings of Pamp and Thurner (2017) travel quite well to the Interwar period while suggesting that the commercial orientation of the period’s arms trade may be overstated (see also Mehrl, Seussler, and Thurner 2022).

Conclusion

This article introduces IPITA, a new dataset on arms transfers in the Interwar years. IPITA provides dyadic information on weapons sent between independent countries, as well as some colonies and non-state actors, for the period 1920-39, covering military aircraft, ships, armoured fighting vehicles, as well as small arms and light weapons including artillery, ammunition, and explosives. Here, we have made the data collection process transparent, offered some descriptive insights on the arms transfer system of the Interwar years, and applied the data to show that in the period, imports of SALW, but not MCW, were associated with increased military expenditures whereas there is some evidence that exports of MCW, but not SALW, resulted in reduced defence spending.

As with all data, IPITA also has some limitations that require discussion. The small arms and light weapons data rely on customs data, meaning that they depend on countries’ reporting, are subject to reporters’ different rules on what goes (and goes not) through customs, and miss transfers that occurred clandestinely or through other non-customs channels. Because the source documents offer no standardization across reporting units, we had to standardize across weapon types and different currencies as well as estimate financial values for transfers where none were given, potentially inducing measurement error. The MCW data, covering military aircraft, ships, and armoured fighting vehicles, were collected from a wide range of primary and secondary sources. They may thus miss transactions not reported in any of these sources whereas for some transfers, sources gave delivery dates and numbers only as ranges of years or amounts. Again, this introduces some imprecision and measurement error. And finally, it is currently difficult to compare transfer volumes of different weapon types with each other as IPITA includes no catch-all volume indicator such as SIPRI’s (2023) TIV. Extending the TIV to Interwar military technology and/or small arms and light weapons thus appears as an important methodological next step for arms trade research.

Since the IPITA data provide a significant extension to the period for which data on MCW and SALW transfers is available, we next highlight some important considerations for researchers seeking to combine Interwar and post-WW2 data and use them together in their analyses. On a technical level, this combination is possible, albeit not straightforward, because, in the case of MCW, different datasets employ different units of measurement: as discussed above, we report unit numbers for MCW whereas the WMEAT data report financial values and SIPRI mainly relies on trend indicator values. That being said, SIPRI data also enumerate unit numbers, thus facilitating the combination of datasets, while our data not only mirror NISAT in the use of financial values for SALW, but even apply the NISAT weapons coding scheme. In addition, for applications that require binary data, such as most network analysis approaches, the different datasets can be binarized separately using thresholds the researchers judge to be comparable and then combined. From a technical perspective, combining these datasets hence requires some work but is possible by using unit numbers as measurement unit for MCW and financial values, standardized to the same currency and year, for SALW, or by binarizing the data.

However, and depending on the specific application, researchers should also consider whether it is appropriate to combine pre- and post-World War II data in the same analysis. On one hand, weapons technology, especially in the case of combat aircraft and armoured fighting vehicles, underwent rapid changes between the 1930s and 1950s with, for instance, the introduction of jet engines and air-to-air missiles in the former, that were reflected in changes in the cost, strategic value, and longevity of weapons. Comparing unit numbers across periods thus has clear limitations, though applying TIV valuations to the IPITA data would assuage this point. But on the other hand, the international political system within which arms transfers happen changed significantly after World War II. As such, some important covariates for inferential analyses of the post-World War II arms trade, such as United Nations voting behavior (see e.g. Martínez-Zarzoso and Johannsen 2019), are unavailable for the Interwar period while more generally, it may require substantial theoretical assumptions to model the effects of other covariates (including, potentially, arms transfers themselves) as if they were the same during these different historical periods. We thus recommend that researchers think carefully about whether it is theoretically appropriate to combine Interwar and post-World War II arms trade data within the same descriptive or inferential analysis. Alternatively, they may follow Mehrl, Seussler, and Thurner (2023) and present separate models testing the same hypotheses for these different periods which, while more complicated in terms of modelling and presentation, allows accounting for their differences in available data and expected theoretical dynamics.

Taken together, IPITA contributes to research in political science, economics, and history. Political scientists can, for instance, use IPITA to study how and whether arms transfers were used as foreign policy tool or contributed to war and peace within and between countries in the multipolar setting of the Interwar years. Economists can employ IPITA to research the demand for arms imports, when arms producers decide to export weapons regardless of their potential negative security externalities, and how these transfers affect importers’ and exporters’ spending on other goods. And for historians of the Interwar years, IPITA allows situating detailed archive-based research on, e.g., armaments procurement decisions, military technological cooperation or shifting foreign policies, within a broader, comprehensive picture of the period’s global arms trade and security interdependencies. IPITA will thus contribute to our understanding of an important historical period as well as of contemporary phenomena in political and economic international relations.

Supplemental Material

Supplemental Material

Supplemental Material