Necessity is the Mother of Invention: Temporary Shocks and Technical Change∗ PRELIMINARY DRAFT W. Walker Hanlon June 5, 2011

Abstract The relationship between short-term volatility and long-term growth is an important topic in economics. Central to this debate is whether temporary shocks can affect innovation rates. This study provides causal evidence on this relationship by focusing on a large exogenous shock affecting the 19th century British economy. The episode was caused by the U.S. Civil War, which sharply reduced raw cotton supplies, causing a depression in the important British cotton textile industry from 1861-1865. Using British patent data, we observe a sharp increase in cotton-textile related patents during this period, while no similar pattern is observed in other similar technology categories. Moreover, we observe that the type of innovation shifts towards technologies used in the early stage of the textile production process, in order to take advantage of lower-quality cotton from other suppliers. These results suggest that innovation is counter-cyclical, at least for a common class of economic shocks, and provide evidence of directed technical change of the type suggested by Acemoglu (2002). Additional patent data from India and the U.S. suggest that the shock increased the number of overseas patents filed by British inventors during 1861-1865, particularly those in cotton-related technologies and those from inventors located in the cotton textile manufacturing districts. ∗

Thanks to my advisers, Don Davis and Eric Verhoogen. Thanks also to Tom Nicholas for generously sharing some of the data used in this project, and for his helpful comments, and to participants at the Columbia International Trade Colloquium. This project was supported by funds from the National Science Foundation (grant No. 0962545) and the Economic History Association. Part of the research was done while the author was visiting the London School of Economics as an AMID fellow.

1

1

Introduction

The relationship between short term economic fluctuations and long-term growth is a hotly debated topic in economics, with considerable implications for economic policies.1 Central to this debate is whether, and how, temporary economic shocks affect investments, particularly investments in innovation. However, the relationship between temporary economic shocks and innovation is theoretically ambiguous. For example, innovation will be procyclical in theories in which innovation depends on learning-by-doing (Lucas (1988), Young (1991), Martin & Rogers (2000)) or on investments by credit-constrained firms (Aghion et al. (2005)). On the other hand, innovation will be counter-cyclical in theories in which innovation and production compete for inputs or when a reduction in output reduces in some other way the opportunity cost of changing to more advanced technologies (e.g., Hall (1991), Caballero & Hammour (1996), and Aghion & Saint-Paul (1998)). Still others suggest that temporary economic shocks may affect the nature of innovation, through directed technical change (Hicks (1932), Acemoglu (2002)). This paper provides empirical evidence that will allow us to differentiate between some of these theories in order to gain a better understanding of the relationship between temporary shocks and innovation rates. Existing empirical work has taken two basic approaches to studying the relationship between economic volatility, innovation, and growth. One approach uses cross-country data to investigate the relationship between economic volatility, either at the economy or industry level, and growth (Ramey & Ramey (1995), Imbs (2007), Ranciere et al. (2008), Koren & Tenreyro (2007)). A second approach uses data within a single economy, usually the U.S., to study the relationship between output and innovation, measured using R&D expenditures, at either the economy or industry level (Barlevy (2007), Ouyang (forthcoming)). The weakness of both of these approaches is that they cannot provide causal evidence on the relationship between economic fluctuations and growth due to the possibility of reverse causality, i.e., that changes in innovation rates may be causing changes in output levels. Given the prominent role that “technology shocks” are often given in theories explaining short-term economic fluctuations, this concern is of first-order importance. The goal of this project is to provide causal evidence on the relationship between temporary economic shocks and innovation rates. This is done by focusing on one particular incident, a large, exogenous, temporary, industry-specific shock in the 19th century. The shock was caused by the U.S. Civil War, which sharply reduced world raw cotton supplies. Our main analysis focuses on Britain, which was heavily impacted by the shortage through its cotton textile industry, which, at the time, was Britain’s largest manufacturing industry and most important export. The shortage of raw cotton caused a severe depression in out1

See, e.g., Lucas (2003), Epaulard & Pommeret (2003), Barlevy (2004), and Jones et al. (2005).

2

put in the industry, lasting from about 1861-1865, with production dropping by up to half from peak to trough. Importantly, the cotton textile industry rebounded rapidly following the end of hostilities, quickly resuming its original growth path. Focusing on this particular example allows us to clearly observe the causal effects of the temporary shock on innovation rates. Of course, the shock in question was caused by a sharp increase in input prices, so the results of this study are most relevant for events of this type. Similar negative shocks driven by increases in input prices are, however, quite common, with oil price shocks being one prominent example. Thus, while this study may not be relevant for all types of economic shocks, it does help us understand the effects of one prominent class of them, in more detail, and with more confidence, than was previously available. As in other recent studies using historical data from this period (e.g., Donaldson (2010)), the empirical setting used in this study was chosen because it has some unique features that help us more clearly identify the effects of interest. To begin with, the U.S. Civil War was clearly exogenous, and the impacts were largely unexpected. Moreover, the shock was temporary, both in retrospect, and in the view of many people at the time. Contemporary reports show that during the shock period there was a generally held expectation that the war would soon end and that the cotton textile industry would rapidly rebound. This expectation turned out to be correct, as cotton textile production quickly reverted to its original growth path following the end of hostilities. Another important feature is that, despite the massive size of the economic disruption, there was virtually no government policy response. This was largely due to the strong free-market ideology that prevailed in Britain at this time, particularly in the northern manufacturing districts that bore the brunt of the negative effects. The lack of government response is important in allowing us to identify the true effects of the shock. It would be surprising if an event of similar magnitude could occur today without a significant government response. In order to measure the effect of the depression on innovation, an extensive database of British patents was constructed, covering 118,863 patents for the years 1855-1883. Each of these patents was classified by the British Patent Office (BPO) into technology categories based on the type of invention they represented. These classifications have not been used previously. They were collected and digitized from around 1,500 pages of original BPO records for the purposes of this study. These technology classifications are crucial in allowing us to identify the particular types of technologies represented by each patent. This allows us to track how the shock affected innovation in particular types of technologies at a fine level of detail. Two main results emerge. First, the period of depression in the cotton textile industry,

3

1861-1865, is characterized by a sharp increase in the number of patents of technologies used by that industry. In particular, the data show an increase of around 10-15% in patents of spinning technologies, one of the two key technology categories used by the cotton textile industry. Moreover, we observe an increase of around 40-50% in patents of spinning technologies spicifically related to cotton. No similar effects are observed for technologies related to the other major textile inputs: wool, linen/flax, or silk. Thus, we observe that the downturn in the cotton textile industry was accompanied by a sharp increase in innovation in cotton-textile-related technologies. This finding suggests that innovation is counter-cyclical, in contrast to the findings of previous studies, such as Barlevy (2007) and Ouyang (forthcoming). Second, using more detailed data, we are able to observe a shift in the nature of innovation in cotton-textile-related technologies. In particular, the data show a significant shift in innovation towards technologies related to the preparation and cleaning of raw cotton. This shift appears to be a response to an increase in the abundance of low-quality Indian cotton, relative to cleaner and higher-quality American cotton, during the Civil War period. This finding suggests that directed technical change (induced innovation) took place as a result of the change in relative factor prices caused by the war. In particular, we find that technical change was directed towards taking advantage of the relatively more abundant out of two factors (low quality Indian cotton vs. high quality American cotton). This supports the predictions of Acemoglu (2002), which suggests that, when the elasticity of substitution between two factors is high, technical change will be directed towards taking advantage of the more abundant factor. These results are supplemented using Indian patent data, gathered from original sources, covering 1859-1879. Indian patents were fewer in number, with only 1,114 filed during this period, but they contain detailed information on the identity and location of the inventory as well as the nature of the invention. The experience of India is particularly interesting because, as a raw cotton supplier competing with U.S. producers, it experienced a large demand shock during the U.S. Civil War. The Indian patent data suggest that there was an increase in Indian patents of cotton-related technologies during the Civil War period. Almost all of these patents were for technologies used to clean or ship raw cotton. This increase was due to both an increase of textile-related patent by inventors based in India, as well as an increase in Indian patent filings by British inventors. Finally, the share of Indian patents by British inventors based in the cotton-textile manufacturing districts increased during the Civil War period, relative to other British inventors, suggesting that the shock also influenced the geographic location of innovative activity. We also use U.S. patent data in order to investigate change in the pattern of overseas patent filings by British inventors. Some 1,842 U.S. patents were filed by British inventors 4

from 1850-1873, accounting from 1.33% of total U.S. patents.2 These data suggest that number of U.S. patents filed by British inventors increased during the Civil War period, relative to those filed by U.S. or other European inventors. The number of U.S. patents filed by British inventors based in the cotton-textile districts, relative to British inventors from other regions, also increased during this period. There is also a sharp increase in the share cotton-textile related technologies in total U.S. patents filed by British inventors. Together, the Indian and U.S. data suggest that the increasing the number of cottontextile-related patents filed in Britain was accompanied by an increase in cotton textilerelated patents filed by British inventors abroad, as well as a shift in innovation towards the main cotton-textile producing regions within Britain. Because of the additional expense required to file patents overseas, this suggests that at least some of the additional technologies generated in Britain during the Civil War period were sufficiently valuable to merit patenting abroad. Moreover, the increase in patents by inventors based in India suggests that positive demand shocks may also act to spur additional innovative activity. While this paper studies the effect of a temporary shock on innovation, it is related to literature studying how innovation rates are impacted by more permanent changes, particularly changes related to trade. A recent example is Bloom et al. (2009), which uses data on patents, IT, and R&D expenditure to show that the recent increase in Chinese imports resulted in an increase in innovation and technological upgrading by those European firms that compete with them.3 The main implication of the current paper, vis a vis this literature, is that trade may affect innovation both through permanent changes (e.g., liberalization) and through temporary shocks, and that both of these avenues should be considered when evaluating the relationship between trade and technological change. To date, almost all of the focus has been on the effect of long-term changes, despite results suggesting that trade openness can also affect the vulnerability of economies to temporary trade shocks (di Giovanni & Levchenko (2009)). Because we find evidence of directed technical change, this paper is also related to literature on that topic (Newell et al. (1999), Popp (2002), Linn (2008), Allen (2009), etc.). The main difference between this and previous work is that by using a single large exogenous shock, we are able to identify evidence of directed technical change more clearly than some existing studies. The next section of this paper provides background on the empirical setting and the shock, followed by some qualitative evidence on the impact of the shock in Section 3. Section 4 introduces the data used. The analysis begins in Section 5, which investigates 2

British inventors were the largest group of non-American patentees during this period. Their paper also surveys a larger literature studying the relationship between trade and technological change. 3

5

the impact of the shock on overall innovation rates in cotton-related textile technologies. Section 6 then looks more closely at the changes in the nature of innovation within these technologies. Supporting evidence from Indian and U.S. patent data is presented in Section 7, while Section 8 concludes.

2

Empirical setting

The cotton textile industry played a key role in the British economy starting in the 18th century and extending through the early 20th century. During the time period considered in this study, 1855-1880, cotton textiles were generally Britain’s largest export and raw cotton was Britain’s largest import. For example, in 1860, cotton textile exports were valued at £52 million, dwarfing the next largest export categories (wool textile exports, £15.7 million and iron and steel, at £13.6 million).4

2.1

The Cotton Textile Production Process

This study focuses on innovation in technologies used in the textile production process, so it is helpful to have a rough understanding of this process, and the technologies involved, before proceeding. There are four main steps in the cotton textile production process once the raw cotton has been harvested from the cotton plants. The first step of the process involves separating the cotton fibers from the seeds, using gins, cleaning the cotton, and preparing it for shipping using balers and packers. These processes could take place either in the raw cotton exporting country or in the textile manufacturing centers. In the second stage in the production process, the raw cotton was spun into yarn. This stage generally occurred in manufacturing centers, such as those in Britain or the Northern U.S., and was generally done using large powered spinning machines such as mules. This yarn thus produced could then be further processed into fabric, through weaving, the third stage of the cotton textile production process, or exported directly as yarn. At the time of this study, most of this weaving was done using large power looms. Once woven, fabric could be exported directly, or it could undergo the final stage in the production process, finishing, which included bleaching, dying, or printing. All of these production stages relied heavily on technology. As a result, Britain became the home of a large, important, and innovative textile machinery sector. Table 1 shows that, during the 1855-1883 period, the two main textile production technology categories, Spinning and Weaving, were in the top 5 patent technology categories, out of a total of 146, 4

Data from Mitchell & Deane (1962).

6

Table 1: Top ten BPO technology categories by patent count, 1855-1883

Rank 1 2 3 4 5

Technology Category Metals, Cutting, etc Furnaces Spinning Steam engines Weaving

No. Patents 7,017 6,157 6,009 4,809 4,807

Rank 6 7 8 9 10

Technology Category Railway etc. vehicles Steam generators Furniture Mechanisms Ships, Div. I (fittings, etc.)

No. Patents 4,184 4,065 3,216 3,120 3,051

in terms of the number of patents granted. They made up around 6% and 5%, respectively, of all British patents. Within Britain, the cotton textile industry was heavily concentrated in the Northwest region, particularly in Lancashire county. Manchester, the largest city in the region was the central hub of the industry, while Liverpool, the largest port, acted as the world market for raw cotton. While cotton was the largest textile industry in Britain, textile industries based on wool, linen/flax, and silk were also of significant size in Britain during this period. The technology and other inputs used by these industries was generally similar to that used by the cotton textile industry, with the greatest differences being in the early stages of production. For example, modified versions of spinning and weaving machinery that was originally developed for the cotton textile industry were also used to spin other fabrics. Furthermore, some mixed fabrics were made by weaving together yarns based on different inputs.

2.2

The impact of the U.S. Civil War

The British cotton textile industry was entirely dependent on imported raw cotton, as growing cotton in Britain was infeasible. At the beginning of the study period, the cotton textile industry was heavily dependent on cotton growers in the U.S. south, as is evident in Figure 1. The onset of the U.S. Civil War in 1861, which included a naval blockade of southern ports initiated in April 1861, caused a massive disruption in raw cotton imports to Britain. While other suppliers, such as India, Egypt, and Brazil, attempted to increase output, they were not able to increase their production rapidly enough to replace the flows from the U.S. The right-hand panel of Figure 1 shows that the share of British cotton imports coming from the U.S. dropped sharply during the war, while other suppliers, India chief among them, expanded. As Figure 2 shows, the reduction in raw cotton supplies led to a dramatic spike in input prices (left panel) and a sharp drop in production (right panel), 7

which is measured using British domestic consumption of raw cotton.5 Figure 1: British raw cotton imports and import shares by supplier British cotton imports, 1815-1910

Share of imports by supplier, 1850-1880

Data from Mitchell & Deane (1962).

Data from Ellison (1886).

Figure 2: British raw cotton prices and domestic consumption 1815-1910 Raw cotton price on the Liverpool market

Raw cotton consumption

Data from Mitchell & Deane (1962).

Two important facts about this shock are visible from these figures. First, the shock was unexpected. This is seen most clearly in the raw cotton prices, which show no run-up 5

The reduction in production also led to massive unemployment in the cotton textile districts, resulting in the “Lancashire Cotton Famine”.

8

prior to the onset of the war.6 Second, the episode was temporary in nature. This is visible in the rapid rebound of British domestic cotton production to its original trend. Not only can we observe the temporary nature of the shock looking back, but there is also evidence that the raw cotton shortage was widely considered temporary during the Civil War. For example, British cotton textile producers continued to invest in new machinery and mills, even while their existing mills stood idle. The number of installed spindles for spinning cotton yarn in England and Wales increased by 2.13 million from 1861-1868, or about 7.5% of installed capacity.7 Another feature of this shock is that it was largely transmitted through the cotton textile industry, rather than being a broad-based economic shock. Once raw cotton imports are removed, total British imports do not appear to be affected during the shock period.8 Similarly, once textile exports are excluded, British manufacturing exports also fail to show any large effect from the shock. Other main textile industries, based on wool, linen/flax, or silk inputs, showed no negative effects of the shock.9 If anything, these sectors benefited from the reduced competition from cotton textiles.

3

Qualitative evidence

Before undertaking our statistical analysis, it is worth sampling some of the qualitative evidence which is available on the impact of the U.S. Civil War on British innovation. A good place to start is with historians of the period. D.A. Farnie, in his authoritative history of the English cotton industry, writes the following (Farnie (1979) (p. 152-153)).

6

A good example of this attitude is given by Ollerenshaw (1870), who remarked in his presentation to the Manchester Statistical Society, that, “The American War commenced on April 5th, 1861, but for many months it had little effect on commerce - being generally regarded as merely temporary...” 7 Data from Forwood (1870). 8 See Figure 18 in Appendix A. 9 Graphs showing exports in these other sectors are available in Appendix A.

9

The shortage of American cotton compelled employers to re-equip their mills in order to spin Surat [Indian cotton], and especially to improve their preparatory processes...The process of opening the tightly packed raw material become wholly automated through the use of the Crighton Opener, invented in 1861, as did the subsequent process of scutching through the application of the ingenious piano-feed regulator developed in 1862...The reorganization of the preparatory processes entailed such an extensive investment of capital that it amounted almost to the creation of a new industry...Those innovations gave a great stimulus to the textile engineering industry and consolidated the technical supremacy of the Lancashire cotton industry in the world. Three points are worth noting here. First, Farnie argues that textile producers were “compelled to re-equip” their mills with new and different machines. Second, he focuses on machinery, such as openers and scutchers, which operate in an early stage of the textile production process, which suggests that technical change may have been directed towards these technologies in order to aid the use of Indian cotton. Third, Farnie suggests that the additional innovation had a long-term impact on the competitive position of British cotton textile producers. Another source of qualitative information on the period is available from contemporary sources. One such source is Ellison (1886), who wrote the following in his book, The Cotton Trade in Great Britain. The high prices caused by the cotton famine, however, gave an impetus to the culture [of cotton] in India which it would not otherwise have obtained, and thereby secured to Europe a permanent increase in supply. Moreover, the quality of the cotton has been so materially improved by the introduction of better methods of handling the crop, that “Surats” are no longer despised as they were up to within a few years ago. Ellison is pointing out the fact that the shortage of high quality American cotton caused British producers to look elsewhere, particularly to India, for their supplies. Moreover, he suggests that this led to an increase in the quality of Indian cotton. Presumably machines played some role in this improvement. Finally, additional information can be gleaned from the patents themselves. Though most of the patents provide only a simple description of the mechanisms involved, a few also mention the motivation behind the new technology. One example is given in Figure 3, which describes a patent filed in Britain in 1862. This patent was classified in the spinning technology category and the “Openers & Scutchers, etc.” subcategory, and was also identified based on the patent title search as a cotton-related patent. It describes an invention created specifically in order to help producers deal with the unique characteristics of the Indian cotton in the cotton preparation process. 10

Figure 3: An Example: Patent No. 2162 from 1862

From British Patent Abstracts, Class 120, 1855-1866. Available from the British Library.

While disparate pieces of evidence such as these can be informative, a more rigorous analysis is needed in order to determine the generality of the patterns they suggest. The next section describes the data that we will use for this analysis.

4

Data

The primary data used to measure innovation in this study come from British patent records. While imperfect, patent data is one of the best available quantifiable measures of technological advance. Modern patent data has been widely used in recent studies of innovation, building on seminal work by Schmookler (1966), Scherer (1982), Griliches (1984), and Jaffe et al. (1993). Hall et al. (2001) provide a helpful review of the advantages of using patent data, including that (1) patents contain highly detailed information, (2) there are a large number of patents available to study, and (3) patents are provided on a voluntary basis under a clearly defined set of incentives. This study is able to take advantage of thousands of patents and will draw heavily on the detailed information available in the patent descriptions. Importantly, while British patent laws changed in 1852 and 1883, they were stable during the period of this study. Of course, one disadvantage of using patent data is that it will not capture all types of innovation. Thus, results based on patent data may not con11

stitute a good test for certain classes of models, such as those based on learning-by-doing, which may be driven by types of innovation which are not normally patented. Most of the data used in this study was collected for the purpose of this project from around 1,500 pages of printed British patent records. To begin, we constructed a database covering all of the patents granted in Britain between 1855 and 1883, 118,863 in all.10 This database includes the patent application year and patent number, which uniquely identify each patent, as well as the name of the inventor. Conveniently, the dates given in the data represent the date of the patent application, rather than the date at which the patent was ultimately granted. In the British system at this time, patent applications could be made using only basic information on the invention. The application provided the applicant with provisional protection and could aid them in establishing the seniority of their invention. The applicant was then responsible for supplying full patent specifications with a certain time period, before the full patent could be granted.11 Crucially, each patent in the data is classified into one or more of 146 technology categories by the British Patent Office (BPO). The purpose of this categorization was to aid inventors in identifying previously patented inventions in order to determine whether an invention was in fact new. In order to get a sense of these technology categories, Table 1 presents the top ten categories by number of patents. Our focus will primarily be on the BPO spinning and weaving categories. The spinning category includes technologies related to the preparation of raw cotton, such as cotton gins and carding machines, machines used in the spinning process, such as mules, yarn types, and other related technologies. The weaving category includes technologies such as looms, types of fabrics, and fabric treatments. These data were supplemented with information from the A Cradle of Invention database, which has been used in previous research (e.g., Brunt et al. (2008)).12 These data provide additional information on the title of the patents, which will be used to generate more detailed classifications of the technology represented by each patent. This data set also provides information on the month of the patent application, allowing analysis at the quarterly level.13 Consistent patent titles are available from 1853-1870, after which there was a clear structural change in the naming conventions used. Using the patent titles allows us to run keyword searches in order to identify patents 10

These data include both granted patents, and those which received provisional production but where a patent was not ultimately granted. 11 For more information on the British patenting system during this period see Van Dulken (1999). 12 These data are available through MFIS LTD (finpubs.demon.co.uk). These data match the primary database well, with over 98% of patents in the two databases matching. 13 Analysis at the monthly level is undesirable given the high level of month-to-month volatility in the number of patent applications filed.

12

representing particular types of technologies, such as those related to cotton, wool, etc.14 In particular, we undertake keyword searches to identify patents related to the main textile inputs: cotton, wool, linen/flax, and silk. Some summary statistics from these patents are provided in Table 2. We can see that the majority of those patents listing one of these keywords are also classified into the BPO spinning technology category, while a few are listed in the weaving category, and some other fall into categories other than spinning and weaving. Keyword searches are also used to identify those patents that state that they are related to spinning, or weaving. Most patents with “spinning” in the title are also listed in the BPO spinning technology category, while most of those mentioning “weaving” are classified in the BPO weaving category. This suggests that the keyword search approach is reliable, though more restrictive than the BPO categories. The final category, labeled “Spinning, Weaving, etc.” captures titles including a number of textile-technology related terms.15 This entry suggests that around half of those patents classified into the spinning or weaving category by the BPO can be identified as such using title keyword searches.

Table 2: Summary statistics from patent title keyword searches, 1855-1870 Title search topic Cotton Wool Linen Silk Spinning Weaving Spinning, Weaving, etc.

Total hits 1,230 998 518 392 976 1,245 3,302

Number in BPO Spinning 892 651 397 279 935 42 1,569

Share in BPO Spinning 73% 65% 77% 71% 96% 3% 48%

Share of BPO Spinning 29% 21% 13% 9% 30% 1% 50%

Number in BPO Weaving 61 57 21 36 25 1,200 1,343

Share in BPO Weaving 5% 6% 4% 9% 3% 96% 41%

Share of BPO Weaving 2% 2% 1% 1% 1% 46% 52%

Within each BPO technology category, patents may also be listed in various technology subcategories. For example, within the BPO spinning technology category, we are able to identify patents falling into subcategories such as “Gins”, “Mules and Twiners”, “Carding Machines”, etc. For this study, data were collected for a number of the more common spinning subcategories. These data will allows us to track changes related to the shock at a relatively detailed technology level. These data are discussed in more detail in Section 6.

14

This technique has been used for these data previously by Brunt et al. (2008). These terms, Spinning, Weaving, Loom, Spindle, Bobbin, Slubbing, Roving, Combing, Winding, Shuttle, and Thread, were identified through a manual review of a sample of patents. 15

13

5

Temporary shocks and innovation rates

We begin our analysis of the effect of the shock by considering the effects on broad BPO technology categories, spinning and weaving. We then consider the effects on technologies related specifically to the primary textile inputs: cotton, wool, linen/flax, and silk. Our first glimpse of the data is presented in Figure 4. The top panels graph patent counts for the BPO spinning and weaving technology categories. The left-hand panel contains annual data from 1855-1883 while the right-hand panel shows quarterly data, smoothed using a four-quarter moving average, from 1855-1873.16 The bottom two panels show similar data for all BPO technology categories except spinning and weaving. Figure 4: Patent counts from BPO technology categories Annual Spinning & Weaving Patents

Quarterly Spinning & Weaving Patents

Annual - All Other Technologies

Quarterly - All Other Technologies

16

The month of the patent application is missing for some years after 1873, so consistent quarterly data is available only for 1855-1873.

14

The most striking feature of these graphs is that the onset of the U.S. Civil War corresponded to a sharp increase in patents of spinning-related technologies, and that this unusually high level of spinning patents was sustained throughout most of the 1861-1865 period. This high level of spinning-related technology patents only falls toward the end of the Civil War, which also corresponds with a financial crises that struck England and was particularly severe in the northern cotton manufacturing districts. No similar increase appears in weaving technologies, nor do other technology categories show similar effects. We are able to look at the effected technologies in more detail by running keyword searches on the patent titles. As described, this allows us to look specifically at patents related to the four main textile inputs: cotton, wool, linen/flax, and silk. One limitation on this analysis is that the usable data extend only until 1870, after which there is a clear structural break in the patent naming conventions, with the input material being mentioned much less often after that year. Figure 5 shows the count of patents mentioning each of the four key textile inputs. Each graph also shows the number of patents mentioning each of these inputs which are also listed in the BPO spinning category. The message from these graphs is clear; there was a sharp and sustained increase in cotton-textile related patents during the 1861-1865 period. No similar increase appears for other main textile inputs. Next, the patterns identified above are tested more rigorously using some simple regressions. In particular, we run OLS regressions based on Equation 1, where Pt represents the log count of patents in period t, St represents a dummy for the shock period (1861-1865), Tt represents a time-trend, and Xt gives the total number of patents in period t. When considering quarterly data, we also include an indicator for the quarter, Qt , to capture seasonal effects.

Pit = β0 + β1 Sit + β2 Tt + β3 Xt + β4 Qt + it

(1)

Regressions are run separately for each category of patents, e.g., BPO spinning patents. Results are presented in Table 3 below. These results suggest that the shock caused a significant increase in the number of patents listed in the BPO spinning category (column (1)), patents mentioning cotton in the title (column (3)), as well as BPO spinning patents that also mentioned cotton in the title (column (4)). These effects are large; the shock is associated with an increase of spinning technology patents of around 15%, and an increase in cotton-related patents of 40-50%. No similar increase is observed for BPO weaving patents (column (2)), or patents related mentioning wool, linen/flax, or silk in the title (columns (5), (6), and (7), respectively). Moreover, these results are not driven by the increase in 1861 alone. This is apparent in the results presented in Table 4, in which the same regressions are run with the shock period flag indicating only the 1862-1865 period. These suggest 15

that BPO spinning category patents increased by over 10% in the 1862-1865 period, and cotton-related patents increased by 34-39%. All of these results are highly significant, and none of them are sensative to the exclusion of any of the control variables. Figure 5: Count of patents with titles mentioning main textile inputs, 1853-1870 Cotton-related patents

Wool-related patents

Linen/flax-related patents

Silk-related patents

Quarterly data smoothed using four-quarter moving averages.

Before leaving this section, it is worth exploring in more detail pattern of cotton-related patents over time. This is done in Figure 6, which takes advantage of the availability of monthly patent data from 1855-1870. This figure reveals that there was an enormous spike in cotton-related patents in April 1861, the month in which the U.S. Civil War broke out. Giving the timing of these patent applications, they could not reflect new innovation undertaken in response to the change in conditions. Rather, it seems clear that this spike was due to an increase in patent applications of already-existing technologies which were previously not worth patenting, but which became worthwhile to patent given the change 16

Table 3: Effect of the shock (1861-1865) on patent count – by patent type

Shock flag (1861-1865) Log Total Pats. Time trend Quarter flag Constant Observations R-squared

BPO Spinning patents (1) 0.147*** (0.0466) 0.958*** (0.269) Yes Yes Yes 72 0.425

BPO Weaving patents (2) -0.0811* (0.0464) 0.550** (0.267) Yes Yes Yes 72 0.203

DV: Log Cottonrelated patents (3) 0.424*** (0.0622) 0.865** (0.399) Yes Yes Yes 60 0.524

Standard errors in parentheses

of patent CottonSpinning patents (4) 0.511*** (0.0823) 1.035* (0.528) Yes Yes Yes 60 0.473

count Woolrelated patents (5) -0.0685 (0.0852) 0.217 (0.547) Yes Yes Yes 60 0.196

Linenrelated patents (6) -0.0549 (0.112) 1.011 (0.716) Yes Yes Yes 60 0.196

Silkrelated patents (7) -0.0548 (0.142) 0.987 (0.911) Yes Yes Yes 60 0.188

*** p