Room 23

 

DO PATENT POOLS ENCOURAGE INNOVATION?

EVIDENCE FROM THE 19TH CENTURY SEWING MACHINE INDUSTRY

by

Ryan L. Lampe

Petra Moser  

 

NATIONAL BUREAU OF ECONOMIC RESEARCH

1050 Massachusetts Avenue

Cambridge, MA 02138

June 2009  

 

We wish to thank the Wisconsin State Historical Society for granting access to the Singer Archives and Eric Hilt, Jim Bessen, Paul David, Ross Thomson and seminar participants at the Santa Fe Institute, the Stanford Junior Faculty Lunch, and Stanford’s Social Science History Workshop for helpful comments. Luke Brennan and Marina Kutyavina provided valuable research assistance. The views expressed herein are those of the author(s) and do not necessarily reflect the views of the National Bureau of Economic Research.

© 2009 by Ryan L. Lampe and Petra Moser. All rights reserved. Short sections of text, not to exceed two paragraphs, may be quoted without explicit permission provided that full credit, including © notice, is given to the source.  

ABSTRACT

Members of a patent pool agree to use a set of patents as if they were jointly owned by all members and license them as a package to other firms. Regulators favor pools as a means to encourage innovation:

Pools are expected to reduce litigation risks for their members and lower license fees and transactions costs for other firms. This paper uses the example of the first patent pool in U.S. history, the Sewing Machine Combination (1856-1877) to perform the first empirical test of the effects of a patent pool on innovation. Contrary to theoretical predictions, the sewing machine pool appears to have discouraged patenting and innovation, in particular for the members of the pool. Data on stitches per minute, as an objectively quantifiable measure of innovation, confirm these findings. Innovation for both members and outside firms slowed as soon as the pool had been established and resumed only after it had dissolved.

Ryan L. Lampe

Department of Economics Stanford University 579 Serra Mall Stanford, CA 94305 rlampe@stanford.edu

Petra Moser

Department of Economics Stanford University 579 Serra Mall Stanford, CA 94305-6072 and NBER pmoser@stanford.edu  

 

On the eve of the United States’ entry into World War I, the aeronautics industry was in disarray. While the Wright Brothers owned a patent on the plane, Glenn Curtiss held overlapping patents that were crucial for improving the design, engines and controls of even the most basic planes.1

Since the Wright Brothers’ patent covered any aircraft using these improvements, they were able to block Curtiss from producing planes.2

Faced with the need to purchase planes from Europe to participate in a European war, Congress created a committee to address the problem. This Committee, headed by Franklin D. Roosevelt, recommended that Curtiss and the Wright Brothers form a patent pool: an arrangement in which “two or more parties agree to pool their respective technologies and license them as a package”.3

The Manufacturers’ Aircraft Association, formed on 12 July, 1917, combined all patents that were needed to build a plane and made them available for licensing. Three weeks later, the United States entered World War I; U.S. production soared from 83 aircraft in 1916 to 1.807 in 1917 and 11.950 in 1918.

Almost one hundred years later, patent pools have re-emerged as a remedy for industries that are plagued by litigation and patent blocking, which occurs when owners of competing patents prevent the commercialization of new technologies.5

Since July 2008, UNITAID, an international drug purchasing facility, has been developing a patent pool to facilitate the development of improved treatments for HIV and AIDS.6

1 Dykman, “Patenting Licensing” and Heller, Gridlock Economy.
2 Bittlingmayer, “Property Rights”. In the example of the airplane pool, patents were “one-way blocking”. Wright's patent could block the use of Curtiss' improvements, but Curtiss could not prevent the Wright Brothers from producing planes, as long as they did not include his improvements. See Gilbert, “Antitrust”, for a detailed discussion of blocking patents.
3 Gaule, “Towards Patent Pools”; and Merges, “Contracting into Liability Rules”.
4 Stubbs, “Race”, p. 133.
5 See, e.g. Shapiro, “Navigating the Patent Thicket”.
See http://www.stopaidssocieties.org.uk/2008/07/patent-pool-plans-progress/.

2

 

In February 2009, the pharmaceutical company GlaxoSmithKline (GSK) announced its support for a pool to develop treatments and cures for “neglected” diseases, including malaria and cholera. In addition, four parties holding key patent applications related to the development of SARS vaccines expressed their willingness to form a pool.7

Pools have also been suggested as a means to advance diagnostic testing for breast cancer.8

These plans follow the formation of four successful pools in the IT sector during the 1990s: the MPEG-2 pool, the 3G platform and two DVD pools.9

In 2001, the value of U.S. goods produced under pooling agreements exceeded $100 billion.10

Antitrust authorities in the United States favor pools because they “provide procompetitive benefits by integrating complementary technologies, reducing transaction costs, clearing blocking positions and avoiding costly infringement litigation”.11

Theoretical models of patent pools predict that pools encourage innovation. Specifically, the prospect of a patent pool increases firms’ incentives to invest in R&D because lower risks of litigation and improved licensing schemes increase expected profits for participating firms.12


7 Simon et al., “Managing Severe Acute Respiratory Syndrome (SARS) Intellectual Property Rights.”
8 Verbeure et al., “Patent Pools.”
9 See e.g., Merges, “Institutions.” These pools combine complementary patents that form a technological standard.
10 Clarkson, “Objective Identification.”
11 U.S. Federal Trade Commission and Department of Justice, “Antitrust Guidelines.” Regulators have even credited pools with making it possible to produce inventions that would otherwise be blocked by overlapping patent grants: “The pooling of the patents, licensing all patents in the pool collectively and sharing royalties is not necessarily an antitrust violation. In a case involving blocking patents, such an arrangement is the only reasonable method for making the invention available to the public” (International Mfg. Co. v. Landon, 336 F.2d 723, 729, 9th Cir. 1964).
12 Lerner and Tirole, “Efficient Patent Pools”; Choi, “Patent Pools”; Gilbert, “Antitrust”; Shapiro, “Navigating the Patent Thicket” and Merges, “Institutions”. Industry experts echo this expectation. For example, one of the founding members of the 2005 radio frequency pool, Stan Drobac,writes that “If you're a licensee, you need to negotiate with five entities to get five licenses, instead of negotiating with 20 to get 20 licenses” (RFID Journal, 10 April, 2008). Pools that combine complementary patents lower licensing fees for outside firms because they avoid “royalty stacking”, when the same product is covered by multiple patents (e.g., Lemley and Shapiro, “Patent Holdup”; Shapiro, “Navigating the Patent Thicket” and Lerner and Tirole, “Efficient Patent Pools”).

3

 

The prospect of a patent pool may also increase the speed of innovation: If the number of patents needed to form the pool is limited, firms may race to develop the technologies that are included in the pool.13

A lack of contemporary data, however, has made it difficult to establish these effects empirically. Existing empirical work has sharpened our understanding of institutional characteristics, such as the types of licensing deals that pools offer to nonmembers and rent-sharing agreements among pool members.14

Nevertheless, there has been little empirical evidence on the effects of patent pools on innovation, mostly because such analyses require much longer time series of data than are available for contemporary pools. Moreover, pools that resolve conflicts about overlapping patents have been extremely rare in recent history.15

This paper uses the example of the first patent pool in U.S. history, the Sewing Machine Combination (1856-1877), to examine whether patent pools encourage
innovation.
16

One key advantage of the historical setting is that we can observe an industry from its birth to technical maturity, including more than 30 years of data to analyze the long-term effects of a pool.17

Another advantage is that the sewing machine pool operated in the complete absence of regulation, which allows us to examine how pools behave when regulators give them free reign.18

13 Dequiedt and Versaevel, “Patent Pools”.
14 Layne-Farrar and Lerner, “Patent Pool Participation” and Strojwas, Lerner and Tirole, “Design”.
15 Bessen, “Imperfect Property Rights.” The only recent example of a pool intended to resolve overlapping patents is the laser surgery pool, which was accused of price-fixing in 1998 (Clark et al., “Patent Pools”). This pool was formed between Summit and VISX, the only two firms with FDA approval to license photorefractive keratectomy (PRK) laser equipment for vision disorders. Based on an FTC antitrust
complaint, the two firms eventually made their patents available for non-exclusive licensing to third parties. Bessen, “Imperfect Property Rights,” argues that pools of overlapping patent rights are less likely to form because prospective members do not know what share of expected profits will accrue to them.
16 See Thomson, “Path,” for a detailed history of the sewing machine industry and, in particular, the industry’s importance to the mechanization of shoe manufacturing.
17 See e.g., Clark et al., “Patent Pools” and Heller and Eisenberg, “Can Patents Deter Innovation”?
18 The first major antitrust case regarding patent pools was E. Bement & Sons v. National Harrow Co. 186, U.S. 70, 91 (1902). Six different firms had formed the National Harrow to end crippling litigation. Their pool combined 85 patents for float spring tooth harrows – a tool to soften top soil for cultivation. National Harrow grew to include 22 firms that covered over 90 percent of the market. Bement was a licensee who sued the pools over the terms of the license, which required licensees to adhere to a uniform price schedule. In 1902, the Supreme Court decided that the pool’s licensing terms were lawful, arguing that the benefits it conferred - preventing litigation over patent scope and validity - outweighed the costs that price-fixing created for the industry (Gilbert, “Antitrust,” pp. 2-3).

4

 

In the first step of the analysis, we use data on litigation and on patenting to test the predictions of existing theoretical models of patent pools. Litigation data confirm that pools lowered litigation risks for pool members.19

They also substantiate the prediction that the prospect of a pool encourages patenting among prospective members.20

Interestingly, however, the data suggest that pool members patented less while the pool was active and only began to patent more aggressively again after the pool dissolved in 1877. Non-members, which account for the vast majority of patents, appear to have been similarly affected by the creation of the pool; patenting increases immediately after 1856, but declines soon afterwards and remains at low levels until the pool dissolves in 1877. These results are robust to the inclusion of time- and firm-fixed effects, as well as alternative controls for demand and the Civil War. One potential problem with using patent data to measure innovations is that firms may have used patents strategically to negotiate a more advantageous position in the pool, or to protect themselves from litigation by the pool.21

19 See e.g., Choi, “Patent Pools”; and Shapiro, “Navigating the Patent Thicket”.
20 Dequiedt and Versaevel, “Patent Pools”.
21 See Cohen et al., “Protecting Their Intellectual Assets”; Moser “Why Don’t Inventors Patent” and Hall and Ziedonis, “Patent Paradox”.

5

 

To address this issue, we construct an alternative measure of innovation that objectively measures improvements in the performance of sewing machines. Specifically, we use information in company records, including trade cards, advertising and internal records to measure the number of stitches per minute that a sewing machine was able to perform.22

Data on sewing speeds confirm the results of the patent data: Innovation slowed soon after the pool had been established and did not recover until the pool had been dissolved. Thus, data on annual patenting and innovation suggest that theoretical models should be extended to capture the ex post effects of patent pools on their members and incorporate effects of pool on other, non-members, which appear to play an important role as drivers of innovation for the entire industry. The data suggest two ways in which the sewing machine pool may have discouraged innovation. First, by creating a more formidable opponent in court, the pool may have intensified the threat of litigation for outside firms, which lowered expected profits and discouraged innovation. Consistent with this idea, litigation data show that non-members were at a greater risk of being sued while the pool was active; pool members acted as plaintiffs in most of these cases. Second, the existence of a pool that aggressively defended patents may have shifted innovation by non-members towards substitute technologies that were not covered by the pool. In the case of the sewing machine industry, outside firms shifted both patenting and innovation (as measured by firm entry with new models of sewing machines) towards an inferior stitching mechanism, that was not covered by the pool.23

22 Measuring effects on innovation, in addition to effects on patenting, is particularly important if innovations differ systematically from patents. Data on innovations at 19th century world’s fairs indicate that inventors’ propensity to patent varies significantly across industries and over time (Moser, “Why Don’t Inventors Patent”?).
23 For a detailed analysis of the potential effects of patent pools on the direction of technical change see Lampe and Moser, “Patent Pools and the Direction of Technical Change”.

6

 

The rest of this paper is organized as follows. Section II presents a short history of the sewing machine industry and its patent pool. Section III describes data on litigation, patents and sewing speed as an alternative measure of innovation. Section IV and V test predictions of existing theoretical models and Section VI concludes.

THE EARLY SEWING MACHINE INDUSTRY, 1846-1885

On 10 September 1846, Elias Howe was granted U.S. patent No. 4.750 for
“Improvements in the Sewing Machine” – the lock-stitch. Although the patent was initially contested, Howe began to charge a hefty license fee ($25, almost 50 percent of the average price) as soon as the court had upheld his patent in 1853.
24

The Sewing Machine “Wars” (1846-1856)

Despite its high price, Howe’s license did not cover all parts that were needed to
build a functional sewing machine. Mirroring the experience of the early 20
th century aircraft industry, manufacturers soon filed suits to assert their rights to different parts of the sewing machine. Litigation during the “sewing machine wars” threatened to stop production and sales.25

For example, the I. M. Singer Company warned consumers not to buy sewing machines from Wheeler & Wilson, or Grover & Baker: 

The unparalleled success of our Sewing Machines has induced several fraudulent imitations of them, besides numerous infringements of our patents-of which we own sixteen. Suits for the infringement of our patents have recently been decided in the U.S. Circuit Courts in N.Y. and New Jersey. In the suit the great principle of HOLDING DOWN THE FABRIC TO BE SEWED TO THE SURFACE OF THE MACHINE, BY A YIELDING PRESSURE, which is used in all Sewing Machines-has been fully established. The Wheeler & Wilson, and the Grover & Bakers Sewing Machines, as we allege, each infringe three distinct patents owned by us.-We have suits against them in several of the U.S. Courts, which will soon be tried. We hereby caution the public not to buy infringing Machines, as they can be compelled by law to stop using them, and to pay costs and damages (Daily Scioto Gazette, 25 October, 1856).

24 Bays, Encyclopedia, p. 15.
25 Cooper, Sewing Machine.

8

 

The Albany Agreement

To resolve such litigation, Singer, along with Wheeler & Wilson, Grover & Baker and Elias Howe formed a patent pool on 24 October, 1856 (Albany Agreement, Singer Papers, Box 225).26

Their pool combined nine complementary patents that were necessary to build a sewing machine. (See Appendix A for a list of the pool patents.) It survived for the duration of all original patents; no additional patents were added after 1856.27

After Howe's lock stitch patent expired in 1867 the pool continued on the strength of Wheeler and Wilson’s patent on the “four-motion” feeding mechanism and Singer's patent on the horizontal surface until 8 May 1877, when the last patent expired.28

Consistent with theoretical predictions, the sewing machine pool lowered license
fees, especially for pool members and in comparison with Howe’s original fees.
29

26 Searches of the New York Times, Chicago Tribune, and Scientific American yielded no indication that the pool was anticipated by the members. As the first patent pool is U.S. history it is unlikely that the pool was anticipated. In fact, Grover & Baker did not contribute any patents to the pool but was included because its
president, Orlando Potter, devised the idea of a patent pool. The settlement of the sewing machine wars (out of court) was announced after the pool had already been established: “Sewing Machines-To the Public- The undersigned hereby give notice that all suits and controversies in relation to the infringement of patents upon sewing machines between the Wheeler & Wilson Manufacturing Company, I. M. Singer & Co and the Grover & Baker Sewing Machine Company have been amicably arranged and settled. WHEELER & WILSON Manf'g Co., I. M. SINGER & CO., GROVER & BAKER, S. M. Co. (
New York Herald, 22 November, 1856).
27 More formally, the sewing machine pool did not incorporate a grantback clause, which requires any additional patents that are granted to pool members after the agreement to be offered to other pool members for licensing without fees (Layne-Farrar and Lerner, “Patent Pool Participation”, p. 10).
28 Wheeler and Wilson's four-motion feed was “so superior but few first-class machines are made without it” (Knight, American Mechanical Dictionary, p. 2102) and made the firm “several million dollars” (New York Times, 26 January, 1875). It was renewed twice and expired in 1873. The great feature of the Singer (Bachelder) patent “was the production of a sewing machine in which the cloth to be sewn is supported horizontally and is fed through the machine perpetually. His machine was the first sewing machine in which the cloth was supported horizontally and advanced by an automatic feed of any kind. It is scarcely possible to estimate sufficiently the importance of such an invention in the art of sewing by machinery” (Potter et al. v. Braunsdorf, F. Cas. 1132 1869; Knight, American Mechanical Dictionary, p. 2102. Singer had purchased this patent from John Bachelder (Cooper, Sewing Machine, p. 23).
29 E.g., Shapiro, “Navigating the Patent Thicket” and Gilbert, “Antitrust.”

8

 

From 1856 onwards, pool members paid $5 per machine and other firms paid $15 while the average sewing machine cost $65.30

The pool further reduced license fees when Howe’s
patent was renewed in 1860: licensing fees decreased to $1 for members and $7 for other
firms. After Howe’s patent expired in 1867, the pool abolished fees for its members and
further reduced fees for other firms to $5.
31 Despite their lower levels, differences across
members and outside firms may, however, have lowered expected profits for outside
firms, and thereby discouraged innovation by outside firms relative to members. The
pool licensed freely to
…any manufacturer who had a meritorious machine that was not an offensive
imitation of the machine of some other licensed manufacturer (Frederick Bourne,
President of the Singer Company, 1883 to 1905, from Depew,
One Hundred Years, p.
530).
In fact, Elias Howe, who did not produce sewing machines, demanded that the pool
patents should be licensed to at least 24 licensed manufacturers at any given point in
time.
32 Once outside firms had bought a license, the pool placed no restrictions; most
importantly, licensees were able to set prices without intervention by the pool.
33
The pool agreement, however, stipulated that part of the license fee was to be set
aside to support its litigation fund, which was to be maintained above $10,000,
approximately a quarter million 2007 dollars, exceeding the annual sales of most small
30
Bays, Encyclopedia, p. 15.
31 Five dollars per machine is the amount that the Florence Sewing Machine Co. paid to the pool in 1869.
Some licensees may have paid even lower fees; Florence sued because the pool appeared to have extended
more favorable terms to another firm. Its complaint, however, was dismissed (
Florence Sewing Machine
Co. v. Singer Manufacturing Co
., 9 F.Cas. 302, 1870).
32
Cooper, Sewing Machine.
33 Hounshell, American System, p. 68; and Depew, One Hundred Years, p. 530. Archival materials suggest
that the pool may have attempted to limit the supply of its licensees. For example, the original
Albany
Agreement
stipulated that licensees should not produce more than 1,000 machines per year. Licensing data,
however, suggest that licensees were able to exceed their allotted quota without punishment. For example,
Ladd & Webster produced 1,788 machines in 1859.

10
manufacturers.
34 This litigation fund more than sufficed to protect the pool’s patents,
and members divided remaining funds between each other.
THE DATA
To examine the pool’s effects on innovation, we have constructed a rich new data
set, which combines annual data on patents by pool members, licensees and other nonmembers with data on improvements in sewing speeds as an objectively quantifiable
measure of innovation. Legal records on litigation, as well as alternative measures for
changes in the demand for sewing machines complement these data.
Sewing Machine Patents by Pool Members, Licensees, and Other Firms
To measure changes in inventive activity over time, we count the number of U.S.
patents for sewing machines per year, distinguishing patents of pool members from
patents of other firms (Figure 2). Patent data are drawn from
Knight’s Mechanical
Dictionary
(1877) and the Annual Reports of the United States Patent Office.35
Data on sewing machine manufacturers, along with their age and other
information are drawn from collectors’ manuals for antique sewing machines. One
hundred and seventy new manufacturers entered the industry between 1845 and 1885.
36
Entry and exit dates are available for 135 of these firms, nearly 80 percent; few
companies, however, survived for more than ten years.
34 Using the Consumer Price Index, $10,000 in 1856 is worth $252,240.88. Other indicators place the value
of the litigation fund between $191,841 and $34,443,798 (Williamson, “Six Ways”).
35 Knight’s Mechanical Dictionary is a useful complement to the records of the United States Patent Office
because Knight divides sewing machine patents for 1842 and 1874 into nine functional categories: (1)
sewing machines making the chain-stitch, (2) sewing machines making the lock-stitch, (3) sewing
machines for sewing leather, (4) feeding devices, (5) button-hole sewing, (6) miscellaneous parts, (7)
attachments, (8) tables and stands, and (9) motors. We exclude data for “table and stands.”
36
Cooper, Sewing Machine.
11
Patents are matched to firms based on firm names (e.g. Wilson, A. B. of Wheeler
and Wilson) as well as information on other inventors who assigned patents to a firm
(e.g. patents assigned by Edward Clark to the Singer Manufacturing Co.). Thirty-five
firms licensed the pool’s patents between 1853 and 1877, 27 percent of 129 total entrants.
To measure the timing of inventions, we record the grant dates for all patents;
filing dates, as an alternative measure, are only available after 1873.
37
Improvements in Sewing Speed
One potential problem with using patents as a measure of innovation is that
inventors may use patents strategically, so that changes in patenting do not accurately
reflect changes in innovation. For example, prospective pool members may patent
existing innovations more aggressively prior to the pool because they want to improve
their negotiating position relative to other members. Similarly, non-members may patent
more aggressively after the pool has formed to protect themselves from litigation.
38
To address this issue, we construct an alternative measure of innovation that
quantifies improvements in the performance of sewing machines for member and nonmember machines. Specifically, we examine changes in the speed of sewing machines,
as measured by the number of stitches that a machine can perform within one minute.
39
37 Grants typically occurred six months to a year after an inventor applied for a patent (e.g., Thomson,
“Learning by Selling,” p. 435), depending on the complexity of applications and the workload of examiners
(Popp et al., “Time in Purgatory”). To measure this lag for 19
th-century sewing machines, we compared
application and grant dates for a random sample of 100 sewing machine patents between 1873 and 1875.
In this sample the average patent was granted roughly six months (140 days) after the application. These
data corroborate the standard assumption that sewing machine patents were issued about six months after
the application date (e.g,, Thomson, “Learning by Selling,” p. 435).
38 See, e.g. Hall and Ziedonis, “Patent Paradox”; and Shapiro, “Navigating the Patent Thicket.”
39 A major benefit of this measure is that speed can be quantified objectively. In comparison, alternative
measures, such as the number of known stitch types, would be significantly more subjective.
Knights
Mechanical Dictionary
, lists 68 distinct stitch types in 1874 (Knight, American Mechanical Dictionary, pp.
12
We have constructed data on sewing speeds for 1845 to 1900 from 19
th-century
issues of the
Scientific American, company records, trade cards, and reports from
international technology shows, such as the Crystal Palace Exhibition of 1851.
40 For
example, a description of work in a clothing factory provides data on sewing speed for
Wilcox & Gibbs and Wheeler & Wilsons in 1889.
The linings are made by a Wilcox & Gibbs machine, the topping is done by a two needle
Union machine which runs at the rate of about 2220 stitches per minute, the closing is done by
a Weed feed machine, and the other work is done by Wheeler & Wilson's machines which run
at the rate of about 1000 stitches per minute (Tullidge,
Tullidge's Histories, p. 368).
Litigation between Pool Members and Outside Firms
Litigation data are drawn from Westlaw’s pre-1945 federal courts database.
Westlaw combines reports from the series
Federal Cases (district and circuit court
decisions decided before 1880, compiled in 1892), the
Federal Reporter (district and
circuit court decisions from 1880 onwards), the
Supreme Court Reports and United States
Reports
(decisions of the Supreme Court).41 Westlaw reports 100 patent disputes
2123-4), but it is difficult to establish how distinct, or useful these stitches are from each other. For
example, Knight’s data include several embroidery stitches, but these stitches may not be as distinct from
each other (or as useful) as two stitches to work with leather versus cloth.
Because the type of a stitch that a sewing machine performs influences its sewing speed, we focus on lockstitch machines, which were known to produce the most durable stitch. Machines producing single thread
stitches, such as Wilcox and Gibbs chain stitch machines were generally faster than lock-stitch machines,
but a single stitch proved so much less durable that all but a few firms eventually abandoned it (James,
Upholstery Tips, p. 86). Among the pool members, Singer and Wheeler & Wilson had focused on lockstitch machines from the start (Bays, Encyclopedia), while Grover & Baker manufactured double thread
chain-stitch machines until 1875, when they sold to a competitor who switched production to lock-stitch
machines (Depew,
One Hundred Years, p. 528). Improvements in the design of shuttles and the adoption
of cranks in drive mechanisms (instead of springs and cams) are the most likely sources of improvements in
the speed of sewing (Thomson,
Path, p. 148). Although Singer introduced a machine with an attached
electric motor in 1889, the foot-powered treadle remained the most effective drive mechanism throughout
the 19th century (Depew,
One Hundred Years, p. 534).
40 A special report in the Thirteenth Annual Report of the Commissioner of Labor (Hand and Machine
Labor)
, contains comparisons of sewing machine and hand productivity for 1895. Unfortunately, these
data cannot be linked to earlier years or specific sewing machines.
41 We search for cases that include the terms “sewing machine” and “patent infringement.” Broader search
terms, such as “patent” or “sewing,” did not produce additional cases.

13
regarding sewing machines between 1850 and 1885.
42 As a robustness check for the
Westlaw data, we also searched William Robinson’s
Law of Patents, which includes two
cases that were omitted from Westlaw. These data indicate that the pool (as a whole)
engaged in 19 cases; pool members (acting independently) engaged in another 23
disputes.
43
A potential concern with litigation data is that they only capture cases that were
decided in court and may therefore underestimate the real risk of litigation. For example,
outside firms may have been more willing to settle and less likely to initiate legal action
because they were afraid to face the pool in court.
44
Potential Sources of Bias and Measurement Error
Most importantly, the firm-level data are subject to survivorship bias, because
19
th-century data on entry and exit are more likely to be available for larger and more
successful firms. Collectors’ manuals record entry and exit dates for 40 percent of all
manufacturers in the U.S. census in 1860 (29 of 74 firms), 63 (21 of 49 firms) in 1870,
and 36 percent (36 of 101 firms) in 1880. Since missing firms are more likely to be
small, our data may underestimate the number of small, outside firms before and after the
42 These reports publish the date of the decision but do not include the filing date. Anecdotal evidence
suggests that decisions were made up to two years after filing. For example, Walter Hunt, who claimed to
be the rightful inventor of the lock-stitch, appealed against the Commissioner of Patents on 24 May, 1854.
This case was decided in Howe’s favor in the Circuit Court of the District of Colombia in February 1855.
See
Hunt v. Howe, 12 F. Cas. 918 (1855). A similar case that Howe filed against William Bradford by
Howe between 1850 and 1851 was decided in 1852 (
Scientific American, 1852, p. 356).
43 The Combination was a plaintiff in one additional case (Potter v. Hicks, 19 F. Cas. 1154, Mossoff,
”Stitch,” p. 40, no dates are available). Ten of these 102 cases were retrials and appeals. The records list an
additional case (
Potter v. Stewart, 7 F. 215) in 1881, four years after the pool dissolved, when pool
members sued for past infringement of Bachelder’s patent (USPTO patent 6,439).
44 Another source of bias is that historical case data are based on contributions of individual reporters,
rather than a centralized system, which may create bias toward general interest decisions. Khan, “Property
Rights,” p. 94.

14
pool relative to pool years, which implies that we may overestimate innovation by outside
firms during the pool.
45
In addition, patent data, especially when electronically collected, are subject to
measurement error. For example, optical character recognition cannot always distinguish
the 19
th-century script for letters R from B, P from F, and U from J. To address this
issue, we hand-checked random samples of the data and replicated every search with the
corresponding misspelled letters (e.g., “Singee” for Singer).
Measures for the Intensity of the Civil War and Changes in Demand
Across industries, disruptions as a result of the Civil War (1861 to 1865) may
have discouraged innovation. For the sewing machine industry, however, the war may
have also encouraged innovation, as it increased demand for machine-made uniforms.
Before 1861 the U.S. Army manufactured its own uniforms and was reluctant to experiment
with machine-made clothing, fearing that the product would be inadequate for combat duty,
especially under rough frontier conditions. The sewing machine was initially confined,
therefore, to stitching caps and chevrons. With the outbreak of war, the contracts signed for
military clothing did not specify a method of manufacture. Suppliers soon turned increasingly
to machine-made apparel to assure standard quality and to meet contract deadlines (Whitten
and Whitten,
Handbook, pp. 91-92).
In addition to increasing its demand for machine-made clothing, the war department
placed direct orders for machines:
the Civil War brought a demand by the War Department for a million uniforms. The
government contracted for thousands of sewing machines; these were loaned free to sewing
circles which worked on the making of uniforms (Crow,
Great American Customer, p. 205).
45 Small sewing machine manufacturers that are missing from the data produced between 500 and 1000
firms over their lifetime (Bays,
Encyclopedia). As an alternative check of the data, we also examined city
directories for Philadelphia (1850, 1855, 1860, 1870, 1880, and 1890) and New York (1850, 1851, 1860,
1869, 1880, and 1890). In the New York directories, most entries are manufacturers, and the data closely
match the information in Cooper (
Sewing Machine). In comparison with the New York directories, the
Philadelphia directories include a large number of sewing machine agents (e.g., Edward Jones, and Henry
Co. in 1860).

15
Thus, 19
th-century accounts suggest that the Civil War resulted in a significant positive
demand shock, which may have encouraged innovation in sewing machines.
46 To
measure the size of changes in demand we include data on increases in the size of the
Union Army between January 1861 (16,267 soldiers) and January 1865 (959,460
soldiers, Figure 4).
47 Most of our estimations focus on increases in the number of Union
soldiers; Confederate uniforms were rarely machine-stitched and both sides reused
uniforms.
48 Results, however, are robust to alternative measures of the intensity of the
war, including the number of Confederate soldiers.
To control for alternative sources of changes in demand, we include data on
population growth and real GDP, which serve as proxies for demand by individual
families.
49
EMPIRICAL RESULTS
Did the Pool Reduce Litigation?
Theoretical models of patent pools predict that pools lower the threat of
litigation for their members.
50 In the case of the sewing machine, the pool's prospective
members had engaged in crippling litigation, which was ended by the pool.
51 Between
46 See e.g., Schmookler, “Invention”; and Sokoloff, “Inventive Activity.”
47 Long and Long, Civil War.
48 Drew and Snow, Eagle’s Talons, p. 113. The daughter of Mrs. Robert M. Patton of Florence, Alabama,
remembers “I was just through college, and I made, together with my mother and a serving woman she had
employed, uniforms, underwear, and several overcoats, so heavy that we had to work on them while lying
on a table. Every stitch was done with our fingers. We had no machines until 1869, when my father bought
me one” (Andrews,
Women, p. 427).
49 United States Census Bureau, 1850-1890, and Johnston, “U.S. GDP.” Godfrey, International, p. 125,
notes that ``the majority of the American sewing machine firms were concerned mainly with producing
small domestic machines for use in the home, while only a small number were engaged in building the
heavier type of machine which could be used for industrial purposes."
50 See e.g., Choi, “Patent Pools.”
51 See e.g., Cooper, “Sewing Machine,” p. 41.
16
1856 and 1877, while the pool was active, it was a defendant in only 3 of 55 total legal
disputes (Figure 7) all of which concerned the same complaint: The pool had granted a
more favorable license to a competitor of the Florence Machine Company, and Florence
sued to get its license fee reduced; the court decided in favor of the pool.
52 Singer was a
defendant in two additional cases in 1859 and 1860.
53 Both were based on the complaint
that Singer had patented another firm's invention and were decided in Singer's favor.
54
Contrary to the results for member firms, litigation data suggest that the pool
increased litigation risks for outside firms. Between 1856 and 1876 the pool initiated 15
legal battles, more than a quarter of all 55 sewing machine cases.
55 Pool members acting
independently initiated another 9 legal disputes. Forty-nine of these 64 cases were
directed at outside firms (Figure 6). Thus, the pool may have increased litigation risks for
outside firms even as it lowered such risks for its members.
Did the Pool Encourage its Members to Patent More?
The most central prediction of the theoretical literature, however, is that patent
pools encourage innovation by increasing expected profits for member firms and
reducing the costs of licensing for outside firms.
56 To test this prediction, we examine
52 Two subsequent cases concerned payments by Florence to a judge and jurisdictional concerns (not all
parties were resided in the same state). See
Florence v. Singer Manufacturing Co., F. Cas. 302 (1870),
Florence v. Singer Manufacturing Co., F. Cas. 310 (1871), and Case of the Sewing Machine Cos., 85 U.S.
553 (1873), and
Florence Machine Co. v. Grover & Baker Sewing Mach. Co., (1872), 110 Mass. 70.
53 Wickersham v. Singer, F. Cas. 1134 (1859) and Wilson v. Singer, F. Cas. 217 (1860).
54
In Wilson v. Singer, F. Cas. 217 (1860), Wilson sought the reissue of a patent that incorporated
improvements which had been patented by Singer. Though the court initially ruled in Wilson’s favor, his
patent was ruled invalid in subsequent litigation (
Potter v. Dixon (1863), 2 Fisher, 381).
55 All but 2 of 15 disputes in which the pool was the plaintiff concerned Wilson’s “four motion” feeding
mechanism (USPTO patent 7,776).
56 See e.g., Dequiedt and Versaevel, “Patent Pools.”
17
whether patenting increased in the years leading up to the pool and continued at a higher
level while the pool was active. While theoretical analyses focus on effects on pool
members, we also extend the analysis to include effects on other firms, which accounted
for the majority of patents and improvements.
If the prospect of a pool encourages a patent race among prospective members,
firms that succeed in joining the pool should have a larger number of patents in the years
leading up to the pool. These predictions are borne out in the data; in 1855, three
members were granted a total of 10 patents, compared with an average of less than three
patents per year for 1850 to 1854 (Figure 1). This increase in patenting, however, is also
consistent with the alternative hypothesis that prospective pool members Singer, Wheeler
& Wilson, Grover & Baker and Elias Howe patented more to protect themselves from
litigation. Prospective members may also have patented existing innovations more
aggressively to improve their bargaining positions in negotiations for the pool.
57 Thus, 9
of the 10 member patents in 1855 were granted to Singer, even though the company
contributed no significant in-house inventions to the pool.
58
Interestingly, these firms began to patent less as soon as the pool had been
established, producing an average of three patents per year from 1857 to 1861 and only
two patents per year from 1866 to 1870. Members continued to patent less until the pool
dissolved in 1877, and quickly resumed patenting afterwards, producing five patents in
1878, nine in 1879, and eight in 1880.
Did the Pool Encourage Outside Firms to Patent More?
57 Dequiedt and Versaevel, “Patent Pools.”
58 Singer purchased its key patent for the horizontal work surface from Bachelder (Cooper, Sewing
Machine
, p. 23).
18
In contrast to patents by members, patents by outside firms spiked
after the pool
had been established (Figure 2). Annual patents jump from 29 patents in 1856 to more
than 100 patents in 1858.
59 Similar to the case of member patents, this spike in patenting
may, however, represent a strategic response to a heightened threat of litigation with a
more powerful opponent, rather than a true increase in innovation.
60
Patenting by non-members began to decline only two years after the pool had
formed. By 1862, annual patents had fallen to 21 patents per year, below pre-pool levels.
Patenting recovered after the Civil War increased demand for sewing machines (to 150
patents in 1873). After this peak, patenting declined again until the pool dissolved in
1877. By 1880, only three years after the pool dissolved, patenting began to rise again,
increasing to nearly 300 patents in 1882.
Comparing sewing machine patents with aggregate counts of U.S. patents also
indicates that the pool discouraged patenting. An initial increase of sewing machines to 3
percent of all U.S. patents in 1858 had eroded to less than 1 percent by 1866 (Figure 3).
Sewing machine patents recovered after the pool dissolved, to 2 percent of all U.S.
patents in 1882. Thus, comparisons with total patent counts confirm the results above
that the pool discouraged patenting by both members and other firms.
Difference-in-Differences Estimation
To examine the patent data more systematically, we estimate difference-indifferences regressions that compare annual patents by pool members relative to other
59 Patenting rates in the late 1850s may be an overestimate relative to the early 1850s, when patent
solicitors successfully lobbied for the removal of examiners whom they thought to be too strict in assessing
novelty. Post (“Liberalizers,” p. 52) argues that the share of patent applications that were granted increased
as a result of this change; grant rates increased from 32 percent in 1853 to 67 percent in 1859.
60 See e.g., Shapiro, “Navigating the Patent Thicket”; and Bessen, “Patent Thickets.”
19
firms during the pool with annual patents by pool members relative to other firms before
and after the pool.
Pool Members are Grover and Baker, Singer, and Wheeler and
Wilson;
Pool equals 1 in years when the pool was active for the entire year (1857-
1876).
61
(1) pit = + PM Pool Memberi + PMP Pool Memberi * Poolt
+ L Licenseei + Age log Firm Ageit + t + i + it
The coefficient PMP on the interaction Pool Member * Pool is our difference-indifferences estimator; it measures the increase in patenting by pool members relative to
other firms. Under the assumption that omitted time-varying effects are uncorrelated
with pool membership this coefficient can be interpreted as the causal effect of the pool:
If
PMP is positive and statistically different from zero, the pool encouraged members to
patent more. Similarly,
licensees may have patented more while the pool was active if
they benefitted from reductions in license fees and transaction costs.
We also control for
Firm Age as a characteristic that may affect patenting
regardless of the pool. Specifically, younger firms may patent more because they are
more likely to be based on novel technologies. Alternatively, older firms may patent
more because they are more established and familiar with the patent system.
62
61 As a result some of our untreated years were actually partially treated by the pool, so that we may
underestimate the real effect of the pool. We miss two months of pool activity in 1856, because the pool
was founded in October 1856, and four months in 1877, because the pool dissolved in May 1877.
62 For sewing machines models between 1853 and 1882, sales have been found to encourage technological
change (Thomson, “Learning by Selling”). In our sample, sales data are available for 30 firms that were
licensed by the pool between 1867 and 1876 (20 percent of all firms, yielding a total of 163 observations).
In this subsample of the data, the coefficient of correlation between annual sales and annual patents is
0.185. The coefficient is 0.192 assuming a two year grant lag and 0.153 assuming a one year grant lag.

20
To account for unobservable factors that may influence patenting we include
annual and firm fixed effects (
t and i ). For example, the dismissal of strict examiners
in the early 1850s or changes in the demand for sewing machines may help to increase
the annual number of patents, independently of the pool. Similarly, certain firms may
patent more because they are more creative regardless of age or the existence of a pool.
Regression results confirm that members patented less while the pool was active.
Estimates for
PMP indicate that members produced an average of 1.6 fewer patents per
year while the pool was active (Table 1, significant at 5 percent). In comparison, across
all years, pool members produced between 2.0 and 2.2 more patents per year. Additional
regressions that estimate year-specific treatment effects indicate that members generated
between 0.8 and 2.3 fewer patents in the last ten years of the pool compared with the preand post-pool period.
63
The data also yield only limited evidence that the pool encouraged licensees to
patent more. Licensees produced up to 0.2 additional patents per year than other firms,
but this effect is not statistically significant (Table 1, I and III). Interactions between
Licensees and Pool are never statistically significant. Another interesting result from the
data is that, all else equal, younger sewing machine manufacturers patented more: A 1
percent decrease in age adds about 0.2 patents per year (Table 1, IV, significant at 5
percent).
How did the Civil War Affect Patenting?
63 Coefficients are negative for all 20 years when the pool was active (1857-76), and they are significant at
the 5 percent level for 12 of these years: 1857, 1860, 1863, 1865 to 68, and 1871 to 75.

21
Alternative specifications control for the Civil War and other factors that may
have influenced demand. Specifically, we examine the potential impact of increases in
the size of the Union Army, real GDP, and population. Anecdotal evidence suggests that
Singer and other pool members benefitted disproportionately from increases in the
demand of sewing machines, with Singer selling “tens of thousands machines.”
64 To
identify these effects, we estimate interaction terms between firms’ affiliation to the pool
(distinguishing members and licensees) and measures for the intensity of the Civil War.
To account that demand effects may operate with a lag, we also include one-year lagged
increases in the size of the Union Army.
65 Linear time trends capture changes in
patenting over time that may be independent of the pool.
66
(2) pit = + P Pooli + PM Pool Memberi + PMP Pool Memberi * Poolt
+ L Licenseei + Age log Firm Ageit + MLF Member Feet
+ NLF Non-member Feet + UA Union Armyt
+ UAP Union Armyt * Pool Memberi + UAL Union Armyt * Licenseei
+ GDP real GDPt + Pop log Populationt + t + it
Regression results from this alternative specification confirm results of
regressions with time fixed effects that the pool discouraged its members from patenting.
Across all years, members produced about 2 additional patents per year compared with
other firms (at 1 percent significance). In comparison, members produced 1.6 fewer
patents per year while the pool was actively (at 1 percent significance, Table 2).
Similar to regressions with time and firm fixed effects, regressions with controls
for specific demand factors yield no significant evidence that licensees patented more
64 “tens of thousands of Singer’s machines stitched uniforms worn by Union armies in the Civil War.”
Botti,
Envy, p. 97
65 Longer lags are not significant and do not change the main results.
66 Coefficients are robust to alternative specifications of time trends up to the 4th order polynomial.
22
while the pool was active or in response to lower license fees. In fact, increases in nonmember license fees are positively correlated with increases in the number of licensee
patents per year. Thus, because license fees might be endogenous to patenting, we
exclude license fees and focus on alternative controls (Table 2, II-V).
Interestingly, the data indicate that the Civil War discouraged innovation in the
sewing machine industry, despite its effects on demand. For every 10,000 increase in the
number of Union Army soldiers, the average non-member firm produced between 0.004
and 0.006 fewer patents per year (at 1 percent significance, Table 2, II-V). Thus, the data
suggest that the disruptions caused by the war outweighed its positive demand effects for
outside firms. Consistent with the narrative evidence, however, the Civil War appears to
have encouraged patenting by pool members. For every 10,000 increase in Union Army
soldiers, pool members produced 0.01 additional patents per year (at 10 percent
significance, Table 2).
67
To account for the large number of zeros in the dependent variable we repeat the
analysis as Poisson and negative binomial regressions.
68 All key results are robust to this
alternative specification. Across the entire sample, pool members patented more than
other firms (producing between 5 and 7 times as many patents per year, at 1 percent
significance, Table 3). The coefficient on the interaction
Pool Memberi * Poolt indicates
that the pool reduced the number of member patents by 50 to 57 percent (at 5 percent
67 The data suggest that the effect was strongest in the same year, perhaps because the war department
demanded uniforms before new soldiers would join the war. Our main results are robust to including
absolute numbers of Union Army soldiers rather than changes in the number of soldiers.
68 In these regressions, patents per year, pit, is assumed to be an exponential function of firm specific
characteristics,
Xit, and a series of controls, it: E[pit|Xit]=exp(Xit+t) where i indexes the firm and t
indexes the year. Negative binomial and Poisson regressions take into account the fact that the distribution
of patents is skewed towards zero; negative binomial regressions also allow the variance of the dependent
variable to be larger than its mean. In our data set, the mean number of patents per year is 0.43 and the
variance is 1.15. This suggests that a negative binomial model is preferred to a Poisson model.

23
significance, Table 3). Similar to OLS, Poisson and negative binomial regressions yield
no evidence that licensees patented more.
69 The coefficient on Union Armyt * Pool
Member
i is significant at 5 percent confirming OLS results and narrative evidence
suggesting that the war encouraged pool members to patent more (Tables 3 and 4).
Did the Pool Encourage Technical Progress?
Two key findings of the patent data suggest that the pool may have encouraged
changes in strategic patenting rather than innovation. First, the increase in member
patents immediately preceding the pool suggests that prospective members may have
patented existing innovations more aggressively to strengthen their bargaining position
relative to other members.
70 Second, the spike in patenting for non-member firms
immediately after the creation of the pool may represent a strategic response by nonmembers to a heightened threat of litigation.
71
To separate changes in strategic patenting from changes in innovation, we
examine data on sewing speeds as an objectively quantifiable measure of performance.
Data on sewing speeds confirm that the pool slowed rather than encouraged rates of
innovation (Figure 5). From 1845 until 1856, the maximum number of stitches that a
sewing machine could perform increased from 200 to 2,000 stitches per minute. As soon
as the pool had been established, innovation appears to have come to a halt, and sewing
speeds stayed roughly constant for the duration of the pool. Improvements in sewing
speeds had not yet reached their natural plateau at this time, and continued to advance
soon after the pool dissolved in 1877. By 1889 the maximum speed of sewing had
69 Interactions between Pool and Licensee are not significant and have been dropped.
70 Dequiedt and Versaevel, “Patent Pools.”
71 See e.g., Shapiro, “Navigating the Patent Thicket”; and Bessen, “Patent Thickets.”
24
increased to 2,500 stitches per minute.
72 Outside firms produced the fastest machines in
1889 and 1890 (Hurtu & Hautin, an outside firm, with 2,500 stitches-per-minute in 1889
and Willcox and Gibbs, a licensee with 4,000 stitches-per-minute in 1890).
73
Contemporary accounts indicate improvements in other characteristics of the
sewing machine were similarly delayed by the pool: “the machine was still noisy,
expensive and inefficient when the Civil War brought a demand by the War Department
for a million uniforms.”
74
CONCLUSIONS
This paper has used the example of the 19th century sewing machine industry to
examine empirically whether patent pools encourage innovation. Our data broadly
confirm theoretical predictions that patent pools lower litigation risks for pool members
and that the prospect of a patent pool encourages prospective members to patent more.
In contrast to theoretical predictions, however, pool members began to patent less
as soon as the pool had been established. For example, difference-in-difference estimates
indicate that pool members patented less while the pool was active both relative to the
pre-and post-pool period and compared with other manufacturers. Similarly, the share of
72 Improvements in shuttles and the adoption of cranks in drive mechanisms (replacing springs and cams)
are the most likely sources of improvements in sewing speeds (Thomson,
Path, p. 148). Electrification sets
in after the sample period; although Singer introduced a machine with an attached electric motor in 1889,
the foot-powered treadle remained the most effective drive mechanism throughout the 19th century
(Depew,
One Hundred Years, p. 534). Basic least square estimations confirm that improvements in speed
slowed during the pool years. Specifically, we estimate least squares regressions of stitches-per-minute
with a linear time trend and the
Pool variable as explanatory variable. In such regressions, the coefficient
on
Pool is negative but not statistically significant. As an additional robustness check, we drop
observations that record the fastest speed during the pool years (2,000 per minute); these regressions also
confirm that improvements in the speed of sewing machines decelerated as long as the pool was active.
73 In comparison, modern industrial lock-stitch machines sew at speeds of 6,000 stitches per minute. See
McGrath and Blachford, “Gale Encyclopedia,” p. 3,350.
74
Crow, Great American Customer, p. 205.
25
sewing machine inventions in the U.S. patent data declined during the pool and recovered
only after the pool had been dissolved.
We are also able to improve existing studies of pools by extending the analysis to
include outside firms, which account for the majority of innovations in the sewing
machine industry. These data suggest that the pool discouraged patenting by outside
firms and that it had at best insignificant positive effects on patenting by licensees.
One limitation of using patent data as a measure of innovation is that observed
changes in patenting may reflect firms’ strategic response, for example to threats of
litigation, rather than true changes in innovation. In the case of the sewing machine
pools, prospective members such as I. M. Singer & Co. may have been more likely to
patent existing innovations in years leading up to the pool to protect themselves from
litigation with other prospective members or to improve their bargaining position at the
creation of the pool. Similarly, outside firms may have begun to patent more after the
pool had been established to protect themselves from litigation with the pool, which made
for a substantially more powerful opponent in a court of law.
To separate strategic effects from increases in innovation, we constructed an
alternative data set of objectively measurable improvements in the performance of
sewing machines (stitches per minute). These data confirm that innovation slowed while
the pool was active and only began to accelerate again after the pool dissolved in 1876.
Interestingly, significant increases in the demand for sewing machines as a result of the
Civil War had only minimal effects on the sewing machine industry as a whole, although
they appear to have encouraged patenting by pool members.

26
Thus, evidence from the sewing machine industry challenges theoretical
predictions and regulators’ expectations that patent pools encourage innovation. What are
the mechanisms by which pools discourage innovation? The experience of the sewing
machine suggests two main channels by which pools may discourage innovation,
particularly by outside firms: increases in litigation risks and changes in the direction of
R&D. Increased litigation risks for outside firms lower expected profits and discourage
investments in R&D. Increases in litigation may, however, also encourage outside firms
to divert their research efforts away from improving key technologies that are covered by
the pool towards substitutes that are still “freely” available.
75
REFERENCES
Andrews, Matthew Page. The Women of the South in War Times. Baltimore: Norman,
Remington Co., 1920.
Bays, Carter.
The Encyclopedia of Early American and Antique Sewing Machines:
Identification and Values.
Kentucky: Collector Books, 2007.
Bessen, James. “Patent Thickets: Strategic Patenting of Complex Technologies.” SSRN
Working Paper, March 2003.
_________. “Imperfect Property Rights.” Working Paper, Research on Innovation
(www.researchoninnovation.org), 2009.
Bittlingmayer, George. “Property Rights, Progress, and the Aircraft Patent Agreement.”
Journal of Law and Economics 31, no. 1 (1998): 227-48.
Botti, Timothy.
Envy of the World: A History of the U.S. Economy & Big Business. New
York: Algora Publishing (2006).
Choi, Jay Pil. “Patent Pools and Cross-Licensing in the Shadow of Patent Litigation.”
CESifo Working Paper No. 1070, 2003.
Clark, Jeanne, Joe Piccolo, Brian Stanton, et al. “Patent Pools: A Solution to the Problem of
Access in Biotechnology Patents?” White Paper, United States Patent Office, 2001.
75 We examine these effects in Lampe and Moser, “Patent Pools and the Direction of Technical Change.”
27
Clarkson, Gavin. “Objective Identification of Patent Thickets: A Network Analytic
Approach for Measuring the Density of Patent Space.” Ph.D. thesis, Harvard University,
2004.
Cohen, Wesley, Richard Nelson, and John Walsh. “Protecting Their Intellectual Assets:
Appropriability Conditions and Why U.S. Manufacturing Firms Patent (Or Not).” NBER
Working Paper W7552, 2000.
Cooper, Grace Rogers.
The Sewing Machine: Its Invention and Development. Washington,
DC: The Smithsonian Press, 1976.
Crow, Carl.
The Great American Customer. New York: Ayer Publishing, 1970.
Depew, Chaunce M.
One Hundred Years of American Commerce. New York, NY: D.O.
Haynes and Co., 1968.
Dequiedt, Vianney, and Bruno Versaevel. “Patent Pools and the Dynamic Incentives to
R&D.” Grenoble Applied Economics Laboratory, 2004.
Drew, Dennis and Donald Snow.
The Eagle’s Talons: The American Experience at War.
Maxwell Air Force Base, AL: Air University Press, 1988.
Dykman, Harry. “Patent Licensing within the Manufacturer's Aircraft Association (MAA).”
Journal of Patent Office Society 46, (1964): 646-59.
Gaule, Patrick. “Towards Patent Pools in Biotechnology?” Mimeo, Ecole Polytechnique
Federale de Lausanne, 2006.
Gilbert, Richard. “Antitrust for Patent Pools: A Century of Policy Evolution.”
Stanford
Technology Law Review
, 2004.
Godfrey, Frank P.
International History of the Sewing Machine. London: Trans-Atlantic
Publications, 1982.
Hall, Bronwyn, and Rosemarie Ziedonis. “The Patent Paradox Revisited: An Empirical Study
of Patenting in the U.S. Semiconductor Industry, 1979-1995.”
RAND Journal of Economics
32, no. 1 (2001): 101-28.
Heller, Michael A. and Rebecca S. Eisenberg. “Can Patents Deter Innovation? The
Anticommons in Biomedical Research.”
Science 280, no. 5364, (1998): pp. 698 – 701.
Heller, Michael.
The Gridlock Economy. New York, NY: Basic Books, 2008.
Hounsell, David.
From the American System to Mass Production, 1800-1932. Baltimore,
MD: John Hopkins University Press, 1984.
James, David.
Upholstery Tips and Hints. Lewes: Guild of Master Craftsman Publications,
2000.

28
Johnston, Louis D., and Samuel H. Williamson. “What Was the U.S. GDP Then?”
MeasuringWorth, 2008.
Khan, B. Zorina. “Property Rights and Patent Litigation in Early Nineteenth-Century
America.”
Journal of Economic History 55, no. 1 (1995): 58-97.
Knight, Edward. Knight's American Mechanical Dictionary. New York: Hurd and
Houghton, 1877.
Lampe, Ryan and Petra Moser. “Patent Pools and the Direction of Technical Change:
Evidence from the Sewing Machine Combination, 1856-1877.” Working paper, Stanford,
2009.
Layne-Farrar, Anne and Josh Lerner. “To Join Or Not To Join: Examining Patent Pool
Participation and Rent Sharing Rules.” Working Paper, 2008.
Lemley, Mark and Carl Shapiro. “Patent Holdup and Royalty Stacking.”
Texas Law Review
85 (2007).
Lerner, Josh. "Patenting in the Shadow of Competitors."
Journal of Law and Economics 38,
no. 2 (1995): 463-95.
Lerner, Josh, and Jean Tirole. “Efficient Patent Pools.”
American Economic Review 94, no.
3, (2004): 691-711.
Lerner, Josh, Marcin Strojwas, and Jean Tirole. “The Design of Patent Pools: The
Determinants of Licensing Rules.”
RAND Journal of Economics 38, no. 3, (2007): 610-625.
Long, E.B. and Barbara Long.
The Civil War day by day; an almanac, 1861-1865. Garden
City, New York: Double Day & Company, 1971.
McGrath, Kimberley, and Stacey Blachford.
The Gale Encyclopedia of Science:
Phosphorus-Spectrum
. Detroit, MI: Gale Group, 2001.
Merges, Robert. “Contracting into Liability Rules: Intellectual Property Rights and Collective
Rights Organisations.”
California Law Review 84, no. 5 (1996): 1293-1385.
_________. “Institutions for Intellectual Property Transactions: The Case of Patent Pools.”
University of California at Berkeley Working Paper, 1999.
Moser, Petra. “Why Don't Inventors Patent?” Working paper, Stanford University,
Available at SSRN:
http://ssrn.com/abstract=930241, 2008.
Mossoff, Adam. “A Stitch in Time: The Rise and Fall of the Sewing Machine Patent
Thicket.” George Mason Law & Economics Research Paper No. 09-19. Available at SSRN:
http://ssrn.com/abstract=1354849, 2009.
National Museum of American History, Library of the Smithsonian Institution Libraries,
Trade Literature Collection. Washington, DC.
29
Popp, David, Ted Juhl and Daniel Johnson. “Time in Purgatory: Examining the Grant Lag for
U.S. Patent Applications."
Topics in Economic Analysis & Policy 4, no. 1 (2004): Article 29.
Post, Robert. “ ‘Liberalizers’ versus ‘Scientific Mean’ in the Antebellum Patent Office.”
Technology and Culture 17, no. 1 (1976): 24-54.
Robinson, William.
The Law of Patents for Useful Inventions. Boston, MA: Little, Brown,
and Company, 1890.
Schmookler, Jacob.
Invention and Economic Growth. Cambridge, Massachusetts: Harvard
University Press, 1966.
Schmookler, Jacob.
Patents, Invention and Economic Growth - Data and
Selected Essays, Grilliches, Zvi and Hurvicz, Leonid (eds.).
Cambridge:
Harvard University Press, 1972.
Shapiro, Carl. “Navigating the Patent Thicket: Cross Licenses, Patent Pools, and Standard
Setting.” In Jaffe, Lerner, Stern, eds,
Innovation Policy and the Economy, vol. 1, Cambridge,
MIT Press, 2001.
Simon, James, Eric Claassen, Carmen Correa, and Albert Osterhaus. “Managing Severe
Acute Respiratory Syndrome (SARS) Intellectual Property Rights: The Possible Role of
Patent Pooling.”
Bulletin of the World Health Organization 83, (2005): 707-10.
Sokoloff, Kenneth. “Inventive Activity in Early Industrial America: Evidence from Patent
Records, 1790-1846.”
Journal of Economic History 68, no. 4 (1988): 813-850.
State Historical Society of Wisconsin, Papers of the Singer Manufacturing Company,
Madison, Wisconsin, various dates.
Stubbs, Kevin D.
Race to the Front: The Materiel Foundations of Coalition Strategy in the
Great War
. Greenwood Publishing Group, 2002
Thomson, Ross. “Learning by Selling and Invention: The Case of the Sewing Machine.”
Journal of Economic History 47, no. 2 (1987): 433-45.
_________.
The Path to Mechanized Shoe Production in the United States. Chapel Hill:
University of North Carolina Press, 1989.
Tullidge, Edward.
Tullidge's Histories, (volume II) Containing the History of All the
Northern, Eastern and Western Counties of Utah: Also the Counties of Southern Idaho. Salt
Lake City, Utah: Press of the Juvenile Instructor, 1889.
United States Department of Justice and Federal Trade Commission. Antitrust Guidelines for
the Licensing of Intellectual Property.
DOJ and FTC, 1995.
Verbeure, Birgit, Esther van Zimmeren, Gert Matthijs, et al. “Patent Pools and Diagnostic
Testing.”
Trends in Biotechnology 24, no. 3 (2006): 115-120.
30
Whitten, David O. and Bessie Emrick Whitten.
Handbook of American Business History.
New York: Greenwood Publishing Group, 2000.
Williamson, Samuel. “Six Ways to Compute the Relative Value of a U.S. Dollar Amount,
1790 to Present.” MeasuringWorth, 2008.

TABLE 1 – OLS REGRESSIONS WITH FIRM AND YEAR FIXED EFFECTS
DEPENDENT VARIABLE IS PATENTS PER FIRM AND YEAR (1850-1885)
(I) (II) (III) (IV)
Pool Member
a 2.084** 2.016** 2.207**
(0.501) (0.504) (0.513)
Pool * Pool Member -1.659** -1.689** -1.621** -1.513*
(0.549) (0.546) (0.554) (0.612)
(log) Firm Age 0.093* 0.122* -0.213*
(0.046) (0.057) (0.097)
Licensee
b 0.112 0.192
(0.115) (0.126)
Constant 0.408** 0.383** 0.560** 2.027**
(0.109) (0.111) (0.102) (0.599)
Observations 1051 1051 1051 1051
Firm Fixed Effects N N N Y
Year Fixed Effects Y Y Y Y
R-squared 0.17 0.17 0.16 0.32
Notes: Data from Long and Long (1971), and Johnston and Williamson (2008). Standard errors are clustered at the firm level.
** denotes significant at 1 percent level, * denotes significance at 5 percent level.
a: Pool Member = 1 for Singer, Grover & Baker, and Wheeler & Wilson.
b: Licensee=1 for firms who purchased a license from Howe or the pool.

TABLE 2 – OLS REGRESSIONS WITH CONTROLS FOR CIVIL WAR, GDP, AND POPULATION
DEPENDENT VARIABLE IS PATENTS PER FIRM AND YEAR (1850-1885)
(I) (II) (III) (IV) (V)
Pool -0.189 0.015 0.011 0.032 0.122
(0.161) (0.104) (0.103) (0.111) (0.099)
Pool Member
a 2.072** 2.048** 2.047** 1.977** 1.964**
(0.496) (0.504) (0.504) (0.503) (0.500)
Pool * Pool Member -1.610** -1.586** -1.586** -1.627** -1.611**
(0.518) (0.523) (0.522) (0.520) (0.515)
Log Age 0.082 0.081 0.080 0.112* 0.115*
(0.045) (0.045) (0.045) (0.055) (0.055)
Union Army
b -0.004 -0.005* -0.005* -0.006* -0.006*
(0.003) (0.002) (0.002) (0.002) (0.002)
Union Army Lag 0.001 -0.000 -0.000 -0.001 -0.002
(0.002) (0.002) (0.002) (0.002) (0.002)
Union Army * Member 0.010 0.010 0.010 0.011 0.011
(0.006) (0.006) (0.006) (0.006) (0.006)
Union Army (-1) * Member -0.003 -0.003 -0.003 -0.003 -0.004
(0.003) (0.003) (0.003) (0.003) (0.003)
Real GDP
c 6.674 9.498* 9.398* 8.772* 8.463*
(4.418) (4.141) (4.119) (3.917) (3.906)
Log Population 2.049 7.810 7.966 7.329
(5.448) (4.717) (4.632) (4.752)
Licensee
d 0.137 0.130 0.123
(0.127) (0.125) (0.118)
Union Army * Licensee -0.000 -0.001
(0.002) (0.002)
Union Army (-1) * Licensee -0.001
(0.002)
Member Fee -0.040
(0.023)
Non-member Fee 0.047
(0.025)
Linear Time Trend -0.060 -0.222 -0.225 -0.208 -0.027
(0.147) (0.119) (0.117) (0.120) (0.018)
Constant -21.203 -79.243 -80.815 -74.338 -0.339
(54.867) (47.635) (46.787) (47.993) (0.238)
Observations 1051 1051 1051 1051 1051
R-squared 0.15 0.15 0.15 0.15 0.15
Notes: Data from Long and Long (1971), and Johnston and Williamson (2008). Standard errors are clustered at the firm
level. ** denotes significant at 1 percent level, * denotes significance at 5 percent level.
a: Pool Member = 1 for Singer, Grover & Baker, and Wheeler & Wilson.
b: Increase in Union Army soldiers in 10,000s.
c: In millions of year 2000 dollars.
d: Licensee=1 for firms who purchased a license from Howe or the pool.

TABLE 3 – NEGATIVE BINOMIAL REGRESSIONS WITH CONTROLS FOR CIVIL WAR, GDP, AND POPULATION
DEPENDENT VARIABLE IS PATENTS PER FIRM AND YEAR (1850-1885)
(I) (II) (III) (IV) (V)
Pool -0.632 -0.261 -0.255 -0.148 0.028
(0.423) (0.242) (0.239) (0.267) (0.228)
Pool Member
a 1.918** 1.886** 1.887** 1.586** 1.551**
(0.308) (0.300) (0.300) (0.273) (0.282)
Pool * Pool Member -0.720* -0.700** -0.701** -0.836** -0.806**
(0.279) (0.268) (0.268) (0.282) (0.299)
Log Age 0.037 0.039 0.039 0.170 0.182
(0.117) (0.117) (0.117) (0.126) (0.123)
Union Army
b -0.044* -0.043* -0.042* -0.043* -0.043*
(0.022) (0.021) (0.021) (0.022) (0.022)
Union Army (-1) 0.003 0.001 0.003 0.002 -0.000
(0.011) (0.010) (0.012) (0.012) (0.012)
Union Army * Member 0.050* 0.047* 0.046* 0.047* 0.047*
(0.021) (0.021) (0.021) (0.021) (0.021)
Union Army (-1) * Member -0.005 -0.004 -0.005 -0.005 -0.006
(0.005) (0.005) (0.005) (0.005) (0.005)
Real GDP
c 11.941 16.353* 16.453* 14.264* 13.555
(8.169) (7.733) (7.755) (7.168) (7.289)
Log Population 12.786 21.278* 21.055* 16.136
(15.038) (10.059) (9.931) (10.448)
Licensee
d 0.465 0.439 0.446
(0.304) (0.300) (0.294)
Union Army * Licensee 0.006 0.002
(0.009) (0.005)
Union Army (-1)* Licensee -0.005
(0.008)
Member Fee -0.068
(0.054)
Non-member Fee 0.077
(0.060)
Linear Time Trend -0.337 -0.575* -0.570* -0.446 -0.050
(0.394) (0.254) (0.251) (0.260) (0.037)
Constant -131.394 -217.008* -214.759* -164.910 -1.932**
(151.543) (101.543) (100.257) (105.470) (0.402)
Observations 1051 1051 1051 1051 1051
R-squared 0.15 0.15 0.15 0.15 0.15
Notes: Data from Long and Long (1971), and Johnston and Williamson (2008). Standard errors are clustered at the
firm level. ** denotes significant at 1 percent level, * denotes significance at 5 percent level.
a: Pool Member = 1 for Singer, Grover & Baker, and Wheeler & Wilson.
b: Increase in Union Army soldiers in 10,000s.
c: In millions of year 2000 dollars.
d: Licensee=1 for firms who purchased a license from Howe or the pool.

TABLE 4 – POISSON REGRESSIONS WITH CONTROLS FOR CIVIL WAR, GDP, AND POPULATION
DEPENDENT VARIABLE IS PATENTS PER YEAR (1850-1885)
(I) (II) (III) (IV) (V)
Pool -0.591 -0.218 -0.214 -0.153 0.010
(0.360) (0.222) (0.218) (0.234) (0.199)
Member
a 1.854** 1.816** 1.817** 1.533** 1.517**
(0.308) (0.308) (0.308) (0.287) (0.277)
Pool * Member -0.776** -0.747** -0.749** -0.845** -0.831**
(0.265) (0.269) (0.269) (0.270) (0.267)
Log Age 0.104 0.111 0.111 0.229 0.240
(0.114) (0.113) (0.114) (0.131) (0.126)
Union Army
b -0.045 -0.044 -0.044 -0.045 -0.045
(0.025) (0.025) (0.023) (0.024) (0.024)
Union Army (-1) 0.004 0.002 0.002 0.001 -0.001
(0.009) (0.009) (0.011) (0.011) (0.011)
Union Army * Member 0.052* 0.050* 0.049* 0.051* 0.051*
(0.024) (0.023) (0.022) (0.023) (0.023)
Union Army (-1) * Member -0.005 -0.005 -0.005 -0.005 -0.006
(0.005) (0.005) (0.005) (0.005) (0.005)
Real GDP
c 11.581 15.772* 15.811* 14.542* 13.488
(7.724) (7.132) (7.166) (6.851) (6.981)
Log Population 8.541 16.994 16.855 14.625
(12.577) (9.503) (9.331) (9.803)
Licensee
d 0.410 0.383 0.386
(0.321) (0.317) (0.314)
Army * Licensee 0.005 0.001
(0.008) (0.005)
Army (-1) * Licensee -0.004
(0.008)
Member Fee -0.070
(0.048)
Non-member Fee 0.081
(0.050)
Linear Time Trend -0.231 -0.468 -0.465 -0.411 -0.049
(0.338) (0.244) (0.239) (0.249) (0.037)
Constant -88.629 -173.812 -172.404 -149.730 -2.032**
(126.738) (95.883) (94.140) (98.934) (0.344)
Observations 1051 1051 1051 1051 1051
R-squared 0.15 0.15 0.15 0.15 0.15
Notes: Data from Long and Long (1971), and Johnston and Williamson (2008). Standard errors are clustered at the firm
level. ** denotes significant at 1 percent level, * denotes significance at 5 percent level.
a: Pool Member = 1 for Singer, Grover & Baker, and Wheeler & Wilson.
b: Increase in Union Army soldiers in 10,000s.
c: In millions of year 2000 dollars.
d: Licensee=1 for firms who purchased a license from Howe or the pool.

FIGURE 1 – SEWING MACHINE PATENTS: POOL MEMBERS
Notes: Data from Knight (1876) and www.patents.google.com include any patent that was either granted or
assigned to a pool member when it was granted.
FIGURE 2 – U.S. PATENTS FOR SEWING MACHINES
Notes: Patent grants as reported in the Annual Reports of the United States Patent Office and Knight (1876).
The solid line plots a fourth-order polynomial trend.

FIGURE 3 - SHARE OF SEWING MACHINE PATENTS AMONG ALL U.S. PATENTS
Notes: Patent grants as reported in the Annual Reports of the United States Patent Office. The solid line plots a
fourth-order polynomial trend. Series excludes patents granted for tables and stands.
FIGURE 4 – THE CIVIL WAR: UNION AND CONFEDERATE ARMY SOLDIERS
Notes: Total number of soldiers, including deserters wearing their uniform (Long and Long, 1971).
FIGURE 5 - STITCHES PER MINUTE
Notes: Data from the Scientific American (1846-1869), exhibition catalogues, such as the “United States
Commissioners Report to the Universal Exposition in Paris,” “The Report of the Twenty-seventh Exhibition of
American Manufactures, Held in the City of Philadelphia,” ads in contemporary trade publications, including
“The Textile American;” and historical industry analysis, such as
Uniting the Tailors: Trade Unionism
amongst the Tailoring Workers of London and Leeds, 1870-1939
. A complete list of references for each
observation is available upon request. The solid line plots a fourth-order polynomial trend.
FIGURE 6 – LITIGATION: POOL MEMBERS
Notes: Data from Westlaw Pre-1945 Federal Courts Database and Robinson (1890).
FIGURE 7 – LITIGATION: DEFENDANTS IN SEWING-MACHINE CASES
Notes: Data from the Westlaw Pre-1945 Federal Courts Database and Robinson (1890).
APPENDIX A – POOL PATENTS
Number Issue Date Inventor Innovation
4,750 Sept. 10, 1846 E. Howe, Jr. Eye-pointed needle and lock-stitch.
6,099 Feb. 6, 1849 C. Morey & J.
Johnson (re-issued to
Singer and Clark)
Barbed needle.
6,439 May 8, 1849 J. Bachelder
(owned by Singer)
Continuous feeding device, horizontal table.
7,659 Sept. 24, 1850 J. Bachelder
(owned by Singer)
Vibrating loop chain-stitch.
7,776 Nov. 12, 1850 A. B. Wilson Lock-stitch using vibratory shuttle.
8,294 Aug. 12, 1851 I. M. Singer Feed-wheel, thread controller.
12,116 Dec. 19, 1854 A. B. Wilson Four-motion feed.
12,233 Jan. 16, 1855 Conant Rotary cloth-feeder, two needle chain-stitch.
16,030 Nov. 4, 1856 I. M. Singer Rotary lock-stitch improvements.
Notes: Data from the Singer Papers, Box 225, Knight (1876), and Depew (1968).  

 

 

 

As reproduction of historical newspaper articles and/or historical sources and/or historical artifacts, this works may contain errors of spelling and/or missing words and/or missing pages and/or poor pictures, etc.