The Acquisition and Commercialization of Inventions in American Manufacturing: Incidence and impact

The survey I will cover was funded by the National Science Foundation and the Kauffman Foundation. This is the first time it is being presented in Japan. It is still a work in progress.

It begins by discussing the importance of open innovation and markets for technology in an innovation economy. For U.S. firms, we still don't know the extent to which outside sources are a source of invention, which firms then commercialize. Previous specific studies examined the importance of universities or customers, for example, but a broad overview of the ecology of innovation has not been developed. There is also much we don't know about the impact of externally supplied inventions on the innovation rate and the relative importance, value, and impact of different sources.

The paper accompanying the survey provides broad evidence on the extent to which innovations introduced by U.S. manufacturing firms rely on external sources for their inventions and provides insight into what we refer to as the "division of innovative labor" for the United States. The distinction we make between invention (creation of new technology) and innovation (practical implementation; e.g., in the market) is important. We estimate the importance of these external sources to innovative performance and their incidence, value, and impact on innovation.

We hope to better understand innovation and expand our knowledge of its fundamental determinants resulting in a model of its drivers, so we have a base model of what are the drivers of innovation. There's also the social welfare and policy concern, well-established in the literature, regarding the importance of gains from trade in the economies of specialization. What may be more crucial is understanding the ecology of firms that are available to both generate and commercialize inventions. It is critical for us to understand the ecology of firms that are able to both generate and commercialize inventions.

The data were taken from a survey of product innovation for firms in manufacturing and selected service industries (mainly manufacturing). We focused on firms which commercialize products in order to find those that have implemented innovation. We also asked for the respondents first to focus on a specific line of business in answering all of the questions to allow categorization by industry. For the period 2007-2009, we asked whether the firm had any innovations, and, if so, what was the most significant one? If they innovated, we asked where the innovation came from—did it come from suppliers, customers, universities, etc.—and how they acquired it—from licensing, cooperative research, mergers and acquisitions, and so on.

The sample came from Dun & Bradstreet firm lists. We stratify by industry, firm size, and startup/non-startup status. We started with a list of about 22,000 firms with oversampling of larger firms, startup firms, and firms in the most innovative industries. We focused on surveying personnel who are most likely to know about their products.

We received an approximately 30% response rate—over 6,000 responses. Firms not involved in manufacturing and very small firms (fewer than 10 employees) were eliminated, leaving about 5,000 cases. We then adjusted the sampling weights to account for the non-response bias and the sampling strategy. A 30% response rate is modest and there can be concerns about response bias, so we compared the sample with the population for the data in our possession, and, generally, the sample is representative of the population.

In 2009, we asked respondents whether they had earned revenue from any new (meaning new to the respondent) or significantly improved goods or services introduced in their industries since 2007. If the answer was yes, we asked whether any of these new or improved goods or services were patented. We also asked whether the respondent was the first to introduce the relevant new or improved good or service. If the respondent answered yes to either of those questions, that respondent was considered to be a "new to the market innovator."

About 43% of respondents stated that they had earned revenue from a product that was new to the firm during the previous three years. About 18% said that they introduced a product that was new to the relevant market during the same period. The Business R&D and Innovation Survey (BRDIS), a census introduced at about the same time, recorded a "new to the firm" percentage of about 22%, but nothing for "new to the market." A German survey from about the same time gave responses of 49% and 22% for new-to-firm and new-to-market, respectively. As these are somewhat subjective, we also compared whether respondents' answers correlated with the research and development (R&D)-intensiveness of their industries, patenting rates in the said industries, etc. In general, these correlated at a ratio of 0.7 or above. The only correlation below this was BRDIS R&D-intensity (r=0.6).

An average of 20% of respondents' total sales derived from new or significantly improved products. About 40% of new-to-market innovations were patented. About 40% of respondents answered that they had brought a new-to-the-firm product to market. Of these, about 40% stated that this product was also new to the market. This data raise many questions. The innovation rate varies substantially across industries, but the imitation rate seems much more stable. We also see variation in patent propensity—innovators are much more likely to have patented than are imitators. About 4% of respondents answered that they had developed technology for another firm, but they themselves did not introduce any innovation.

With regard to firm size—large firms (more than 1,000 employees), medium-sized firms (100 to 1,000 employees), and small firms (more than 10 but less than 100 employees)—rates of innovation are higher for large firms. The share of sales coming from innovation tends to be very skewed; the most significant innovations tend to account for the bulk of sales for all innovations. It must be said that "small firms" does not necessarily mean startups. About 180 startups were included in our data, but many long-established small businesses also participated.

We asked whether suppliers, customers, other firms in the industry, consultants, commercial labs or service providers, independent inventors, and universities or government labs generated the innovation (i.e., created the overall design, developed the prototype, or conceptualized the technology). About half of the respondents said at least one of these sources was the source of their innovation. Customers were the most common answer followed by suppliers. Technology specialists (consultants, commercial labs or service providers, independent inventors, universities or government labs) accounted for about 18% of innovations.

Dependence on external sources is relatively high and seems fairly stable across both industries and size classes. Customers are generally the most pervasive source. More R&D-intensive industries tend to cite suppliers and customers less and cite universities more. When the interviewed company provides intermediate goods, customers are more likely sources of innovation.

Startups constitute only about 2.5% of our sample, but are reported to be a source of innovation in 14% of the cases. They are a disproportionate source of innovations commercialized by other organizations. Large firms especially are likely to receive innovations from suppliers. Small firms are more likely to use independent inventors as a source.

Mergers, joint ventures, cooperative research, licensing, service contracts, and informal sources are all possible innovation acquisition channels. Mergers, licenses, and service contracts are market channels. Cooperation in research is an important channel—innovation is often acquired as jointly-developed technology. Licensing, mergers, and service contracts represent just over one-third. Only about 16% involved exclusively market-based channels (not jointly developed, etc.). Markets for technology were responsible for a minority of the transfers. An exception is pharmaceuticals companies, which depend heavily on market channels, especially licenses and mergers.

Small firms tend toward informal channels, whereas large firms favor licensing and mergers and acquisitions. If the source was a customer, it's especially likely to have been either informal or cooperative. If the source was one of the technology specialists, market channels are more common. We also compared the cost and value of inventions obtained from customers and from specialists.

To determine value, we asked where inventions originated and how valuable they are (share of a business unit's sales attributable to the innovation). We also asked whether a company needed to invest in new personnel, new equipment, or new distribution channels to commercialize the innovation and whether the company patented it. These are all indicators of greater value. We used regression models to compare internal innovations to each of the external sources with controls for selection.

Customers are less valuable than internal innovation while specialists are more valuable. Specialist-sourced innovations are more likely to be patented than customer-sourced innovations.

We also conducted a counterfactual exercise to estimate the overall impact of external innovation sources. What if innovations were unavailable from any particular source, for example, suppliers or customers, or from any external sources at all? We used a multinomial logit estimation framework to see how the sources are distributed across the set. If customers were unavailable, you would have an 18% reduction in the overall innovation rate. If specialists were unavailable, this would be 11%. All told, there would be about a 43% reduction in the approximately 18% rate of innovation in the absence of external sources.

About half of the most significant new-to-market innovations originated from an external source, the most prevalent of which are customers and suppliers. Technology specialists tend to be more important for high-tech industries and also for smaller innovators. The acquisition channels often include collaboration. Market-based channels are much less likely to be used overall, but are more likely to be used in high-tech industries. While customers are the most likely source, they are also of lower value than specialists as sources. They are also lower cost. Reduced access to external sources significantly lowers the overall innovation rate.

GOTO Akira
Professor Walsh's paper showed for the first time through very elaborate surveys and econometrics that the separation between invention and innovation is far greater than we had imagined. In other words, open innovation is very widespread in the U.S. manufacturing industry. Under the old model, invention and innovation are done within the same large company, while, under the new model, they are done separately. The presentation also claims that the new model can generate more innovation because of the labor specialization permitted by vertical separation. Specialist firms create inventions and pass them on to innovating firms through technology markets. Invention by specialists is high-cost and high-quality while invention by customers is the opposite.

It has also been said that outsourcing of research was very common in the U.S. manufacturing industry in the late 19th and early 20th centuries. The new model or the recent vertical separation trend then may actually be a return to the old model.

There are two explanations for the birth of the new model. First, information and communications technology (ICT) and biotechnology demanded this kind of innovation system. Second, institutional change (including strong patent protection and weaker anti-trust enforcement, etc.) in the United States has also led to this kind of arrangement and, in fact, may be responsible for the birth of this new model.

There are two reasons—one theoretical and one empirical—to doubt the claim that the new model is more productive in terms of creating innovation. We all know that technology and information are very difficult commodities to buy and sell across the market as they involve large transaction costs. A market for technology cannot be expected to be perfectly competitive and cannot work efficiently. Also, empirically, the old model was not so bad. For example, in the age of the old model, where big firms played a major role in the pharmaceutical industry, quite a few blockbuster drugs were introduced. Some scholars have claimed that the introduction of new drugs has not increased since the introduction of the new model. I wonder whether the new model or vertical separation of invention and innovation is really a good idea.

John P. WALSH
Thank you very much for those very insightful comments. On whether the new model is new, I'm very sympathetic to your argument, which is that we don't know whether it is or if this 49% of externally sourced innovation is higher or lower than it would have been 20 years ago. There's good reason to think there was always a lot of external acquisition, and there might have been more in such prior periods. We should be careful about making claims that somehow this is novel, but because there has been much discussion that this structure might be new and important, what we wanted to do was acquire broad data on just how common this is. We know there is a biotech industry and that pharmaceutical companies work closely with it, and we know that small IT firms become technology sources for large integrated IT firms, particularly in software. The more general point is that there is also a gradient between the acquisition of a completed technology put into the market versus simply hearing about an idea and then developing it into a new technology in-house. It's likely that there are some gray areas there.

Assuming there has been an increase: why? One argument is that some industries by nature are compatible with markets for technology. I think institutions probably play a key role. One of the important findings is that many are not patent-mediated. More generally, they're not market-mediated. This point about transaction costs and the problems of market prices suggests that to the extent that this division of innovative labor is viewed as a market for technology, it has these concerns. Transaction costs are often substantial, and the price mechanism may not work very well. In general, this division of innovative labor is less market-like. The mechanism seems to be much more in the line of relational transfers between parties with existing ties and ongoing relations.

Is this any better than integrated innovation? The argument is that it is complementary to internal innovation or non-innovating—i.e., the vast majority of firms don't innovate. Should I make this myself or should I let someone else make it? There's probably good reason to have both, but it is an open question about whether outsourcing alone is a good system, and the same is true of exclusively having internal innovation. Having a mix is probably better, but there are many varieties, and it's not at all clear what mix would work best.

Q1. If a company patents at least part of an innovation, you call it "new-to-market," but in some industries, such as electronics, thousands of patents are involved in a product. That's not very new to the market. There are very large industrial differences. How do you justify this definition? Second, you talked about patents in terms of licensing as a tool for the transmission of new information and new inventions. But, on the other hand, patents or other intellectual property rights can be a deterrent to the diffusion of inventions. Does your study or your survey give any hints or ideas on the extent of that kind of negative effect of patents?

John P. WALSH
We have a composite measure to determine whether or not a product was new to the market. If we limit it simply to whether it was the first in the market, the results will be largely the same, so the model is not very sensitive to whether or not we use the patent question. The extent to which using the composite measure changes the results may vary by industry so we may want to simplify it for the next round of revisions.

Do the patents deter rather than facilitate transfers? Very few of these are patented—around one-fourth—meaning that the source had a patent, which is what you are referring to. High-tech inventions and those sourced from technology specialists were particularly likely to have been patented, which is consistent with the idea that a patent is a vehicle to allow the transfer and receive some return from the invention. We also have cooperative transfer and other transfers which might be deterred by the existence of patents. Patents are not the bulk of the transfers so it is clear that they are always necessary for transfers.

Q2. I was very interested in Mr. Goto's comments. Institutional changes such as patent policy, anti-trust policy, and higher education were discussed. Do either of you believe that other possible factors, institutional changes, or government policies could have affected the formation of the so-called new model?

John P. WALSH
In terms of another actor, for example, although the overall share of startups in this division of innovative labor is not that high, it is disproportionately high. This suggests that an infrastructure that encourages startups may be an important mechanism, in terms of venture capital and other aspects that make it easier to start businesses. The biotech industry is the classic example. Encouraging the establishment of firms that utilize technology specialists would be an additional policy component.

GOTO Akira
One possible policy would be a law to protect trade secrets. In Japan, strengthening the Unfair Competition Prevention Law would have an impact. Another possible institutional change would be a policy establishing a government research institute, which could have a major impact on the innovation system. For instance, the U.S. National Institutes of Health (NIH) has a huge budget—over $30 billion. The total budget for science and technology in Japan is about 3.6 trillion yen. It plays a major role in promoting basic science related to health and also it makes it possible to educate young and talented scientists.

Q3. The importance of startups, government institutions, and education was mentioned. These are all external invention sources. Are there any policy measures conducive to better non-market transfer mechanisms?

John P. WALSH
Weakening trade secret law would result in more non-market transactions. However, changes related to, for example, anti-trust and eliminating prohibitions on cooperative R&D might be another possibility. There was a change in U.S. patent law that basically facilitated cooperative research between universities and industry by reducing the probability that prior art in the course of internal collaborations within this collaboration team would be generated accidentally. Basically, the modal non-market transaction looks like either a firm's customer or supplier developing a technology and saying, "We would like you to implement this." The supplier might offer very generous terms for the transfer. As an example, the supplier might be very generous in the terms of transfer. An aluminum company may introduce a new lining for cans and teach canning companies how to make cans with the new lining. The canning companies will buy more aluminum, thus the aluminum company does not need to make money on the technology as it will make it on selling the aluminum. I'm not sure of specific policy mechanisms that would encourage this, but I think the main point is to not do anything obstructive.

GOTO Akira
One possibility is to liberalize labor law to allow engineers to move from one company to another more easily. In most cases, inventions and ideas are embodied by engineers, and mobility allows their ideas to move with them. This would promote the flow of inventions and ideas between companies.

John P. WALSH
As an example of that is, due to the high cost of health insurance in the United States, which is tied to your employer, if you had an existing illness and tried to move to another company, the new company's insurance would not cover any expenses related to that pre-existing condition. That limits labor mobility. One component in the new healthcare law in the United States tries to eliminate those mobility costs. That's one example of a policy change to make it easier.

Q4. Innovation and invention are key to the revival and further development of the Japanese economy. In this regard, why are all of the survey results focused on the United States, one of the most advanced countries in the world in terms of innovation and invention? What if we looked at China? Wouldn't it be different? Also, we basically assume innovation and invention are done by people. Artificial intelligence (AI) is developing very rapidly. Innovations by an AI would not be attributable to people but to a network of computers or to the artificial intelligence. What would be the impact and implications of that?

John P. WALSH
How this varies across countries is a very interesting topic. Particularly how this would be different in developing countries is also important, and how this would differ in China which is investing in a lot of R&D is especially interesting. I'm quite sure it would be different, but I am not sure how. We need to learn more. On artificial intelligence and robot innovation, I met a mathematician who wrote software to create proofs, and the computer program had a couple of publications in journals. Robot scientists do exist. In the future, it's conceivable that they would be technology specialists which might specialize in offering solutions and possibly be incorporated into the regular activity of a larger firm.