Immigration and high-tech startups: not correlated, after all?

Some colleagues have just published a new analysis of high-tech entrepreneurs, and the results are  surprising. There has been a lot of talk  lately that high tech firms are disproportionally founded by immigrants.  But according to my interpretation of this new, and more carefully done, study, that’s  not true.  Here’s my summary of their results:

Only about 3 percent  of the founders of high-impact, high-tech companies are foreigners (about 60 out of 2034).   97 percent are US citizens, and specifically 87 percent are US-born, while 10 percent are naturalized US citizens.    Furthermore, most foreign-born founders lived in the US for decades.  These founders are statistically very similar to the average US population in terms of birth and immigration status.

What’s the evidence? The study is by David M. Hart, Zoltan J. Acs, and Spencer L. Tracy, Jr.High-tech Immigrant Entrepreneurship in the United States A summary of their report is at http://www.sba.gov/advo/research/rs349.pdf

Here’s the odd part: the authors draw a rather different conclusion from their data than I do. Specifically their lead is “The central finding of the study is that about 16% of the nationally representative sample of high-impact, high-tech businesses that we surveyed count at least one foreign-born person among their founding team.” And indeed, their data does say this. But they failed to make a comparison to the US population in general. And classifying as “immigrant founded” a firm with any foreign-born founders is inherently biasing the results toward a higher number. They report that “of the 205 Immigrant Founded Companies in the sample, more than half [ie. at least 8% of the firms in the sample] were founded only by foreign-born entrepreneurs,” which seems to me more informative than the 16% number.

For what it’s worth, I’m very sympathetic to the view that immigrants have a positive effect on technology and entrepreneurship. My nationalistic perspective is that we (the US) should educate in our universities the smartest kids we can get our hands on worldwide , and then keep them in the US.  We’ve always been a nation of immigrants, and there is no reason to change. Of course, there are many nuances and ramifications of this debate.

Here’s my comparison data on the general US population, based on latest-available information from US Census.  In 2003  about  11.7 percent of the population was foreign-born.    (http://www.census.gov/prod/2004pubs/p20-551.pdf)  Of these, only about 40 percent are naturalized US citizens (http://www.census.gov/prod/2003pubs/c2kbr-34.pdf). But corrected for the same average age (or residence time in the US), it looks like very roughly 70 percent of immigrants are  naturalized.  And, if we adjusted for region of residence, since startups and immigrants both disproportionally reside in some regions, the fraction of immigrants in their sample would probably be even lower than the general population in the same region.

On the other hand, it might be more relevant to look at the status of company founders in the year of founding, rather than today. Presumably fewer of them were US citizens at the time (although the percentage foreign born would be the same). But a similar correction would apply to the immigrant population of the US as a whole, so it’s still not clear that the two would differ meaningfully.

Finally, this research looks well-done to me. It’s only their spin on the executive summary  that I disagree with. They seem to have done a much more careful job than previous analysis on this topic, and if you read their report carefully they don’t overstate their results.

[A good comment on this research is at http://www.usnews.com/blogs/capital-commerce/2009/07/16/immigrants-a-driving-force-behind-innovative-firms.html#3323010 . The commenter has delved into the report and extracted some numbers I didn’t notice, that support the view that this research is NOT pro-immigration, at least not directly.]

Advertisements

3 thoughts on “Immigration and high-tech startups: not correlated, after all?

  1. Roger, it’s not that simple. There are many issues which make the comparisons you cite from the Hart and Acs paper invalid. Here are a few:

    1. DEFINITION OF “HIGH-TECH”
    The researchers created their own definition by taking an antiquated BLS definition, subtracting the largest industry group in the BLS definition and including their own SIC codes. The resulting definition of “high-tech”
    includes — cigarette, crude petroleum and natural gas, pulp mills, soap, paints, agricultural chemicals, reclaimed rubber, photographic equipment and supplies, etc. What do these manufacturing industries have to do with what we, today, call “tech”? Plus I would really be surprised if you or any other academic would consider the mining, paper or chemical industries to even be “high-impact” let alone “high-tech”? Maybe these industries did make an impact on the economy at the turn of the century and there are some old definitions which include these, but let’s get real here.

    2. DATAFRAME USED/AGE OF COMPANIES SURVEYED
    The dataset they used is rarely used by academic researchers for understandable reasons. It is “Corporate Research Board’s American Corporate Statistical Library” and is derived from Dunn and Bradstreet data. The researchers selected companies which doubled sales/employment from 2002-2006 from this dataset.

    The problem is that this includes companies which are decades old and largely irrelevant to the policy arguments which the report seeks to make.
    And it turns out, 70% of the companies surveyed were over 10 years old and 12% were more than 30 years old. The world is changing rapidly and we need to look at what is happening today, not decades ago.

    I doubt that anyone would agree that a 30-year old cigarette manufacturer or mining company is relevant to today’s debate about skilled tech immigrants.

    3. DEFINITION OF FOUNDER/METHODOLOGY
    The researchers define a company founder as “the person or people who owned part of the firm when it first began to cover all salaries and wages”. This would mean that any seed investor including friends and family and venture capitalists were “founders”. They called companies and said they wanted to speak to anyone “who knows about the founding history of the company”. This person was then asked about the company’s R&D labs, patents filed, sponsored research, etc. and then the citizenship, ownership, education of the ” 5 most important founders”., etc.

    Anyone who knows about tech companies knows that founders don’t stay this long and corporate memory is very short. How could the research team possibly have obtained information such as how the founding team came together, education level and history of each founder, year each founder came to the U.S. and obtained citizenship — from 30 year old companies?
    Even in 10 year old companies, hardly anyone knows this. I personally founded 2 technology companies and didn’t even know all this about my co-founders the day I founded the companies!

    4. RESPONSE-RATE STATISTICS
    The report says that of the total population of 24,000 companies in their original dataset, they surveyed 2,668. It says “1,415 provided completed responses, giving us a response rate of 53%”. This is misleading — the 53% was a “survey completion rate”, not a “response rate”. I would be really interested in learning what percentage of the total population of companies contacted actually took the survey. I believe that in a survey like this, a high-response rate is critical to the credibility of the resulting analysis.

    5. VALUING THE CONTRIBUTION OF IMMIGRANTS
    In a high-technology startup, there are only 2 people who are really critical to the success of the venture — the CEO and the CTO (or lead technologist). They are equally important and can be the same person (like Bill Gates), but one can’t succeed without the other. (Key product developers are also important, but these are not always company founders).

    This report counts anyone who provided seed funding as a founder and gives them equal importance. Then it compares native founded companies with immigrant founded. The native founded company has to have all natives, but the immigrant founded company can have just one foreign-born and the rest can be natives. How is this a meaningful comparison? So what if the accountant or lawyer was an immigrant? How can you make policy recommendations based on this comparison?

    As a result of these flaws, the analysis produces some really incredible findings:

    1. Have a look at Figure 1 in the report. If you exclude Mining, Paper, Chemicals, Machinery, Transportation Equipment, and look at the industries which could realistically be considered “high-technology” — Electronics, Instruments, Communications Equipment, Business Services, and Engineering Services, you see an amazing trend. In every one of these industries, immigrant founded companies greatly outnumber the native founded companies! How can this be?? It seems the report finds that the vast majority of “high-impact” “real high-tech” companies created over the last few decades in the U.S. have immigrant founders.

    2. They found that Indians were the largest of the immigrant founding groups and as you noted below, the researchers claimed that most foreign-born founders lived in the U.S. for decades. Yet if you look at census data (http://www.census.gov/prod/cen2000/doc/sf4.pdf), you see that as of 2000, 54% of the India-born in the U.S. came to the U.S. after 1990, and an additional 27.8% arrived from 1980-1990. Only 18.2% were here for “decades”.
    It is the same with many other Asian immigrant groups. So something isn’t right with the Hart report’s findings.

    3. They found that 30% of immigrant founded companies had women founders vs. 20.5% of native founded. This itself would be headline news if it was the reality.

    Roger, I am a fan of your research and I think that we agree on the importance of skilled immigration. Let’s continue to make our rightful arguments with facts and quality data.

    Regards,

    Vivek Wadhwa
    Visiting Scholar, School of Information, UC-Berkeley Associate Director, Center for Entrepreneurship and Research Commercialization, Duke University
    Senior Research Associate, Labor and Worklife Program, Harvard law School
    Columnist, BusinessWeek.com

  2. Dear colleagues:

    Vivek Wadhwa harshly criticized our recent report on this blog. (The report can be found at http://www.sba.gov/advo/research/rs349tot.pdf.) In what follows we respond in detail to his critique. For those who may not want to read this post in full, we summarize our response here. Most of Wadhwa’s criticisms should be ignored by scholars. He makes elementary mistakes of math and methodology. He frequently makes assertions that have no foundation. He exaggerates and demeans, ultimately impugning no one but himself.

    We believe our study provides a better estimate of high-tech immigrant entrepreneurship, as those terms are ordinarily used by scholars, than Wadhwa et al. 2007. However, there is room for the figures in the two studies to be reconciled, depending on one’s definitions of the key terms and what one thinks the relevant population of firms is. We would be more than happy to entertain a serious scholarly debate on this subject and look forward to engaging in one.

    We refer at several points in our response to the Kauffman Firm Survey (KFS). We couldn’t hope to match on our modest budget the methodological sophistication and level of effort that this multi-million dollar project achieved. But we reference it as the gold standard for survey research on entrepreneurship.

    DEFINITION OF HIGH-TECH. We use a standard definition of high-technology found in the economics literature, based primarily on the ratio of R&D employment to total employment and referencing BLS data. The KFS uses the same definition, although it defines three groups of SICs, high-technology, medium technology, and all other industries; our high-tech category basically combines KFS’s high- and medium-tech categories, as we state in the report. Wadhwa et al 2007 use the following definition: “the main work of the company is to use technology or engineering to design or manufacture products or services.”

    The only SICs in our high-tech category that are not in KFS’s high- and medium-tech categories are 303 (reclaimed rubber), 374 (railroads), 489 (communication services, n.e.c.), and 874 (management and public relations). 303 and 374 account for less than .05% of the firms in our population. 489 accounts for almost 1%; this group may include emerging ICT-based service sectors that fit conventional notions of high-tech. But the key point is that these 3 SICs are too small to affect the results that we report in any significant way.

    874 is a special case. As we describe in the report, it would have dominated our results had we left it in, accounting for more than a third of all firms in our population. We judged it better to omit this industry and tell the reader that we did so. Perhaps others might find it interesting to explore the role of foreign-born entrepreneurs in management and public relations at some future date.

    We report the distribution of immigrant-founded and native-founded firms by broad sector (manufacturing vs. services) in Table 3 and by two-digit SIC in Figure 1. In both comparisons, the patterns are similar between the two groups of firms. Wadhwa misunderstands Figure 1 in his comment. Each bar in the graph shows the share of IFCs or NFCs in that sector divided by the total number of IFCs or NFCs respectively. Because there are many more NFCs than IFCs in the sample, there is no sector in which IFCs out-number NFCs.

    USE OF DUN & BRADSTREET DATA. D&B is the best source from which a sample like ours could have been pulled. KFS used D&B for its sample. Indeed, Wadhwa et al. 2007 used D&B as well. Census and IRS data are undoubtedly richer sources, but they are only made available to researchers under strict confidentiality conditions, and it is not possible to identify individual firms for survey research purposes.

    AGE OF FIRMS. A prior study by two of the co-authors of this study (Acs, Parsons, and Tracy 2008) shows that a small fraction of firms account for the bulk of net job creation in the U.S. One surprise in that study was that the age range of these firms is wide. It seems that there are relatively old firms as well as young ones that for some reason or another “catch fire” and grow rapidly for at least several years. That growth is happening today, not thirty years ago, as Wadhwa implies.

    In any case, as Table 4 reports, there is no statistical relationship between age of firm and immigrant founding in our sample.

    DEFINITION OF FOUNDER. For the most part, we let our respondents identify the firm’s founders without a definitional prompt. However, if the occasional respondent asked for a definition, the one that our survey contractor supplied, which Wadhwa quotes in his post, is the one that is used in the Panel Study of Entrepreneurial Dynamics (PSED). Paul Reynolds, the PI for PSED, is one of the country’s leading authority on survey research related to entrepreneurship and has invested considerable effort in creating this definition.

    We point out in the report, as Wadhwa acknowledges in his comment, that CEOs and CTOs who were the subjects of his study are not necessarily founders. The Stanford Project on Emerging Companies, for instance, found that about half of founders were not employed in either of these roles. We thus come to a semantic question: who is an “entrepreneur” – the founders about whom we asked or the senior executives about whom Wadhwa et al 2007 asked? The vast majority of scholars equate “entrepreneur” with founder, although more encompassing definitions that include senior executives can be found. We believe that firm founders and senior executives perform different functions in society – both important, but different. They must be treated separately if we are going to understand firms and economies.

    RESPONSE RATE. We used the response rate calculator made available on-line at aapor.org by the American Association for Public Opinion Research (AAPOR). That calculator provides 14 alternative definitions. We use the one that we thought our colleagues and readers would find most intuitive: of those firms that our survey contractor actually reached, what proportion finished the survey? Technically, this is the “cooperation rate.” (All four definitions of the cooperation rate in the AAPOR calculator yield the same value, 53%, for our sample.) If we include all of the other dispositions of phone numbers called by our survey contractor, such as fax numbers, busy signals, non-working numbers, and the like, the rate drops to between 29% and 34%, depending on which definition (AAPOR 1-4) is used. We were surprised to get such a high response rate, however calculated. Survey researchers whom we consulted before we began the project told us to expect a response rate of 10%
    or even less. (One of the most intimidating things that we learned was that the KFS had made 375,000 phone calls to obtain 3,782 completed interviews.)

    But this discussion is something of a digression. The response rate does not determine whether a survey sample is representative. A representative sample is unbiased. We addressed bias by validating the sample against the population data that we had in hand from D&B. We also used the population data to weight the observations in the two regressions that we report. And we obtained a large enough sample to keep random error low. Of course, it is always possible that there is an unknown variable that makes the respondents different from the non-respondents. That is a hazard in all survey research. But we did what could reasonably be done to have confidence that our sample is representative: worked with a professional survey research organization, validated the data, and obtained a large sample. We certainly believe our data collection method is more reliable and valid than that described on p. 9 of Wadhwa et al. 2007.

    POLICY RECOMMENDATIONS. We don’t make policy recommendations in this report. We seek to add information that is relevant to the policy debate.

    INDIAN FOUNDERS. We report (Table 20) that 16% of the founders in our sample were born in India. Table 16 shows that 25% of the founders in our sample had been in the U.S. for 15 years or less, and another 25% had been here for 25 years or less. There’s certainly no mathematical problem here with the timing of Indian immigration to the U.S. We do think the issue of mapping the country of origin and timing of firm founding onto the changing demographics of immigration is a good issue for further study, and we have been working on that.

    FEMALE FOUNDERS. This issue is also worthy of further study. Wadhwa does not state a reason for doubting our findings but simply makes unfounded assertions.

    THE COMMUNITY OF SCHOLARS. We certainly agree that the growth in research on the role of the foreign born in the U.S. innovation system and in U.S. entrepreneurship is an excellent development. We look forward to participating in further discussions with those scholars, who, like us, take care in designing and carrying out research and who engage in serious and evidence-based debate.

    WADHWA’S RESTRAINT IN MAKING NEGATIVE COMMENTS ABOUT US TO THE PRESS. We are disappointed that Wadhwa did not discuss our work with reporters. We would have been happy to rebut his views, even though we understand that the rules of the game are different in the media than in the scholarly community. As the old saying goes, “any publicity is good publicity.”

  3. David, I am not sure if I am the Police or the Professor here, but I agree with our President that it is best to transcend rhetoric and focus on the
    substantive differences.

    You felt it necessary to launch personal attacks against me, but I stand by what I wrote. Yes I did misinterpret Figure 1, but everything else was
    correct. There are many definitions of “founder”, “hi-tech”, “high-impact” and “response-rate”. In every case, you have used definitions which will
    mislead the audience for which this Small Business Administration advocacy paper was written — small business people and policy makers. Plus in the
    paper itself, you tried to discredit all research which came before yours.

    Ultimately, future scholars will judge who was right.

    Regards,

    Vivek

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s