Difference between revisions of "Measuring High-Growth High-Technology Entrepreneurship Ecosystems"
Line 7: | Line 7: | ||
==Current Version== | ==Current Version== | ||
− | The current version. which | + | The current version. which lead to a 2nd R&R at Research Policy is: |
− | <pdf>File:MeasuringHGHTEntrepreneurshipEcosystemsV3-3.pdf</pdf> | + | <pdf>File:MeasuringHGHTEntrepreneurshipEcosystemsV3-3.pdf</pdf> |
==Files== | ==Files== |
Revision as of 12:56, 10 December 2020
Academic Paper | |
---|---|
Title | Measuring High-Growth High-Technology Entrepreneurship Ecosystems |
Author | Ed Egan |
Status | R and R |
© edegan.com, 2016 |
Contents
Current Version
The current version. which lead to a 2nd R&R at Research Policy is:
Files
The files are in:
E:\projects\MeasuringHGHTEcosystems /bulk/vcdb4
Notice
This paper was broken into two:
- Measuring HGHT Entrepreneurship Ecosystems: This now contains the definitions, measures, and example. It is an informal, by-example theory paper.
- Determinants of Future Investment in U.S. Startup Cities: The empirical analysis of ESOs is now in this paper!
2nd R&R
Note that there Reviewer 2 never returned any comments.
Summarizing reviewer 1's comment (reading between the lines):
- The writing is currently pretty good: it is well done... the paper is polished... very nicely done.
- It works as a whole. The reviewer didn't want anything cut: It is a collection of case studies and definitions, and I don't have ... major comments.
Reviewer 3's comments are more problematic. As is often the case, I wondered whether the reviewer actually read the paper:
- The paper advances seven new measures, not 15 as the reviewer claims
- Policy cartels are introduced in section 4.1 (out of 5), and aren't the main focus of the paper per se
- I never use the entire battery of measures -- different measures are applicable in different contexts
- A key point of the paper is to stop organizations from self-selecting into measures that they do well on
- All bar two sentences of the 'substantive material' in the review (i.e., from "First..." to the end) don't mention anything to do with the paper!
At this point in my career, I'm uniquely placed to give the response that every academic really wants to give to the 'self-aggrandizing idiot' that somehow always ends up controlling the fate of our hard work, namely:
- "Dear Reviewer. After carefully considering your comments, I would like the offer the following response: I find your suggestions for my beautiful paper to puerile/irrelevant/narcissistic/useless/stupid/all-of-the-above (delete as appropriate), so I'm going to ignore them and, by extension, you. Up yours."
However, it does seem that some of reviewer 3's comments would lead to improvements in the paper. These are:
- Being clearer that standardized measures aren't a panacea
- A better discussion of gaming and incentives
- Better framing in the front end of why these measures make sense (likely using the points from RFP for the special issue)
I would need to address Reviewer 3's comments point-by-point, so here they are (re-ordered):
- Discuss when "standardized measurement" improves outcomes, when it does, and perhaps how.
- Discuss systematic bias due to gaming metrics incentivized by rewards attached to the measurement (old and new). Specifically, note that this behavioral change might have nothing to do with the underlying phenomenon of interest. (Think "Rewarding A While Hoping For B", etc., multitask, etc.)
- Justify the choices in the reduction of the measurement space. Measurement is often reductive: What is left out? Is it important? Discuss the difference between the conceptual phenomenon and how it is operationalized.
- Consider possible downsides of each individual metric
- Consider possible downsides in using the entire battery of measures
- Test alternatives to this measurement approach
- Test the effects of using a framework on various outcomes.
Of these, 1,2, and 3 are reasonable suggestions and could be addressed. Point 4 is more problematic but probably possible in some sense. Point 5 could be interpreted as the downside of the framework as a whole (see below) and then could be possible. Points 6 and 7 might be excused by explaining that the editors and I have agreed that this won't be a testing paper. Nevertheless, the paper will show examples of not using the measurement framework through-out.
Also for reference, here are the measures from the paper:
- Measure 1 (Startup Ranking)
- Measure 2 (Apportioned investment and exit value)
- Measure 3 (MOOMI ratio)
- Measure 4 (Pipeline)
- Measure 5 (Raise Rate)
- Measure 6 (Repeat VC)
- Measure 7 (ESO Expertise)
Measure 1 is exclusive to all others. Measure 2 underpins 3, but 2 & 3 stand-alone. Measures 4 and 5 go hand in hand, and can be refined by 6. However, 7 exists a proxy for 5 and 6, because they can be hard to calculate. So, the reviewers request to "consider possible downsides in using the entire battery of measures" could only mean to consider the downside of the framework as a whole, because it isn't possible to use "the entire battery of measures" together.
In my letter to the reviewer, I can explain that this paper is for a special issue and that the editors and I have agreed that this should not be and an empirical paper and that should not contain regressions or other tests. That might tamp down their vitriol somewhat. Whether I deliver back a "major revision" is in the eye of the beholder, and there's no need to draw attention to this demand.
RP Constraints
Although no-one has asked me to, I need to cut down the length of the paper. RP asks for a maximum word count of 8,000-10,000, though does allow exceptions. Using pdftotext and wc, or doing a count in the tex file, it looks like I'm up near 17k. However, both approaches drastically inflate the total with table entries, TeX commands, etc. I expect the current true count is closer to 15k. I need to get it down to at most 12k.
A different way of coming at the problem is to say that I'm using 12pt font with 1" margins and one-half line spacing (as implemented using \onehalfspacing in LaTeX, which isn't the same as 1.5 spacing in word). AER restricts to 40 pages using this format[1], and Management Science restricts to 32 pages[2]. With averages of perhaps 15 words per line and 32 lines per page, this would give just shy of 500 words per page, making 10k words, 20 pages of dense text, and perhaps 30 pages with figures, tables, quotes, headings, definitions, etc.
Because the last round was also an R&R, I presumably met all the other constraints.
Special Issue Questions
The special issue posed a series of questions (the full list is below) that I might rephrase to statements as follows:
- Startups (venture-backed and pre-venture backed) are the fundamental constructs, so they are what should be measured, and cities (census places geographically define ecosystems)
- Consistent measurement is key!
- The pipeline framework lays out the key relationships as it maps the mechanisms by which ecosystems get established, mature, decline, or get renewed.
- Venture capital measures add quality dimensions. Using the pipeline framework on top of this quality measure gives more finely grained evaluations of the effectiveness of policy instruments
- Venture capital has timing issues but exits are noisiers and longer term (short vs. long term measures)
- What do existing rankings for ecosystems measure?
- Beyond direct pipeline components, we should think about universities, corporations, and other participants.
- The measures should be able to assess the question: To what extent is government policy accelerating or inhibiting the progress of entrepreneurial ecosystems?
Reviewers Comments
The raw reviewer's comments are as follows:
Reviewer 1's Comments
First of all, my sincere apologies to you and the authors for my very late report on the manuscript. Please excuse me for the delay.
The authors study measures of high-growth, high-tech entrepreneurship activity across the U.S. and provide concrete examples of how municipalities can use these to assess various policy initiatives. The authors claim that policy interventions at the municipal level have a significant impact on pre-venture startups, and that this has been missing from the extant academic literature.
Regarding my take on the paper: it is well done and given that it has already been revised once, I don't have many more major comments, but I do have one thing I would like to bring to your attention -- although the paper is polished, it doesn't seem like an academic paper with extensive empirical analysis (in fact, it doesn't have even a single regression) or with an analytical model that enhances our understanding of theory (doesn't have a single equation either). It is a collection of case studies and definitions, albeit very nicely done. Given my area of expertise in entrepreneurial finance and my own experience publishing an empirical study in ResPol, I feel ill-equipped to offer a recommendation on whether or not this paper fits the scope of ResPol. I believe that is largely an editorial decision and you would be the best judge for it. So, I will let you decide on this bigger question and will skip my minor suggestions for the author. I hope that is alright.
Reviewer 3's Comments
Reviewer #3: I'm recommending a major revision to the paper, which would likely constitute a lot of additional work, but I feel would greatly strengthen its contribution.
The objective of this paper is establishing fifteen measures of HGHT entrepreneurship activity, and giving examples of their application and potential usefulness especially with regard to the behavior of what the author terms policy cartels.
The defining of key terms is a useful contribution of this paper, as are the identification of potentially useful metrics of HGHT entrepreneurship. Furthermore, the examples are often helpful in highlighting their various applications.
In my view, the major flaws in the study are not (1) considering possible downsides of each individual metric, (2) considering possible downsides in using the entire battery of measures, and (3) testing alternatives to this measurement approach. I briefly elaborate on these three points in the following paragraphs.
In the conclusion, the author states that two antecedents to improved policy are "standardized measures, to reduce the information asymmetry between policymakers and constituents…and…a simple but grounded framework that can reduce the expertise required to develop and enact productive startup policy."
First, it is not necessarily true that "standardized measurement" improves outcomes, such as policy, overall. Measurements are rarely if ever neutral (vis-à-vis behavior). Two examples are,
- First, we often implicitly assume that more information in the hands of decision makers is unambiguously good. But this requires a lot of assumptions that do not hold in real life. An additional measure, say, would influence decision making but perhaps the distortion is welfare decreasing. Nassim Taleb gives the example of "value at risk" (VAR), a metric commonly used in finance, in the foreword of the book Lecturing Birds on Flying (pp. xvi).
- Second, systematic bias due to gaming metrics incentivized by rewards attached to the measurement. Said differently, an agent would have the incentive to change behavior in the least costly way in order to maximize the payoff associated with the measurement---this behavioral change might have nothing to do with the underlying phenomenon of interest. By way of example, Weisbrod, Ballou, Asch in their book Mission and Money, discuss various measures used in university rankings published by US News & World Report, and how these are often finessed by schools in ways that have little to do with education (see pp. 64-65 for one such discussion).
The point is, any conceived standardized framework is not necessarily better than nothing, and not all frameworks would be of equal value. I would have liked to see a more rigorous discussion of the merits of the framework purposed, which analysis would include at least the following:
- Discussion of alternative calculations of a measure when applicable. For example, the paper's Measure 1 is a composite measure of three sub-measures: why was this particular normalization of sub-measures (i.e., ranking) used? Why was this particular aggregation method (i.e., summation) used? What are the upsides and downsides of this and other approaches? For example, "the flow of dollars" would have a long right tail, which is obliterated when transformed to a rank. Is this a good or bad thing and why?
- Measurement is often reductive (in the sense that they constitute a mapping from a high-dimensional, potentially complex space, to a far simpler space). What is left out? Is it important? In part this translates to discussing the difference between the conceptual phenomenon and how it is operationalized?
- Might a measure be systematically biased or lead to bias if implemented?
- Tests of the effects of using a framework on various outcomes.
In summary, more skepticism about the usefulness of the metrics and the framework, and more empiricism is necessary.
Timeline of Submission
This paper came from a presentation that I made to the Kauffman UMM Grant Cohort. I originally attempted to add empirics but this approach necessarily reduced the coverage of material: Although the framework is simple and used in practice, it is also on the frontier of research, so there aren't any published academic papers with the empirics. So I opted to break the original submission in two - breaking the empirics back out and leaving this as the best attempt I could make at a narrative-based exploration of the whole framework. It is, as a consequence, a very unusual paper. But most people I showed it to were enthusiastic. It is also reference-bait. Outside the review process, some readers were both amused and worried about its snarky tone, which I'm still trying to address.
This paper had a storied submission process:
- The deadline for resubmission was June 15th, 2020. Before this deadline, I emailed the editors and offered them either this version of the paper, which contains no empirics, or an empirical paper without examples or definitions. I received no response.
- This version of the paper was submitted as an R&R to a Special Issue of Research Policy on June 10th, 2020, with manuscript number RESPOL-D-19-01438R1.
- On September 15th, I sent an email to the editors requesting information but received no response.
- I wrote to the editors again on October 27th, this time using the Elvesier form, to request another update. The last status reported by Elvesier (https://ees.elsevier.com/respol/default.asp) was 'Required Reviews Complete' on October 9th, 2020.
- On October 28th, I received an email saying: "Hello Ed. I hope to get back to you shortly. I have two good reviews and I’m waiting on a third. This most definitely will be another R&R. More soon"
- On November 8th, I got an official email about the paper that said: "We have now received the referees' reports on your paper, copies of which I enclose below for your information. As you will see, the referees make various comments and suggestions for improvement. I have given up on the third reviewer and want to return the paper to you." However, this email only contained one review. I requested clarification and noted that Reviewer 3 had asked for empirics.
- On November 11th I got an email that said: "Hello Ed, This is strange. The comments were in the comments to the editor. Here they are and they are not worth that much. This special issue has a specific purpose. You do not need to run regressions!" (Note that the comments are below as Reviewer 1. They indicate that the reviewer accepted the paper.)
Research Policy Special Issue
This paper is for a Special Issue of Research Policy, organized by/for the UMM grant cohort. The deadline for submission is Nov 30th, 2019. See: https://www.journals.elsevier.com/research-policy/call-for-papers/uncommon-methods-and-metrics
Examples of questions that papers could address are:
- What fundamental constructs or elements might constitute a theory or theoretical base for the geographically defined entrepreneurial ecosystem?
- What are general definitions of entrepreneurial ecosystems so that entrepreneurial ecosystems can be measured in a consistent way across all sectors?
- What key relationships need to be captured at the entrepreneurial ecosystem level?
- How should the impact of local entrepreneurial ecosystems on economic growth at the national level be measured?
- Whose performance (and what) should be measured? Should researchers look at a mix of short- and long-term measures?
- Do existing rankings for entrepreneurial ecosystems measure what they claim to measure?
- To what extent are entrepreneurial ecosystems and innovation related?
- What are the salient levels of analysis (e.g., cultural, institutional, spatial) to consider when analyzing entrepreneurial behavior?How do the characteristics of entrepreneurial ecosystems vary by country?
- By which mechanisms do entrepreneurial ecosystems get established, mature, decline, or get renewed?
- What are the trade-offs between attracting entrepreneurs to a city, and solving urban problems such as affordable housing?
- Under what circumstances could a university be considered an ecosystem, and how does this interact with entrepreneurial ecosystems?What are more finely grained evaluations of the effectiveness of policy instruments that capture connections and ties across entrepreneurial ecosystems?
- To what extent is government policy accelerating or inhibiting the progress of entrepreneurial ecosystems?
Data and Analysis
The paper uses vcdb4 and US Startup City Ranking, as well as a wealth of old McNair material. Sources include (copied to the project folder unless otherwise noted):
- Hubs: Hubs Data v2_'16.xlsx
- Federal Grant Data, including NIH, NSF and other grant data, especially SBIR/STTR. Possibly also contract data.
- Agglomeration, including the locality indicators, and American Community Survey (ACS) Data
- Market vs. non-Market? E:\mcnair\Projects\Houston\MarketNonMarket
- Location of VCs (foriegn vs. domestic, local vs. not, etc.) E:\mcnair\Projects\Houston\Houston Ecosystem Recommendations\2017ReportV1.xlsx
- Pipeline and raise rate for Houston: E:\mcnair\Projects\Houston\Acc Rank (IB) -- moved to subfolder pipeline
- U.S. Seed Accelerators, and also other source material for Determinants of Seed Accelerator Performance: The Horse, the Jockey, and the Racetrack. Likely just the load of the file to rule them all...
- Incubator Seed Data
- Carnegie Classifications of Institutes of Higher Education (see University Patents). A new public data file was downloaded from http://carnegieclassifications.iu.edu/ and put in the folder (CCIHE2018-PublicData.xlsx).