Sunday, July 12, 2020
By:
This internship is going by so quickly. Covid-19 has changed my perception of time, such that the beginning of lock-down in late March seems a lifetime ago; while this internship seems to have only been four weeks. I am shocked that I am writing about week six, the first discrete step on the downhill race to the end of this life changing internship.
My research paper is going along smoothly. This Friday I received the first batch of information from research economists at DoC. I decided to work this Sunday to better understand the data I am working with. As one would expect, .xls and .csv tables of job codes, median incomes, ect. My reasoning is that this research project is really my own work; only I decide if I publish a paper after this internship ends and I know I can make this paper happen.
The DoC economists Nico and Stephen are amazing. I met with them on Tuesday the 7th, and our meeting was incredibly productive! Not only were they happy to help, but they had great ideas as well. One of these was to consider USPTO application and citation rates as independent variables for measuring innovation. We discussed how to leverage using both NAICS (North American Industry Classification System) and SOC (Standard Occupational Classification) systems, and how both had resolution to the county level.
At the full-office meeting, I announced that my meeting with Nico and Stephen went well and that I was looking forward to writing a paper. It was a little intimidating to have my supervisors’ director and deputy director question me, but I think it went well! When I met with my supervisors afterwards for our smaller meeting, the communications director for the Office of Advanced Manufacturing (OAM) was also on the line and interested in my research. She asked me to write an abstract level white paper, and this is what I sent her (minus footnotes):
Shortest:
I would like to publish a quantitative analysis on areas of friction in the advanced manufacturing labor supply chain, and to test for above average growth attributable to the establishment of an institute. Data will be gathered with the help of Nico and Steph, from sources such as the Bureau of Labor Statistics Standard Occupational Classification (BLS's SOC) system, and from EMSI, etc.
Shortest:
I would like to conduct a quantitative analysis on areas of friction in the advanced manufacturing labor supply chain, and to test whether growth is attributable to the establishment of an institute. Data will be gathered with the help of NIST MEP economists, Nico and Steve, from sources such as the Bureau of Labor Statistics Standard Occupational Classification (BLS's SOC) system, and from EMSI, etc. In conjunction with UC Berkley and AIP (American Institute of Physics), I will seek to publish the results of this analysis for independent peer review.
Abstract level:
This paper examines the treatment effect of institutes of advanced manufacturing seeded by the interagency Manufacturing USA program. With tens of millions of dollars awarded annually in grants resulting in even more private sector investment. With hundreds of individuals involved with each institute’s enterprise, it is expected that there is an empirical (and positive) impact because of these manufacturing innovation institutes.
In order to test this hypothesis, we establish two independent controls. First, a standard regression analysis establishing the average national trends for our metrics of choice (unemployment, median wages, productivity, ect) is performed on data obtained from both the Census Bureau’s (CB) county business patterns database, and through EMSI (a data aggregation utility used by U.S. Department of Commerce staff). Our data scales geographically where it is granular at the county level, and also scales by the 5-6 tiers (or “digits”) of industry and employment taxonomies respectively adopted by NAICS (used by CB) and SOC (used by the U.S.BLS)
Our second control method replaces national behavior with the 5-10 applicants which applied for recognition as a Manufacturing USA institute for the same funding now used by our treatment sample. We expect more homogeneity amongst peer applicants than all county data aggregated to the national level, and that this can be used to identify the effect of treatment. Although this method will rely upon significantly smaller sample sizes, it automatically removes national outliers in our data (e.g. the extreme variance in population densities between mega-metropolis such as NYC, SF, and LA and the myriad sparsely populated counties resting between the Mississippi and the Sierra Navada). Finally, time permitting, least-squares analysis will be compared with a more robust method of analysis known as synthetic control. This is a form of analysis that is well-established and is commonly used in physics. It has begun to gain traction with economists in the last 10-20 years.
A final method of analysis in consideration is a regression discontinuity design. Many institutes have satellite campuses which were incorporated at later dates. We would look at whether introduction into the Manufacturing USA program has a similar impact across a multi-year scale for any given organization—not simply institute headquarters. It is uncertain whether these subsidiaries’ signals will be discernible beyond stochastic noise, but it is a viable path of analysis.
Finally, this research will be original and novel, as our analysis will include both standard foundational measures of economic performance such as unemployment, GDP, etc; and unique metrics which allow specific targeting of our data. One prototype which Nico and Stevephan both liked and knew they could obtain data for, was measuring unfulfilled job openings, but only for a handful occupations we identify as maximally relevant for advanced manufacturing. Nico and Stevephan have told me that they can provide time-series ordered data that can be customized by region, occupation, industry, ect. Finally, because of this, we expect that the tiered nature of both NAICS and SOC can be leveraged when comparing the national data analysis to the regional data analysis. One example: at the national scale there are similar distributions for all engineers and for the combined average of all engineering specialty sub-types. However, if we are interested in knowing the unique impact that NIIMBL has had on it’s regional economy and community, then we can look at whether there is a statistically significant clustering of bioengineers around Newark, DE (headquarters), or NIIMBL’s satellite institutions, or even the locations of members.
If we are successful in showing that the utility derived from the establishment and funding of advanced manufacturing programs is real and measurable, then we are also simultaneously providing tangible evidence in support of federal budgets which continue to fund the office’s ongoing mission and the next generation of manufacturing programs that AMNPO and MEP will sheppard.
Max Dornfest