Graduate education beats on-the-job learning as the predominant driver of new inventions in tech firms, according to research by James D Adams, to be presented at the Royal Economic Society''s annual conference at the University of Bristol in April 2017.
His study analyses the patents and scientific papers of 800 researchers who were employed by five software and internet services companies between 1991 and 2013. It tracks the number of researcher inventions, the scientific papers they produced while they were in graduate education, and the patents their working groups generated.
The analysis shows that researcher inventions increased by 0.5% for each 1% increase in papers (the ''School of Hard Knocks'' effect), and by 0.25% for each 1% increase in working group patents (the ''Rubbing Shoulders'' effect). It also measures the output of papers by thesis advisers, and finds that the effect was tiny, but still positive: inventions increased by 0.04% for each 1% increase in adviser papers (the ''Curmudgeon'' effect).
The author concludes: ''The source of researcher inventions is predominantly graduate education and motivation, with a smaller contribution from experience with working groups in companies, and an even smaller contribution from thesis advisers.''
This research looks at inventions by industrial researchers. It asks: how important for invention is graduate education and researcher motivation compared with experience in firms? Do thesis advisers contribute to invention by researchers, their one-time students? Can inventions in turn be assigned to products or processes, which are the purpose of invention?
To provide answers, we study a sample of PhD researchers employed in software companies. The sample hand-matches researcher data with firms to their graduate education. The sample consists of eight hundred researchers and their patents and scientific papers, who are employed on average for 6.5 years by five important companies in the software and internet services industry. The period is 1991 to 2013.
To capture graduate education and researcher motivation, we use researchers'' scientific papers from their PhD degree. We find that researcher inventions increase by 0.5% for each 1% increase in PhD papers. Let us call this the ''School of Hard Knocks'' effect.
To assess research experience in firms, we exploit patents of working groups where researchers are assigned. We find that researcher inventions increase by 0.25% for each 1% increase in working group patents. Call this one the ''Rubbing Shoulders'' effect.
To measure contributions of graduate advisers, we make use of advisers'' scientific papers. We find that researcher inventions increase by 0.04% for each 1% increase in adviser papers. While small, the adviser effect is distinguishable from zero. In recognition of loyal service, we dub this the ''Curmudgeon'' effect.
Combining the three, the School of Hard Knocks effect overwhelms the Curmudgeon effect, and the School of Hard Knocks effect is twice the Rubbing Shoulders effect. These differences are real. Expressed differently, the source of researcher inventions is predominantly graduate education and motivation, with a smaller contribution from experience with working groups in companies, and an even smaller contribution from thesis advisers, which is nevertheless positive.
To conclude, consider the final question: can inventions be assigned to products or processes? The answer is yes. Researchers are members of working groups that are assigned to specific products or production processes that improve products.
The important consequence is that patents of researchers can be combined to form rates of innovation. The evidence is that innovation by companies is rising sharply and that it is becoming more diversified across product or process areas. Research builds on prior growth, and it aids future growth by diversifying and improving product offerings.
Theorems, Lemmas, and Apps: The Making of Industrial Software – James D. Adams