In this paper, we explore the conditions under which learning at the individual level transfers to new firms. We use novel survey data from entrepreneurs combined with USPTO patent data for their firms to examine performance of startup firms as a measure of outcomes produced by learning acquired from prior founding experience. The results show that performance-enhancing learning varies with shifts in the industry and with the type of experience. Results indicate that at moderate levels of technological novelty prior experience is especially helpful, however, after significant technological changes in the electronics and software industries, this benefit becomes a disadvantage.
Can entrepreneurship be learned? Individuals with prior founding experience are often written about in the popular press and even seen as celebrities. They are frequently pursued by venture capitalists due to their perceived importance for the successful commercialization of new technology through entrepreneurial firms. For example, consider Bob Langer, an MIT Institute professor who has started at least twenty-four companies. Experienced founders are also celebrated as fostering and investing in future generations of entrepreneurs. Larry Bock has founded eleven life sciences companies, including nine that went public. He boasts having founded, co-founded or invested in forty companies with a cumulative market capitalization over $40 billion. However, is prior startup experience always beneficial? For each successful serial entrepreneur, there may be scores of individuals who have a string of failures. What determines these different outcomes? Relative to its importance, relatively little academic attention has been pointed towards experienced founders and the conditions when they outperform first time entrepreneurs (Hsu, 2007).
We take advantage of differences between those entrepreneurs who have and have not had the prior experience of founding a firm. We use the dotcom boom and several periods of significant technological change in the electronics industry to test the impact of these events on the impact of learning on new firms. We illustrate the history of the microprocessor segment of the electronics industry for several reasons. First, it was at the forefront of rapid changes in the industry, driving changes and opportunities in components, peripherals, and much of the electronics ecosystem, including other uses for these emerging microprocessors. This sector is rife with turmoil and high-velocity, fast-paced change, yet with the benefits of hindsight, we select three time periods of particularly dramatic changes and important technological innovations. Second, it is a large, cohesive segment of the industry that is well documented. Finally, MIT played key roles in the development of this area, and we have access to unique data from MIT to use in our analyses. Many MIT alumni entrepreneurs came from the Electrical Engineering and Computer Science Department and the electronics firms they founded are likely to have been influenced by these time periods of rapid change in microprocessors and the broader electronics industry. The three time periods we highlight are the years following 1981 (the introduction of RISC and 32-bit chips), 1988 (the IBM PS/2 and “microprocessor wars”), and 1990 (the introduction of MIPS, 64-bit chips, and the “clone wars”). These time periods clearly highlight the phases of discontinuity and industry ferment characterized by the models of industry evolution (Utterback & Abernathy, 1975).
The early 1980s was the age of significant advances due to “RISC” processors and 32-bit computer. David Patterson and Carlo Sequin, both on the faculty at Berkeley started the RISC Project in 1980, emphasizing pipelining and use of register windows to speed up processing. Their first processor, the RISC-I outperformed every other single chip design at the time while using fewer than half the transistors. RISC was quickly adopted throughout the industry and in the next few years, AT&T Bell Labs, Motorola, DEC, and HP all introduced RISC-based processors. In 1981 and during the next few years, major changes and technological developments took place in the electronics and microprocessors industries. The first PC, the first portable computer and the first workstation were all introduced. When IBM released its first PC, with a 4.77 MHz Intel 8088 processor and Microsoft’s MS-DOS operating system, it ignited rapid growth in the market. The first portable computer, the Osborne I, included a 5-inch display, 64 kilobytes of memory, a modem, and two 5 1/4 – inch floppy disk drives, weighing 24 pounds. Apollo Computer introduced the DN100 as the first workstation, optimized to run graphics-intensive engineering software. The introduction of 32-bit microprocessors also revolutionized the industry. AT&T’s Computer division introduced the first single-chip 32-bit microprocessor in 1981 (BELLMAC-32A). Motorola, National Semiconductor and Intel quickly followed suit. Also of note in 1981, John Hennessy and a team at Stanford started working on the first MIPS processor. We will return to this development and its impact below.
By 1982, the operating speed of computers on the market almost doubled when the Cray XMP was introduced, running 420 million operations per second, or megaflops. Also that year, the Commodore 64 was introduced (with 64KB of RAM) and became the greatest selling single computer model of all time. Also during this time, Apple introduced the Lisa as the first personal computer with a graphical user interface (GUI). In 1984, Apple released the first commercially successful computer with a mouse and GUI, the Macintosh. An alumni survey is particularly appropriate because it enables gathering data from a well-defined population of comparable individuals.
By 1988, IBM had introduced its PS/2 computers, making the 3-½ inch floppy disk and a video graphics array standard. The company sold more than 1 million units in the first year and released the OS/2 operating system, allowing a mouse to be used with IBM machines for the first time. By the end of the 1980s, an industry shakeout was underway as prices dropped, volumes shipped hit record numbers and consolidation occurred during the period of time known as the “microprocessor wars”.
By 1991, John Hennessy had left Stanford to start MIPS Computer Systems and had begun shipping the R4000 processor. It was the industry’s first 64-bit microprocessor (he later returned to become University President). MIPS increased computing speed by overcoming a major obstacle to pipelining that required interlocks to be in place to process multiple-clock cycle instructions. The MIPS design prevented the pipeline of loading data from being delayed by eliminating interlocking through enabling all instructions to take only one clock cycle. Similar to the way that 32-bit chips revolutionized the industry in the early 1980s, MIPS and 64-bit chips drove revolutions in electronics during the early 1990s. In 1990, Motorola announced the 32-bit 25 MHz 68040 processor, incorporating 1.2 million transistors. In that same year, Intel introduced the 486 microprocessor, running at 33 MHz and 27 million instructions per second (MIPS). Surpassing it shortly afterwards, in 1991 Intel introduced the first processor to use 0.8-micron technology, the 486 running at 50 MHz and 41 MIPS. IBM introduced the POWER architecture in 1990 and by 1991, the first single-chip PowerPC derivatives were on the market in high volumes. DEC launched the Alpha 21064 in 1992, referring to it as the world’s fastest processor at 200 MHz and 64-bit RISC design. The Intel Pentium chip was first launched the next spring, but initially only ran at 66 MHz. Finally, it was also around this time that the “clone wars” began as Advanced Micro Devices brought down prices by settling a lawsuit with Intel and retaining rights to clone the 386 and 486 chips.
Data and Sample
We use a novel survey administered in 2001 to all 105,928 alumni from the Massachusetts Institute of Technology (MIT) to generate a sample of firms where we have detailed information on founders as well as on firm performance.
This survey generated 43,668 responses. Out of 7,798 alumni who had indicated that they had founded a company, 2,111 founders completed more detailed surveys in 2003, representing a response rate of 25.6%. In previously published work the authors show t-tests of the null hypothesis that the average (observed) characteristics of the responders and non-responders are the same statistically, for both the 2001 and 2003 surveys. Industries covered include aerospace, architecture, biomedical, chemicals, consumer products, consulting, electronics, energy, finance, law, machine tools, publishing, software, telecommunications, other services, as well as other manufacturing. Each founder reported information on firms that he or she had founded or attempted to found up to the date of the survey. This new database was further updated to 2006 data from the records of Compustat (for public companies), United States Patent and Trademark Office (USPTO), and Dun & Bradstreet (private companies). A total of 3,156 alumni had indicated that they had started multiple companies, of whom 960 completed the survey for a multi-founder response rate of 30.4%. A total of 1,107 single-firm founders responded to the survey giving a 21.8% response rate out of the 5,086 single-firm alumni founders. The results of this survey have been reported on previously (Hsu, Roberts, & Eesley, 2007).
An alumni survey is particularly appropriate because it enables gathering data from a well-defined population of comparable individuals.
 In April 1981, Byte Magazine Editor in Chief Chris Morgan mentioned the Osborne I in an article on "Future Trends in Personal Computing." He wrote: "I recently had an opportunity to see the Osborne I in action. I was impressed with its compactness: it will fit under an airplane seat. (Adam Osborne is currently seeking approval from the FAA to operate the unit on board a plane.) One quibble: the screen may be too small for some people´s taste."
 The Lisa´s slow speed and high price ($10,000) led to its ultimate failure. The Lisa ran on a Motorola 68000 microprocessor and came equipped with 1 megabyte of RAM, a 12-inch black-and-white monitor, dual 5 1/4-inch floppy disk drives and a 5 megabyte Profile hard drive. The Xerox Star — which included a system called Smalltalk that involved a mouse, windows, and pop-up menus — inspired the Lisa´s designers.
 Successful matches were found for 80% of the company names in the D&B database. A firm is included in the Dun & Bradstreet database when it needs to obtain a credit rating.