The California Pipelines of Silicon Valley

Last update; Friday 7/30/16

... Why Silicon Valley hires so few Black and Hispanic professionals ...

Silicon Valley's top executives claim that they employ so few Blacks and Hispanics because there are so few minorities in the nation's best computer science programs that are in the Valley's academic "pipelines." Nevertheless, numerous recent studies have found high Black and Hispanic enrollments in many of the nation's leading computer science programs. That's why I decided to explore an alternative hypothesis that the Valley did, indeed, recruit from the nation's best colleges and universities ... but mostly from the ones that are located in California. 

I found that Asian and White employment is substantially greater than their enrollment shares in California's leading academic institutions, that Blacks are under-represented, but that Hispanics are grossly under-represented. Indeed, the lion's share of Asian and White over-representation in employment has mostly been achieved by Hispanic under-representation. 

A. The case for local recruitment
The best known companies in the Valley rose to their current eminence during periods of hyper-growth in highly competitive market sectors. 

• California is home to many of the nation's (and the world's) most esteemed colleges and universities in all computer-related STEM fields, not just computer science narrowly defined. These institutions are critical components of the Valley's tech ecosystem, providing competitive advantages that are difficult, but not impossible, to duplicate elsewhere.
   
• California's leading institutions facilitate a cost-effective application of Granovetter's "Strength of Weak Ties" hypothesis, i.e., the hiring of competent job seekers who know people who know people who already work for the Valley's rapidly growing tech firms. The most likely gatekeepers in these networks are full-time faculty who are consultants for the firms in the Valley and adjunct faculty who are full-time employees of the firms in the Valley.
   
• California students who seek internships and/or permanent employment in the Valley incur minimal transportation, meals, and lodging expenses during their job searches; and local hiring minimizes the subsequent relocation costs and personal disruptions for the students who are ultimately hired. On the other hand, local hiring minimizes the time and other costs to the Valley's firms that need to recruit qualified students as quickly as possible.

• Enrollments in California's leading institutions are large enough to provide a steady stream of candidates that can satisfy the recruitment requirements of the rapidly expanding firms in the Valley. Rapidly expanding firms that need to hire more tech staffs next month or even next week won't have the time or the need to "import" qualified employees from the nation's other large academic centers that are thousands of miles away on the other side of the Rockies; hence there is little or no business justification for doing so. Invoking the classic "80-20 rule", I would expect that more detailed data than I currently have would show that at least 80 percent of the techs in the Valley were recruited from the best colleges and universities in California; at most 20 percent came from institutions outside of California

B. The Data
Table 1 (below) displays enrollment data for the Fall 2014 semester that I obtained from the the IPEDS Data Center by following the procedures described in Appendix #1 at the end of this note. Data for Fall 2014 was the most recent data available from IPEDS as of July 2016 when this note was written.

Table 1. Enrollments at Top California Colleges & Universities

Institution	SAT25	Total	Asian	Black	Hispanic	White	subTotal  ABHW
California Institute of Technolog	770	2209	580	27	173	743	1523
Harvey Mudd College	740	804	167	13	80	352	612
Stanford University	700	16963	2630	633	1749	6218	11230
Claremont McKenna College	690	1324	142	54	154	559	909
Pomona College	690	1650	213	109	232	702	1256
University of Southern California	660	42453	7696	2238	5286	13922	29142
University of California-Berkeley	640	37565	11172	886	4529	11519	28106
University of California-San Dieg	630	30709	10443	388	4493	7755	23079
Santa Clara University	620	9015	1416	238	1283	3691	6628
Scripps College	620	988	181	31	93	483	788
Occidental College	610	2040	254	90	303	1034	1681
University of California-Los Ang	610	41845	11257	1229	7066	12824	32376
San Jose State University	470	32713	10381	1029	7407	7112	25929
Totals for All Selected Institutions		220278	56532	6965	32848	66914	163259

Given the exploratory nature of this small study, I used a simple criterion to identify the most prestigious STEM-related institutions in California ==> I selected the colleges and universities wherein 75 percent of the entering freshmen scored at least 600 on their Math SAT scores (second column). The first twelve institutions that appear in this table contain few surprises, but readers will note that the table also contains a thirteenth institution: San Jose State University (SJSU): 

• Although its 25 percentile score on math SATs is well below 600, the main campus of SJSU occupies a strategic location in the heart of the Valley. This made me suspect that its graduate programs might be a major component of the Valley's academic pipeline even though it wasn't one of the "nation's leading institutions" from which the Valley's top executives claimed they recruited most of their employees. My suspicions received initial confirmation from the prominent banner on SJSU's home page that proclaimed that SJSU is "powering Silicon Valley."
   
• Further confirmation came from Wikipedia's description of SJSU that declared that "More San José State University alumni are hired by Silicon Valley firms than graduates of any other college or university"... Wikipedia backed up this declaration with links to TV reports about a survey conducted by a division of LinkedIn last year: "San Jose State University is Top School for Most Silicon Valley Hires" (NBC Bay Area) and "San Jose State Alums Beat Out Elite School Grads For Tech Jobs" (CBS SF Bay Area).^]  
   
• Confirmation of SJSU's capacity to play a major role in training/retraining the Valley's tech staffs came from its Web pages that list an extensive array of "100" and "200" graduate level courses that will be offered in the Fall 2016 semester by its biomedical engineering, computer engineering, computer science, electrical engineering, graphic design, and software engineering programs. Who's paying to take all of these advanced courses? Common sense suggests two groups: Valley employees who want to keep their tech skills up-to-date and employee wannabes.
   

The last column of Table 1, "subTotal ABHW", contains the subtotal = Asian + Black + Hispanic + White students. This subtotal is always smaller than the data found in "Total" column (third column from left) because the "Total" column also includes IPEDS enrollments for students in other categories, e.g., mixed races, unknown races, Native Americans, and other groups. The last row of Table 1 displays each group's relative share of the subTotal ABHW. For example, the Asian relative share = Asian enrollment divided by Asian + Black + Hispanic + White enrollment at each institution.

Table 2A in Appendix #1 at the end of this note displays the percentage of total employment of Asians, Blacks, Hispanics, and Whites that was reported by four prominent tech firms in the Valley in their most recent diversity reports ==> Google (2016), Facebook (2016), Intel (2015), and Apple (2016). 

Table 2 (below) displays the relative employment shares of Asians, Blacks, Hispanics, and Whites at each company. For example, the Asian relative share = Asian share of total employment divided by Asian + Black + Hispanic + White share of total employment. These relative shares enable comparisons of the relative employment shares of Asians, Blacks, Hispanics, and Whites to their relative enrollment shares that were displayed in Table 1. 


Table 2. Employment Shares & Enrollment Shares

Rosters	Asian	Black	Hispanic	White	Year
Enrollment	35%	4%	20%	41%	2014
Google	38%	1%	3%	58%	2016
Facebook	47%	1%	3%	49%	2016
Intel	37%	3%	8%	52%	2015
Apple	39%	2%	5%	54%	2016

C. Findings
It should be easy for the four Silicon Valley titans to recruit qualified Black and Hispanic students from the California pipelines. I expected the relative employment shares of Blacks and Hispanics would more or less equal their relative enrollment shares ... but they aren't.

The over and under representations of Asians, Blacks, Hispanics, and Whites displayed in Table 3 (below) were calculated by subtracting their relative enrollment shares from their relative employment shares, as shown in Table 2 (above). For example, subtracting the Asian 35% enrollment share from its 38% employment share at Google yields a positive 3% over-representation. Whereas subtracting the Hispanic 20% enrollment share from its 3% employment share at Google yields a negative -17% under-representation. The over and under representations at each company are also displayed in the four charts that follow Table 3 (below). 

• Table 3 and the charts show that Blacks are under-represented at each company, but that Asians and Whites are over-represented at each company
   
• Table 3 and the charts show that Hispanics are grossly under-represented at each of the four companies. Indeed, most of the Asian and White over-representation has been achieved via Hispanic under-representation.

Table 3. Over and Under Representations

Firms	Asian	Black	Hispanic	White
Google	3%	-3%	-17%	17%
Facebook	12%	-3%	-17%	8%
Intel	2%	-1%	-12%	11%
Apple	4%	-2%	-15%	13%

Charts  

D. Recommendations
The official explanation for the under-representation of Blacks and Hispanics in Silicon Valley -- that there weren't enough minorities in the nation's academic pipelines -- has lost credibility with each passing year. That's why I suggest that fast growing firms in the Valley recruit most of their employees from the best colleges and universities in California because California is the nation's most populous state and a state that is home to some of the finest academic institutions in the world. Expanding at warp speed, the Valley's high flyers don't have the time nor the business incentives to launch recruiting campaigns that are truly national coast-to-coast-to-coast operations ... except for recruiting a relatively small number of superstars. 

While I haven't gathered enough data to formally reject the official explanation, my alternative offers two advantages. First, it's more plausible; and second, it suggests ways to reduce the under-representation of Blacks and Hispanics, whereas the Valley's conventional wisdom leads to a bleak dead end.

1) Hire more Hispanic students
Fortunately, this is one of those rare opportunities that returns big gains from modest investments. Hispanics have become the largest racial group in California; non-Hispanic Whites are now the second largest -- 38.8% vs. 38.0%.  And California also has a larger Hispanic population than any other state in the nation, over 15 million. Moreover, the Hispanic population is growing faster than the White population, so the large 20 percent share of Hispanic enrollments in the state's foremost academic institutions is likely to increase in the next few years. 

In other words, if a company wanted to recruit more Hispanic techs, California would be the number one state from which to do so. This suggests that the low Hispanic share of the Valley's tech employment is just a temporary phenomenon that reflects the small number of Hispanic gatekeepers currently available,  i.e., the people who know the people in Valley who make the hiring decisions.

• Intensive efforts to add a few more Hispanics to the Valley's technical staffs will provide a much large number of Hispanic students with contacts in the Valley, which will lead to hiring more Hispanic students, who will provide more contacts, etc., etc., etc.  A similar virtuous cycle facilitated the steady rise of Asian Americans from near zero employment in the Valley in the 1950s before the Supreme Court's 1954 desegregation decision and the subsequent Civil Rights legislation in the 1960s. Courts and Congress removed the legal obstacles to Asian American enrollments in California's most selective public and private colleges and universities.
   
• The impact of these positive feedback loops will be greatly enhanced by the Valley's use of CODE2040 (and similar organizations) to supplement the current deficiencies in its networks of Hispanic gatekeepers. CODE2040's internships, conferences, meet ups, and other initiatives will expose more Hispanic students to more employment opportunities in the Valley and will provide the Valley's decision makers with more opportunities to identify qualified Hispanic recruits.

2) Hire more Black Students
If California's large Hispanic population makes it the best state from which to recruit highly talented Hispanic techs, its small Black population (2.5 million = 6.5% of the state's population) and the even smaller Black share of California's academic pipelines (4%) makes it one of the less promising states for recruiting highly talented Black techs. Nevertheless Google, Facebook, and other major companies should make intensive efforts to hire more Black techs; and they should also engage the assistance of CODE2040 (and similar organizations) to amplify the impact of their recruitment efforts. 

E. Conclusions
When a new steady state is reached five to ten years from now, the total number of techs in the Valley may be twice the number employed today. However, it's likely that the White employment share of the Valley's tech jobs will be substantially lower than today, i.e., closer to current White enrollment shares. Asian employment shares may edge a bit higher. If the Hispanic employment share rises to at least 10 percent, i.e., to more than half its current enrollment share, this might yield more than ten times as many Hispanic techs as today. And if Black employment in the Valley rises to its current enrollment share, i.e., 4 percent, this might yield five or six times as many Black techs, but their numbers will still be substantially lower than the employment levels of the other racial groups.

Until very recently I was annoyed, but not very concerned that there were so few Black techs in Silicon Valley because I knew from personal experience that Black involvement in computer technologies was substantially greater in other parts of the country. But I am gravely concerned by the Valley's sudden emergence as the nation's preeminent center for the commercialization of powerful machine learning and other AI technologies that will drive mega-disruptions in every workplace in every part of the country. I am concerned because there will be relatively few Blacks involved in revolutionary decisions that will determine whose careers will be enhanced by computer algorithms and whose careers will be terminated.

As it happens, the lion's share of Black America's Talented Tenth are still clustered in urban agglomerations that are located thousands of miles away in the Eastern and Southern states. Imagine a parallel universe wherein Silicon Valley had somehow emerged a few miles from New York or Baltimore/Washington or Atlanta, instead a few miles from San Francisco. In that universe, Black under-representation would be the same kind of temporary phenomenon that it will be for Hispanic under-representation in the real Silicon Valley of our universe, an under-representation that will inevitably give rise to substantial representation. Soooooooooo ... All we have to do is figure out how to persuade the real Silicon Valley in our universe to move ... or ... persuade it to establish major satellite operations near urban areas in the Eastern and Southern states, e.g., New York, Baltimore/Washington, and Atlanta ... :-)

Appendix #1 -- IPEDS Data

The data for the 13 institutions in the Table 1 were obtained from the Website of the IPEDS Data Center as follows: 

1. Click "Compare Institutions"
2. Click "Continue" (bottom of page) to use final release data
3. Hover over "By Groups" ... then click "EZ Group" to select the institutions
4. Click "State or other jurisdiction" ... then check "California"
5. Click "Sector" ... then check "Public, 4-year or above" ... check "Private not-for-profit, 4-year or above" ... A line should show the number of institutions that have been selected
6. Click "Search"
7. Click "CONTINUE" (middle of page)
8. Click "Admissions and Test Scores" ... click the second "Admissions and test scores" line ...  click "SAT and ACT test scores" ... check "2014-15" ... and check "SAT Math 25th percentile score" 
9. Go back to the left margin, scroll down ... click "Fall Enrollment" ... click "Race/ethnicity, etc" ... click "Race/ethnicity (new)" 
    ... check "Fall 2014" 
    ... click "Level of student"  ... check "All students total" ... click "Save" 
    ... check "Grand total" ... "Asian total - new" ... "Black or African American total - new" ... "Hispanic total - new" ... and "White total - new"
10. Scroll back to the top of the page and click "Continue" (upper right)
11. Click "Continue" (middle of page, right) ... table should show the six selected variables
12. Click "Continue" on the page that will download the output file in .csv format
13. Provide a name for the file to be saved and click "Save" (on Macs)
14. Use Excel, R, Python, Stata, SPSS, etc on the downloaded output file to select the institutions whose SAT 25th percentile score is 600 or above ... also select the data for San Jose State University

APPENDIX #2 

Table 2A (below) contains two rows for each of the four companies:  The first row contains the Asian, Black, Hispanic, and White percentages of total tech employment, i.e., the raw percentage data that was included in the company's diversity reports: Google (2016), Facebook (2016), Intel (2015), and Apple (2016). The second row calculates each racial group's relative share of the sum of the employment of the four groups in each company. (Note: Apple's report provided counts of employees in each group from which   the percentage of employees in each group were calculated.)

Table 2A. Employment Shares vs. Relative Employment Shares

Firms	Asian	Black	Hispanic	White	Year
Google/Tech (raw)	37%	1%	3%	57%	2016
Google	38%	1%	3%	58%	2016
Facebook/Tech (raw)	46%	1%	3%	48%	2016
Facebook	47%	1%	3%	49%	2016
Intel/Technical U.S. (r	36%	3%	8%	50%	2015
Intel	37%	3%	8%	52%	2015
Apple/Professional (ra	39%	2%	5%	53%	2016
Apple	39%	2%	5%	54%	2016

____________________
Related notes on this blog:

• Best U.S. cities for Blacks in tech????  (July 2016)
• Major Centers for Blacks in Technology (February 2016)