Plumbing Tech’s “Pipeline” Problem
Can changing one set of biases overcome another?
The “Ivy League+” of top tier engineering universities have their own salacious pipeline problem.
Tech companies must favor a different set of universities in their hiring process in order to actually make progress on Diversity.
[As a caveat to this whole post: Diversity and Inclusion, and Systematic racism are topics in which I still have a novice level understanding.
I’ve approached this like I would any other problem as a Product Manager, because that’s the background and skills I have to bring to the table, but there could be something significantly problematic about applying that approach to Diversity hiring.
One of the reasons I’m writing this post is to hopefully get it out there and get feedback from people who know more than me and have more personal experience in these areas, so I can learn from their experiences and be a better ally in the future. Please comment or reach out to me if you’re one of those people!]
Tech loves to talk about how it can’t solve its own diversity problems. Because the trouble starts in high school and college with women and racial minorities choosing not to pursue tech degrees, there’s nothing tech can do about it’s current racial diversity right now. For an industry that prides itself on “disrupting” by finding a way to do things the incumbents said couldn’t be done, this always felt like some pretty disingenuous what-about-ism to me.
Even at the industry darlings that everyone looks to for guidance, like Google, the Diversity & Inclusion (D&I) numbers have barely moved at all over the past 2 years.
Black tech workers at Google only increased by .1 percentage point to 1.5%, and Latinx workers didn’t increase at all from 2.8% between 2017 and 2018. Meanwhile both groups experience a significantly higher rate of attrition from the company compared to other racial groups.
Meanwhile these groups made up 13 and 17% of the population.
(Based on Google’s 2018 Diversity Report, and US Census Data)
At this point, it’s pretty clear that the current popular tactic of unconscious bias training and praying aren’t working.
I wanted to dig into this common hiring “pipeline” excuse more and a stat from Google’s 2018 D&I report caught my eye. The report referenced the number of universities Google recruits from as one of 2 highlighted data points around diversity hiring. While Google’s overall numbers have barely changed at all, there’s a slightly larger improvement in their diversity numbers around recruiting. This data point on number of schools recruited from aligned with more qualitative data points from my own experience. It got my product manager senses tingling that there might be a high ROI, highly scalable solution somewhere in this problem area of the hiring pipeline.
“Measure the impact, not the effort.”
First, a rare success story on the diversity in tech front: Gusto focused 100% of it’s outbound recruiting efforts on hiring women engineers for just 6 months and exceeded it’s diversity hiring goals.
The idea behind this tactic was that, with the demographics of the Bay Area tech industry today, we are never going to run out of white dudes applying to our start-up inbound. Therefore, outbound recruiting efforts should be focused on balancing that organic hiring pipeline in other ways.
This is exactly the same mentality companies take when they get plenty of applicants for a standard “Staff Engineer” role, but are struggling to fill a Senior Robotics role: they focus recruiting efforts on the hard to fill position.
It’s also in line with the concept of “equity not equality” in social justice philosophy: Some groups already have more access to opportunities, and therefore giving everyone equal help just expends a lot of effort while keeping everyone equally as far apart in accessing those opportunities as they were to begin with .
On the other hand, equitable distribution of effort seeks to bring everyone up to the same level of access, with commensurate effort given to groups based on how far away they are from the average. An equitable distribution of effort focuses on making end outcomes more equal across groups, not in giving every group equal effort up front.
To bring this back to startup parlance that’s “Measure the impact, not the effort.”
However, one caveat here is that Gusto got exactly what they measured. By focusing exclusively on hiring women, the company’s racial diversity is still really poor.
The lesson here?
- Focusing resources specifically on diversity hiring does lead to improvements.
- One kind of diversity does not beget more diversity automatically. Diversity is a multi-faceted problem that will take multiple cohesive efforts to solve.
It takes effort to set up the right processes to create and support a pipeline of diverse candidates. It takes different efforts to create a pipeline of Women engineers than Black engineers, and both of those are different than creating a pipeline of Black Women engineers.
So we need to think about: How can we target our efforts to make sure we build a holistic solution to diversity over time? It’s an incredibly important question, because if we don’t put in the effort to make sure that our systems promote and encourage diversity across the board, they will solidify into a state that systematically discourages diversity. This is true simply because startups are always trying to move as fast as possible which usually means doing the easiest, most comfortable thing possible. And right now the easiest, most comfortable thing in Silicon Valley is whatever the white dudes are familiar with. That’s the ecosystem we’re operating in.
(This is mostly an essay about how we need to do better with racially diverse hiring, but I want to give a HUGE shoutout to Gusto and make sure it’s clear what an awesome job they did in the area of gender diversity. The company’s numbers on Women in Engineering have continued to improve since their initial efforts in 2015. 50% of their new graduate hires this year are women! And while their actions revealed more blindspots in their and Silicon Valley’s thinking about diversity overall, it also showed that its possible and effective to try something different. Find out more about what Gusto is doing to succeed at D&I and what they’ve learned on their website.)
The second qualitative story highlights this impact of “organic” systems on a company’s composition.
This company had one early employee who had attended a particular university in Canada. Because they were familiar with it, they set up an internship partnership with that university early on, and now every semester — spring, summer, and fall — there’s usually about six interns from that university working for the company here in SF. As you might have guessed, the company’s full-time workforce is significantly more Canadian than the average San Francisco startup.
The lesson here?
- Pipelines that arise organically will most likely mirror the composition and background of the founders/earliest employees.
- The early-career hiring opportunities a company provides affect it’s future composition beyond the average demographics of the Bay Area.
- Diversity doesn’t happen organically but has to be consciously built into the structure of those opportunities.
If our company is made up of primarily white people from traditional white backgrounds, the processes that arise organically are going to favor people with that background. We can imagine the composition of this company would really break the norm if, for instance, this company had a similar partnership with Spelman College, a historically Black women’s college in Georgia.
If we don’t put in effort to counteract the weighting of our company’s current connections, then we can create a systemically racist hiring pipeline without ever meaning to. That’s the hard thing about the situation Tech finds itself in right now: it takes literally zero effort to be racist or sexist.
The system is already set up in such a way that it lends itself to inequality. All we have to do is go with the flow and we become part of the problem.
“If you choose not to decide, you still have made a choice.“ — Rush, Freewill
Which brings us back to the question I focused on in this research: What is the next step to move diversity forward?
It’s clear that the current ecosystem of talent in the Bay Area and other tech hubs is not helping companies make good decisions. So beyond just how one company can fix their hiring, I wanted to think about:
- How can we counteract this “ecosystem” effect in our companies’ hiring today?
- And moreover, how can we change that ecosystem for the future so that companies maybe could achieve racial and gender diversity more organically? (I like to think of this as the “How can we make this lazy-idiot-proof?” line of product thinking.)
If we’re talking about changing the overall demographic mix of professionals in the tech industry, that means creating more opportunity for young people of color early on in their careers, so there are more senior people of color to hire going forward.
Furthermore, because the demographics of the industry are already so un-diverse today, we need to swing the hiring processes way past “equal representation” to actually affect this change. So I chose not to look “low-throughput” options that are unlikely to effect this significant demographic change. For example, things like unconscious bias training that focus on how we can create fairer representation in our current hiring processes or with our current pipelines were thrown out early on for being low impact.
Instead, I started from a world where we focus the majority of our early-career hiring efforts on bringing people of color into the industry, and asked how would we target those efforts and make sure they created the outcomes we wanted? How could we make these efforts just as effective as our pipelines today at bringing in great talent and diverse talent at the same time?
The data shows there is the potential to put a real dent in the “pipeline problem” by looking outside the tiny subset of Ivy League+ schools.
What I found is that if Tech keeps recruiting from what I’ve dubbed the “Ivy League+” of top tier universities — the Ivy League and other Top Technical and Liberal Arts Schools like MIT, Stanford, and Berkeley— then the industry is correct about there being a “pipeline” problem, because those universities have their own salacious pipeline problem.
In 2017 the New York Times found that even with Affirmative Action and the college-age population in minority groups (i.e. the “Top of the Funnel” in startup parlance) increasing, the number of Black and Latinx students has DECLINED at top colleges over the last 35 years. Meanwhile Forbes covers why that’s unlikely to change despite top universities recognizing the issue, and how university rankings like US News and World Report are driven primarily by factors that reflect the wealth of attendees rather than the actual quality of the school.
There are over 7000 institutions of higher learning in the US, over 3000 Computer Science programs, and MANY of those programs have achieved racial representation or better!
The data shows there is the potential to put a real dent in the “pipeline problem” by looking outside the tiny subset of Ivy League+ schools.
Why the Ivy League won’t fix our diversity problem
Even if diversity had been getting slowly better at top schools, the overall demographics of these universities right now just make them a non-starter for any company that wants to make diversity a priority in recruiting.
Columbia University has the best diversity numbers of the Ivy League+ schools, so let’s use them as an example, to be especially generous.
Columbia University
Claim to Fame: Most diverse of the Ivy League Schools with a 7.6% Black and 13.1% Latinx student body overall.
5% and 10% of Columbia’s 160 CS degrees awarded in the school year ending in spring 2016 went to Black and Latinx students respectively. That’s drop-off from student body diversity to CS degree diversity is about average across Ivy League schools, so Columbia’s is still the best program to consider. Only a 1/3rd of CS degrees were awarded to women, but again, that’s about average.
Assuming this is a typical year for Columbia, that means the number of diverse graduates that would be available to recruit from this school in any given year looks something like this:
8 Black CS Graduates, 2–3 of whom are women
16 Latinx Graduates, ~5 of whom are women
For a total of 24 grads
If we extrapolate this across the pool of Ivy League and Other Top CS Universities that’s a pool of about 300 new diversity grads a year from these schools. That’s a “max” number because remember, the diversity numbers at most of these schools are significantly worse.
Moreover, diversity is one of the hardest aspects to target for early in the hiring process, which makes it one of the more “expensive” aspects in terms of recruiting time to fill a pipeline for. When recruiting from these schools we’ve got less than a 1 in 5 chance of pulling in a diverse candidate to our pipeline. Less than an 1 in 10 chance of getting a woman of color into our pipeline.
Remember it was only in the last year or two that Google expanded the scope of schools it was recruiting from. So when we are looking for more experienced people and targeting prior employees of Google, Apple, Uber and Facebook, we’re basically trying to wring diversity out of that same tiny funnel 2–5 years down the line. These companies’ numbers on diversity are even worse than the Ivy League+’s. So our odds of landing diverse candidates only gets worse and more “expensive” as we start trying to capture more senior talent, unless we grow that talent ourselves.
Simply by picking these schools or companies to focus on in our search or recruit directly from, we are making it “hard” and “expensive” to create a diverse hiring pipeline. We are putting diversity at the bottom of our priority list whether we mean to or not. To make it “easier” and “less expensive”, will require changing some of the fundamental assumptions baked into our thinking.
If we really want to move the numbers on diversity now and in the long-term, the easiest way is to focus on new university grads and growing them into leaders in the future.
By focusing on recruiting from schools that are more likely to produce diverse CS graduates, companies can both improve their diversity today as well as expand the funnel of senior talent in the future, and do it at a lower “cost” of recruiters time and effort per candidate.
Proposing: The D&I Shortlist of Universities
Why aren’t more companies focused on more diverse universities then? In my conversations with recruiters and hiring managers, it’s because they don’t know where to start. Again, they are mostly white people, that went to mostly white schools and…aren’t really familiar with anything outside of that. The Ivy League+ have brand recognition, and even though there are over 7000 institutions of higher learning in the US, plenty of them aren’t any more diverse than the Ivy League+.
Even if we can turn down our snobbery about top schools for a hot minute, there’s still the difficulty of putting aside the confidence-inspiring certainty of the U.S. New & World Report rankings, and figuring out how to find these great diverse schools instead.
How do you even know where to start?
I propose creating an alternate list of “name brand” schools focused on which universities are best if you are looking to recruit diverse candidates. I’ve taken a first stab at what this might look like focused on Computer Science degree programs, and Black and Latinx students. Even within this narrowed scope this list is by no means exhaustive! I only started with a short list of 175 schools from out of the 7,000 in the US: You can view the full sheet of schools I looked at and the underlying data here.
This is a great list to get us started though, not only in being more aware of diverse schools, but understanding how diversity breaks down at the collegiate level so we can focus recruiting programs to achieve specifics kinds of diversity. In addition to over 90 schools recommended for diversity recruiting efforts, the model also includes 25 Ivy League+ schools to show how their numbers compare to diverse schools and how they perform in the model.
How the Model Works
Schools in the sheet appear ranked by their Diversity Recruiting Score. The Diversity Recruiting Score is a factor of the school’s Diversity Score and the number of CS degrees it graduates.
Diversity Score
The “Diversity Score” is a metric I totally just made up! Woohoo!
That being said, it is grounded in real life data points relevant to diversity including racial demographics nationally for the US, each state, for the school’s student body overall, and the school’s CS program specifically.
The weighting is as follows:
- 1 point for having more Black CS grads than the national demographic average (>13%)
- 1 point for having more Latinx CS grads than the national demographic average (>17%)
- 1/2 a point for having a more Black student body than the national demographic average (>13%)
- 1/2 a point for having a more Latinx student body than the national demographic average (>17%)
- 1/2 a point for having a more Black student body than the demographic average of the state the school is located in (varies see “Demographic Matching Data” tab in the sheet)
- 1/2 a point for having a more Latinx student body than the demographic average of the state the school is located in (varies see “Demographic Matching Data” tab in the sheet)
- 1/2 a point for having a graduating CS class of 40% women or more
For a total of potentially 4.5 Diversity Points.
The green cells on the left side of the Diversity Recruiting Score indicated which metrics the school’s points come from. Green means the schools met the bar to earn the point for that metric. (There’s some additional metrics to think about to the right of the diversity recruiting score that aren’t included in the model at all, but are also color coded for some additional clarity.)
# of CS Degrees
The schools are then ranked by the number of CS degrees they graduate every year, with the idea being that schools that graduate 100 CS students and are very diverse are better (read “more efficient”) to recruit from than diverse schools that only graduate 20 CS students a year. Ranking the schools rather than using the raw graduation numbers creates some normalization to prevent schools that just graduate a TON of CS students from getting too much of a bump. (Note that it is an inverse ranking so the bigger the number, the better.)
Diversity Recruiting Score
This rank by number of CS degrees is then multiplied by the Diversity Score to get a final Diversity Recruiting Score for the school.
CS Program Rankings
You may also notice that the data includes the national ranking of the CS program, but I’m not actually using that ranking at all when scoring the schools. This is because I think recruiting and interview processes already over index on this type of information and the point of this research is to provide an alternative to that.
However, I know CS program rankings are a relevant number that people using this data would wonder about and be interested in. Consider this my single capitulation to the stupid “lowering the bar” argument. I think every company has a different bar where they would draw the line and say “We want to look at diverse schools as long as they are in the top 50% of CS programs,” whereas another company might say, “Focus on diverse schools only as long as they are in the top 10% of CS programs”. So the data is there and users are welcome to incorporate it into their decision making more or less as they see fit.
But the model’s opinion is that it doesn’t matter at all. :)
Some Interesting Things of Note Coming Out of this Research
Beyond just a list of potential schools to get started with, there’s several patterns that showed up in this research that are helpful to keep in mind when figuring out how to untangle the diversity pipeline problem for your company.
School diversity does not = CS program diversity
Schools can occasionally have significant drop-off of diversity between their student body numbers and their CS program graduation numbers.
Xavier University of Louisiana in New Orleans for example has a student body that is almost 80% Black but only 40% of it’s CS degrees go to Black students — a drop off in representation of over 36%!!! Recruiting from this school is likely to not result in the diverse pipeline you hoped for.
This also highlights how important it is, if you are going to expand this model for other roles, to specifically use numbers for the degrees programs you want to hire out of, instead of what I’ve included here.
I was really impressed with how comprehensive and specific the IPEDS data set is, and how easy it was to pull the data even just using their manual interface. It only took me an hour to pull program specific data for 125 schools, so it’s totally worth the effort to get specific.
I’m sure with a little more research on the different way to get data out of IPEDS, the process could be more automated, or potentially even integrated into your recruiting tools.
Schools are rarely diverse on more than 1 axis, meaning there’s no silver bullet
As you can see just by running down the list of top 25 schools, schools tend to highly represent either Black or Latinx students, but rarely both. This means that there isn’t one school a company can focus on and still right its diversity numbers. It’s going to take a coalition of schools to create a truly diverse pipeline.
Additionally, the percent of women represented across all of these engineering programs is currently still very problematic, with a few exceptions (Scroll down to find some gems if you are looking to focus on recruiting minority women! Alabama A&M and Spelman College, for example). However this is a known issue and these schools are focusing on gender diversity in their engineering programs just as much as other schools, so don’t use that as a knock against them until you can prove your less diverse schools have a significantly better record.
But let’s be honest, no matter what we first try to get started, we were going to find some problems or things we didn’t consider. By starting from a data-driven approach, at least we can go in with realistic expectations, be able to accurately identify the source of pipeline issues, and the different program demographics that are likely to move the needle.
Why are there more strong Latinx schools at the top of the list?
Most Importantly: This is not an indicator of quality at all. If you’re into CS program rankings it’s actually the opposite. So what’s going on here?
Overall the number of diverse schools in the list is about equal for strong numbers of Black and Latinx students(34 and 36 schools). Strong Black schools tend to be lower on the list because they graduate less CS students, not because of their quality or their overall diversity score.
It’s pretty clear from the list of schools that Latinx schools dominate the top of the list due to a couple demographic coincidences, and also, you know, racism.
Texas and California have excellent public collegiate systems, in close proximity to commercial tech powerhouses, and both states also have a much larger than average populations of Latinx.
California in particular has a huge public collegiate system and has done a lot to make college accessible to the widest variety of people, but the state actually has a much smaller Black population than the national average. Therefore it’s not surprising to see that diverse schools that also graduate large classes of CS students tend to spike on the population of Latinx students.
Due to historical segregation and continuing systemic racism in the collegiate system, strong Black schools tend to be HBCUs, or Historically Black Colleges and Universities. These schools tend to have smaller engineering programs, because even through the 60s and 70s — less than 60 years ago, less years than my parents are old — most companies would not hire Black engineers.
Really, the imbalance in the list only stresses the importance of the work we are talking about here. Creating pipelines so Black engineering students have access to opportunities is something we in the industry can do — now, today — to counteract the results of the overt, systematic anti-black racism that plays such a strong role in our country’s history.
How to Use This List of Schools
Recommended Use Cases
I think there are 3 main places this information could have an impact.
- Bake them into your outbound recruiting search.
- Bake them into your inbound resume parser.
- Reach out to those universities and recruit directly. Go onsite if possible.
I think number 3 is the most effective and impactful, but I understand that it also take the most effort. I personally think there are zero reasons why Bay Area tech companies shouldn’t drop everything right now and be driving to UC-Domingues Hills or Santa Cruz to recruit and set up a partnership. Or SUNY Lehman College or Rutgers-Newark if you’re based in the NYC area. But maybe I’m oversimplifying things.
At the very least, it seems worth the effort to have someone reach out to those programs and get on their job boards.
However, most recruiting software systems or resume parsers now let you set preferences for particular companies, schools, skills, etc. that show up on someone’s background and it takes, like, 5 minutes to update those settings. So there’s literally no reason not to give options 1 and 2 a try.
The sooner you give it a try, the sooner you can start learning what doesn’t quite work and how you can refine the diversity hiring process from there!
What’s in it for you?
Even if you give exactly zero fucks about diversity at your company, if you are involved with hiring at all, you’re eyes should be getting all sparkly about this information.
Your ability to recruit effective talent for your company is potentially only next in line to “product-market fit” in the Maslow’s Hierarchy of your company not crashing and burning. I’d put recruiting ability even above “not being an organizational shitstorm”, because I happen to know from personal experience that hiring enough good people can overcome any amount of assholery or incompetence at the top.
And recruiting is HARD right now. There are more engineering and product management positions open in the Bay Area than there are qualified people to fill them. It’s a blood bath for medium-sized companies competing against each other right now, with everyone competing for the same new grads from the same name brand schools, or for veterans of the same “winners & unicorns” pool of startup…which also have a history of only recruiting from those name brand schools….
The company that can innovate on recruiting is going to have a significant advantage over it’s counterparts. To do that companies must find ways to access and identify the amazing amounts of latent talent outside that tiny Ivy League+ Thunderdome all the other companies are competing in.
This model is just pointing out one way to start doing that.
Wider Implications
I hope this article makes you believe in the power of DOING SOMETHING DIFFERENT in your recruiting process, and that it can create a real impact on diversity at your company. Although it will take effort, the data shows the effort will definitely be worth it.
I hope it makes you want to reach out to amazing organizations like Code 2040, because maybe you don’t have the time to create individual relationships with these universities, but you’ll provide the opportunities if someone can help connect you with the interns.
I hope it makes you realize that data matters and what gets measured gets done. Perhaps you’ll even look into programs like Project Include that not only create outside accountability for D&I goals, but also are working to come up with new, actually effective ways to create D&I it at tech companies.
If nothing else, I hope this research gets a conversation started in people’s minds. That it helps anyone involved in hiring — recruiters, hiring managers, interviewers — double check their assumptions about name brand schools or name brand startups, and the implications about a person’s technical background or abilities. Don’t let your unconscious biases hide in plain sight.
I hope this article opens the door to accepting that until the Ivy League+ schools fix their own pipeline problem, being stuck on name brand schools in your hiring, is participating in systemic racism. And that’s true whether we mean it to be or not, because that’s how systemic racism works. We have to recognize these sometimes seemingly unrelated assumptions must be changed if we really want to move diversity forward in this industry. Otherwise we’re just talking the talk with no walk what so ever.
I hope this research helps to start raising the profile of a lot of these great university’s that have had diversity figured out for years before most of us in the general tech community even knew we needed to start caring about that. In 2016 Bloomberg wrote about Howard University’s stalled efforts to get tech companies to hire their engineering grads. Less than 18 months later Google opened a “Howard West” campus extension on the Googleplex campus. Awareness matters.
At the end of the day, the least you can do is read over this list 3 or 4 times and get the names of these colleges stuck in your head next to the “good schools” label. Then next time you see someone’s resume from Spelman College, you won’t be like “Whatever, I’ve never even heard of that place.”
Instead maybe you’ll be like, “Yeah, I think I remember that from somewhere. Some list of great CS schools, like Cornell and MIT or something? We should talk to this person!”
And here’s to your next class of incoming new grad hires looking like this:
Appendix A: A Review of This Model and It’s Potential Problems
Use of Latinx, Race and Racism versus Hispanic, Ethnicity and Ethnic Discrimination
“Hispanic” is technically an ethnicity, not a race per the U.S. Census bureau. People of Hispanic origin in the U.S.primarily identify as racially White. I’m using race and racism throughout this piece rather than more academically correct or combination terms like “racial and ethnic discrimination” for simplicity and ease of reading.
Most of the demographic data used in this research does use the updated “Hispanic or Latino” to capture ethnicity data. Latinx seems most commonly used by D&I organizations like Code 2040 when referring to this group, so again for brevity’s sake I’ve used Latinx as well.
Why focus on Black and Latinx demographics only?
This was primarily based on the fact that I wanted to try something and get it out there, so I focused on the US’s two largest racial minority groups because I expected it would be easier to find data and college recommendations for them. Initially, when I made the first model based on a lot more vague, proxy data, that was totally true. There even seemed to be significantly more college recommendations resources for Black students than for Latinx students.
In the end, the data set I settled on using includes consistent data across demographic groups (woohoo!) and this model could easily be expanded to take Native Americans, Pacific Islanders, and Asians into account when targeting diversity hiring.
I hope the effort to get something started that others can fork and improve on makes up for any incorrect assumptions here on my part!
Availability Bias
I spent 5 evenings on this work — about 20 hours in total on the original model. I knew almost nothing about it when I started so most of that time was spent trying to even find sources for the kind of data I was looking for, followed by coercing the raw numbers into things that seemed like useful indicators of diversity.
In the end I’m pretty happy with completeness and specificity of the IPEDS dataset from the National Center for Education Statistics for the institutions I ended up looking at, but since I came up with the list of institutions and then found the IPEDS data set much later, there’s a huge availability bias in the list. There are very likely diverse institutions that ought to be represented here and aren’t!
Inappropriate Dataset
Initially, I was using mass media college rankings and student body overall diversity demographics as proxies for CS program quality and CS program diversity. With the IPEDS data I was able to get to a much more accurate indication of the diversity of the Computer Science programs specifically. I’m pretty happy with that, and based on my own knowledge of diversity don’t know off hand how I could improve the dataset.
The problem with mass media rankings is that they only give the top 25–100 schools, which creates a winner take all situation where the top 0.5% of schools get all the points and there’s no way to differentiate between all the rest of the schools. The rankings also tend to be pretty vague, as in there are rankings for “Top Colleges for Black Students” and “Top CS Programs”, but not “Top CS Programs for Black Students”.
Computer-Science-Schools.com is a site I’m not familiar with outside of this research and so can’t vouch for it’s accuracy or methodology. However, I chose it for it’s breadth of representation of programs.This sites rankings were fairly similar to the mass media rankings for the top schools, but then included rankings for all known CS programs (over 3000).
Bad Algorithm
I’m somewhat arbitrarily picking things to “rate” on based on what I think represents succeeding at Diversity at the collegiate level. I could be wrong about those indicators or not being nuanced enough in my weighting.
Mostly ignores the largest minority: Women
For the most part I’m not focusing on women as a minority group in this work — just that 1/2 a point for graduating more than 40% women from the CS program.
This is a personal choice because schools with decent rates of female CS graduates tend to just mirror the list of overall top CS programs, thereby giving that Ivy League+ list of schools an advantage.
I have big concerns about the focus on women in tech as the solution to diversity. I worry that in order to avoid actually solving the diversity problem Tech will just fill the ranks with white women in order to shut down the conversation. Erica Joy’s perspective on the early diversity in Tech efforts explains much better than I ever could:
I think the Ivy League schools having better numbers of women than the average CS program, but significantly worse numbers on racial diversity are an example of this. I think you can also see this backed up in tech diversity numbers from several companies over the last two to three years where all the companies touting their D&I wins refer to the number of women in Technical and Leadership increasing. If you dig into the numbers though there hasn’t been a commensurate increase in the numbers for racial diversity, which leads me to conclude that the improvements aren’t equally available to women of all races.
Example from Slack’s 2016–2017 data:
The number of women in technical roles increased 5.3 percentage points between 2016 and 2017, but Blacks in technical roles only increase .5 percentage point (when using the apples-to-apples comparison data that normalize the difference caused by methodological changes between these years).If both populations were increasing commensurate to their representation in the general population, we would have expected to see a 1.4 percentage point increase in the number of Black members of the technical workforce at the same time we saw a 5.3% increase in the number of women. (Based on Slack’s 2017 Diversity numbers)
Gusto directly admitted this about their hiring efforts and are now performing the same focused recruiting process for racial minority candidates. But not every company is that honest, and even Gusto took 2 years off in the middle there to rest on their laurels before getting concerned about racial diversity.
In my own experience over the last few years, focusing on one form of diversity results in a pretty harmful reductionist take on the entire problem that just pits one group against another. The real problem can not be “how to stop mansplaining” or other overt manifestations of misogyny, because the underlying biases that some people are inherently better than others will just find another way to manifest. It’s about that internal battle in each of us to recognize the uncomfortable truth that we all have some kind of these unfair discriminatory biases: we all prefer to be around people who are like us because that feels easier, and we all need to go through the process of learning to be comfortable being uncomfortable so we can expand beyond that if we actually want to fix this problem.
We all must learn to be more empathetic, compassionate, and able to understand people who aren’t necessarily like us or don’t see the world through the same lenses that we do. And I think only a “preponderance of diversity” brought on by adding lots of different kinds of new voices, rather than a lot of one new voice, can solve this problem, because then there’s sort of nowhere to run and hide. Then being uncomfortable and around people who are different than you is something that you can’t escape and still get you job done. It puts us all on the hook to improve in a measure commensurate to the privilege we wield in the wider world.
So to wrap up this tangent: Because of my personal concerns about how the available women-focused data skewed the results back toward the Ivy Leagues+ schools, this model only focuses on racial diversity.
Why I think it’s good enough anyways
As a product manager I know the data is never going to be perfect — it can never GIVE you the answer. Data is always directional and you have to know when to stop trying to dig up more or better data, and instead start taking action and iterating in the direction that the data indicates — start learning by doing and generating your own data.
At this point my product instinct is telling me this is a good enough place to start: That it will cost less effort and we’ll get higher fidelity answers about what’s actually broken about this model or what other kinds of data we ought to use if we just start trying do something based on what this current data says and track the results.
