As an Economics PhD student at Stanford, I spent years hunting for the natural experiment that would become my job market paper. From my cubicle in Palo Alto I had unrivaled access to top scholars — but almost no visibility into what was happening in the real world. So I scoured newspapers for stories about new programs and policy changes, and cold-emailed the people behind them, hoping one might share data or a foothold for a study. Mostly I hit dead ends and started over. (My lucky break came from a news story about state variation in DNA database laws.) It’s an agonizingly inefficient process, and it’s how the third, fourth, and fifth years of a PhD tend to go — until the stars align, or you give up and drop out.
Fast forward a decade: In the summer of 2023, I joined Arnold Ventures, and my inbox was overflowing. Information that had been rare and precious in academic circles — innovative programs, policy shifts, warm introductions to the mayors and nonprofit leaders and state senators running them — was suddenly everywhere. My policy colleagues found this completely unremarkable, and wondered aloud why rigorous studies of all these interesting interventions weren’t getting done. Did academics not know what the open questions were? They suggested we write detailed research agendas to point the way.
I assured them that wasn’t the constraint. Academics know the open questions perfectly well. What they lack is the foothold — the access and the trust — to answer them. Researchers don’t know practitioners, and practitioners don’t know researchers, and neither side trusts the other. The binding constraint on better research isn’t funding or awareness. It’s relationships. Closing that gap became my primary goal at Arnold Ventures, and I’m now more convinced than ever that a funder’s convening power is the most powerful tool in the kit.
Networking events can, of course, be terrible — we’ve all endured the lukewarm hors d’oeuvres and harsh fluorescent lighting. I’ve always loved the relationship-building side of conferences (it became a running joke that colleagues would find me schmoozing at the hotel bar rather than in the sessions), but even I’d rather visit the dentist than face the standard failure modes: hunting for a seat or a conversation in a sea of people who all seem to know each other already, endless small talk that never becomes real connection, and one person droning on while everyone else scrolls their phones.
And there’s a subtler risk on the funder side. Even if we’re confident we throw a warm, welcoming party, we may just be convening the same people again and again. Pleasant, yes — but if everyone already knows each other, what’s the point? Was that open bar really the best use of limited charitable dollars?
I knew early on that events would be central to my strategy for facilitating more and better research. I wanted to host the kind of gathering that would have been useful to me as a PhD student — one that produced collaborations practitioners valued too, and built connections across siloed networks that could pay off in both the short and long term. I didn’t see anything like it in the field at the time, and I had a hunch it would matter. Our two largest bets along these lines were BRIDGE Days and Innovation Days. I’ll take each in turn.
We hosted BRIDGE (Building Research IDeas and Generating Evidence) Days roughly every other week, beginning in spring 2024. Between May 2024 and December 2025 we held 32 of them — the only weather cancellation being, improbably, a Houston snowstorm (not on my bingo card!). The events are ongoing, but the analysis below covers that window.
Each event brought 12 researchers and 12 practitioners together for one day, focused on a single topic where our team saw an urgent need for causal evidence. We ranged widely: DNA in investigations, supporting victims of violent crime, employer insurance to spur hiring of people with records, electronic monitoring, retail theft, increasing the reporting of sexual assault, scaling mental health care, and lead pipe replacement, among others. Seating was assigned to mix researchers and practitioners at every table, so no one could simply cluster with people they already knew — though they could mingle freely at the closing happy hour.
The day opened with a short primer: what we know in the space (usually very little), and how economists use randomized trials and natural experiments to separate correlation from causation. A practitioner panel followed, surfacing the challenges and novel approaches most on their minds. Those panels always felt too short by design — just enough to show the depth of expertise in the room and get researchers’ minds working. The rest of the day ran on facilitated conversation and breakout groups, each ideally landing on a specific intervention and a way to implement it that would support a rigorous evaluation.
The trust gap was real at the start. Practitioners wanted their expertise respected and had often been burned by overconfident academics in the past; researchers worried practitioners cared more about ideology than evidence. Watching that distrust dissolve over the course of a breakout group was the best part — our carefully curated attendees met the best version of the “other side,” and their sense of what was possible shifted. We closed each day by pointing people toward our standing RFP, emphasizing that strong causal research was the first step to unlocking AV funding to scale what works.
These events were a major logistical lift, but I made the bet deliberately, for three reasons:
Control. Owning the content and guest list let us structure each day around our own goals.
Convening power. A funder’s invitation rarely goes unanswered. We used that pull to get busy, high-value people — top experts and decision-makers — into the room.
Compounding knowledge. The work of deciding whom to invite and what to discuss deepened our own grasp of the issues and players, making us a sharper resource to our networks long after the event ended.
Even 20+ events a year left many interesting stones unturned — and biweekly was already stretching our internal capacity. Could we fund external partners to run their own events, again aimed at producing proposals for our RFP? I expected a lower average return, but with no internal staff time required, a lower cost too — so plausibly still worth it.
So we posted an open RFP for Innovation Days (a different name, to distinguish them from our internal BRIDGE Days). We received far more proposals than we could fund, and ultimately backed about 30 events across 2024 and 2025 on a wide range of topics. We helped organizers broaden their networks — suggesting practitioners for researchers to invite, and vice versa, with introductions on offer — but were otherwise hands-off on agendas and design. Maybe some partners would build better events than we could! I wanted to encourage experimentation.
These events promised several benefits: warming practitioners to impact evaluation, forging long-term researcher–practitioner ties, seeding a culture of experimentation, and pulling junior scholars into the public safety field. But our primary goal was concrete — generating research collaborations that became viable projects — and those projects would land with us as funding proposals. (We practically begged participants to submit letters of interest.) So our first-order outcome was easy to name: the number of high-quality proposals submitted.
Topics and participants weren’t random, but event timing essentially was — a function of when an idea struck us, presenters’ schedules, and room availability. We exploited that as-if-random timing in a stacked difference-in-differences design: for each event, we compared attendees’ proposal submissions before versus after, using people slated for future events as the not-yet-treated control group. That controls for underlying time trends — academics submit more in June after the spring term, say, which would otherwise inflate the apparent effect of a May event. Stacking across all events averaged the effect over the full set.
The contrast was stark. BRIDGE Days produced about 2 additional high-quality submissions per event in the 10 weeks afterward (and similar results out to 30 weeks). Innovation Days produced a precise null — completely ineffective at generating the collaborations we’d hoped for, at least in the short run.
Did Innovation Days have no value at all? Not necessarily. But new research projects were the point, so we pivoted: we pulled the Innovation Day RFP and redirected the resources. We took our own oft-repeated advice to fail fast.
This is a small slice of my team’s convening strategy, and to be clear, we were just trying stuff — none of us are professional event planners. BRIDGE was our first, best attempt, and surely improvable. But it gave us a benchmark: in its current form, generating one new high-quality research collaboration cost about $35,000. That figure lets us compare BRIDGE against other events and partnerships we might try — keeping our eye on testing ideas and finding scalable solutions, rather than rationalizing ineffective strategies. (People told us they loved the Innovation Days, which might have given us an excuse to keep them going.)
I remain convinced that the binding constraint on progress, from research to policy, is relationships. We all live in bubbles, and getting out of them — talking to people with different vantage points and pressures — helps on many fronts. Here it serves a very practical end: helping researchers and practitioners share what they know and find ways to work together. They often share the big goal (safer communities) but bring different strengths (research methods, institutional knowledge) and face different short-term pressures (publishing, budgets, reelection). Showing that we can and want to help each other is how trust gets built — and trust is what leads to more collaboration.
Still, not all convenings add value. I was genuinely surprised the Innovation Days returned nothing; I’d braced for a lower return, not zero. It’s a frustrating reminder that there’s a limit to what any funder can do in-house, and that scaling impact will take real trial and error. I’m eager to keep experimenting.
My hunch is that the standing RFP was a crucial complement to BRIDGE: attendees knew funding was waiting if they pursued an idea sparked at the event. Targeted, time-limited RFPs, by contrast, mostly serve as a roll call — they reveal who’s already working on a topic but rarely spur new work, which takes time and relationships scholars may not yet have. For the people most likely to chase a new idea — PhD students and junior scholars — real-world connections are the better catalyst. Money helps, but a connection that sparks an idea, which becomes a great paper, is far more catalytic on the research side. And catalytic investments are the whole ballgame in philanthropy.
The latest episode of Probable Causation features Juan Vargas, an economics professor at Collegio Carlo Alberto and the University of Turin. We talked about the unintended consequences of a Colombian policy announcement that promised to pay people to stop growing coca (the plants used to make cocaine), beginning on some future date. The goal was to incentivize people to stop growing that crop. The result? Farmers raced to start growing it, to increase their odds of receiving those future incentive payments. This one policy announcement undid decades of progress against the drug trade. Check out my conversation with Juan for more!
Over the past year, my team at Arnold Ventures has been working closely with Jonathan Coates and his London-based Indio Media team to produce two new podcasts. Our goal is to change the public conversation about public safety, to one focused on innovation and evidence over ideology. I wrote a LinkedIn post with a bit more of the backstory. (TL;DR — I met Jonathan at an event and he convinced me that AV should be a media company. Again, events can be catalytic!)
Please check out these new shows wherever you listen to podcasts; new episodes are posted on alternating Thursdsays. If you like them, help us spread the word by sharing with friends or leaving a review. I’m so proud of how these have turned out and hope you enjoy them.
Fighting Crime, hosted by Cristina Quinn. A narrative discussion of pressing problems in the public safety space, and new, innovative approaches that might solve them.
Episode 1: Why Don’t People Leaving Prison Get Jobs?
Episode 2: Little Scandinavia: An American Prison Experiment
Fighting Crime: Field Notes, hosted by me, Julie James, and Kevin Ring. Episodes feature long-form interviews with the most interesting people in today’s public safety arena.
Episode 1: Fixing Broken Prisons, with Mike Thompson
Episode 2: How Lead Exposure in Children can be Fixed, with Robbie Barbaro
Jonathan Wrobleski wrote a thought-provoking review of my new book in his Sentencing Matters Substack.
Crucially, Doleac tells us, the economist would insist on treating their own ideas as mere hypotheses, not moral imperatives or self-evident truths. One of the central themes of Second Chances is that criminal justice policy making can look more like a science lab and less like a press or political office, and when it does, the results will often be less crime at lower cost. Doleac walks through example after example in which policies that sounded tough or compassionate or “common sense” turned out, once tested, to have no effect on crime — or even made things worse. She shows how economists use randomized trials and natural experiments — staggered rollouts, eligibility cutoffs, geographic borders — to separate what we think a policy does from what it actually does in the world. The goal is sound policy: less crime, more justice, at less cost.
And I love his vision of an iterative policy-making process that mirrors regular tech improvements:
This is where my mind wonders back to the Apple developers conference. Each year, Apple doesn’t just unveil shiny hardware; it releases a new version of its operating systems. There’s always a list of changes: bugs fixed, security holes patched, features added, old code eliminated. Some updates are big, some are minor, but the assumption is that the system is never finished. The job is to keep iterating.
Doleac helps us imagine the equivalent of an annual “Justice Developers Conference.” It could be a gathering not just of criminologists or economists — those gatherings actually already take place now — but one where DOJ, other federal and state policymakers, sentencing commissions from across the country, legislators and their staff, and others gather with economists and criminologists not to announce the latest crackdown, but to discuss research and roll out “Criminal Justice 17.0” — a set of evidence‑based updates:
This diversion program was tested in three cities and cut new charges in half for eligible defendants, so we’re expanding it.
This sentencing enhancement for commonplace technology no longer explains any variation in risk, so we’re retiring it.
This reentry program that sounded great on paper didn’t move rearrest rates at all, so we’re sunsetting it and reallocating funds.
Read more here:
No posts
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.