Dallas Salisbury of the Consumer Policy Center breaks down why unbiased retirement research matters—and what it looks like in practice. In this episode, Dallas explains how retirement policy and benefit decisions often get made with incomplete information, and why relying on objective, evidence-first analysis is so critical when the stakes affect millions of workers and retirees.
Jeffrey Snyder, Broadcast Retirement Network
Joining me now is Dallas Salisbury.
He’s a senior fellow at the Consumer Policy Institute, and he’s also the founder of the Employee Benefits Research Institute. Dallas, it’s always great to see you. Thanks for joining us on the program this morning.
Dallas Salisbury, Consumer Policy Center
Wonderful to see you and wonderful to be here.
Jeffrey Snyder, Broadcast Retirement Network
Yeah, as I was recounting to you, I remember seeing you back in 2008 at the SPARC Institute’s annual forum. You were talking about all the great data that you collected. We’re going to get into that in a second, but you have a new gig now.
You’re the senior fellow at the Consumer Policy Institute. What’s your focus now? Is it still retirement and employee benefits?
Dallas Salisbury, Consumer Policy Center
It’s all on employee benefits, personal financial planning, and a component on caregiving and long-term care, nursing homes, assisted living facilities.
Jeffrey Snyder, Broadcast Retirement Network
A lot. I know Social Security has been top of mind for you in the Institute as well.
Dallas Salisbury, Consumer Policy Center
Yes, it has. In fact, the first piece that I have put up on the consumerpolicy.org website is one on Social Security. I have them in process.
Sometime in the next couple of months, we’ll post them on the implications over time of leakage from the retirement system and another looking at flexibility in the retirement system.
Jeffrey Snyder, Broadcast Retirement Network
Very exciting. I look forward to maybe talking to you about it somewhere down the line. Great.
Dallas, the reason why I’d reached out to you is I’m very impressed by the work that EBRI does. I have always been. It’s true research in employee benefits.
I wonder if we could go back in time a little bit. Before the founding of EBRI, was there any research around employee benefits that existed?
Dallas Salisbury, Consumer Policy Center
There was data, but not a whole lot of research. The research that was being done was almost entirely out of the government. In fact, the Social Security Administration created a Bureau and Office of Research and Statistics in 1935.
Then in 1963, they introduced a survey of the aged. In 1969, a retirement history study. Then all of that morphed a bit in the 60s there.
President Kennedy had a Kennedy Committee from 62 to 65, and that really pushed the Social Security Administration to do some early research. Kolodribitz, Reno, Woodruff, and others were publishing very broad data set articles on the coverage of the retirement system and all. There was very little of that in the health area.
Which came later. BLS didn’t introduce its employee benefits survey until 1979. Actuarial firms kept proprietary data, but academic work was very sporadic.
The American Enterprise Institute did some work around ERISA funded by Ford Motor, which was a corporate leader in the ERISA debates. That gap was really the reason EBRI was created in 1978. Post ERISA, SSA funded real university consortiums beginning in 1998 through 2019, but that funding was cut entirely in 2025.
Multiple centers have closed as a result, and there is a remaining center at Boston College with independent funding. But ironically, the web that was created post-tax changes and PBGC changes in the 80s, 90s, and 2000s, most of that has now disappeared again. At a point in time where research on employee benefits is probably more important than ever, the Employee Benefit Research Institute is back to being one of the only entities in the private sector that does it day in, day out, week in, week out.
Jeffrey Snyder, Broadcast Retirement Network
Yeah. And Barb Marder and Bridget Bearden and the team, they’re doing an exceptional job building off of what you did. I want to give them full credit.
How important is this independent research to inform? You mentioned the consortiums that were established, the research within, I think, Social Security you mentioned, but how important is this research to informing policymakers, other government entities within the federal and state governments, as well as the retirement industry, and most importantly, I guess, clients, plan sponsors. How important is it?
Dallas Salisbury, Consumer Policy Center
It’s foundational. All three make high-stakes decisions on incomplete information, policymakers, industry, employers, unions, you name it, not to speak of individuals. Policymakers need pre-legislation evidence and not post hoc anecdotes.
Industry needs benchmarking. Smaller employers especially lack in-house expertise, and they rely on their own vendors. But an example is EBRI produced its first research and published its first book on leakage and lump sum distributions in the early 1980s.
The first piece on funding retiree health care was in 1983. And at the front end of EBRI for the first decade, there was almost nothing else out there. And then you talk about policymaker impact is EBRI was heavily relied upon in the early 1980s, in particular, by the Carter Commission on Retirement Policy, Pension Policy.
It played a key role in testimony and doing data summaries for the Finance Committee and the Ways and Means Committee in the early 80s and creation of IRAs. And then in basically, frankly, the chairman of the Finance Committee saying that it was research from EBRI that preserved 401k plans in the 1986 Tax Act. What was unique was we never said in any of those studies, this is what this says you should do.
The data was the data, the tables were the tables, but they were convincing with regard to the comparative value and use of IRAs, where the Reagan administration in Treasury one proposed eliminating defined contribution plans and replacing them with 7,500 a year contribution, individual retirement accounts without an entitlement to employment. And we now know that that would have been relative to trillions of dollars in assets accumulated, likely a devastating change.
Jeffrey Snyder, Broadcast Retirement Network
Yeah. Well, someone who came after 1986 into the retirement industry, I’m thankful that 401k, 403b, those code sections were preserved. Let me go back on something you mentioned.
How important is unbiased research? Because you mentioned that EBRI’s purpose, sole purpose was to provide the information and let policymakers make their decisions based on the data. How important is that unbiased research?
Dallas Salisbury, Consumer Policy Center
I think it’s the whole ball game. It’s perceived bias gets research discounted on both sides of the aisle, but in every category, industry as well, unions as well. EBRI’s founding principles, essential economic function, ongoing need for objective information, disinterested research beat advocacy in many of these cases.
No lobbying, no legislative positions, let’s competing interests trust the same numbers. And I compare this to many of the academic research entities, the National Bureau of Economic Research, which is now out has closed down its retirement unit because of the loss of social security funding. But it has produced periodic research of very high value analytically for decades.
But the authors of almost all of those studies include and here’s therefore what the solutions are. And by providing what the solutions are, you frequently end up cutting off many people that should read the study from even reading it because that the headline normally starts with that policy change. Bias can be unintentional, funding source, framing shape, conclusions, even without intent.
EBRI’s had to change a bit in that regard is we purposefully during the first 37 years, which was the years that I ran the organization, we were lucky enough and I’ll say lucky enough to have an extremely broad base of support that was employers, it was employee groups, many unions, many public pension funds, private pension funds, the pension funds of Chile and Japan and France and England. And it was at every range on every issue that EBRI did research on. Essentially within our funding base, there were people that totally disagreed with each other on all of those issues.
And basically in creating EBRI, they agreed to set aside all of that and just say, let’s do the numbers and see what the numbers show. And when you do it that way without lobbying and not pushing legislative positions, then the competing interests trust the same numbers and bias can be, as I said, unintentional. So it makes a huge difference.
Jeffrey Snyder, Broadcast Retirement Network
Yeah. And I would say that’s kind of where we’re in alignment with EBRI in that I don’t really have a goal objective other than information, education. I think that’s what people are looking for.
Just give me the facts and let me make the best decision. Let me ask you about the origin of numbers. You mentioned that, you talked about the history, the funding of social security for research like this.
Where do you get the numbers from? So when you’re creating the reports, obviously it’s based on some data. Where does that data set come from?
Dallas Salisbury, Consumer Policy Center
Well, in the early years of the Institute, and this would have been true for the Kennedy Committee, and it was to a significant degree true for the Carter Commission in 79, 80, 81. And the main data sources which have been public and replicable are from BLS. They’re the Fed’s Survey of Consumer Finance, the Census Bureau’s CPS, and SSA records themselves have always been very valuable both from the CMS side on retiree medical and on the retirement side from the Social Security Administration.
Then there’s survey data. And the survey data is always because it’s smaller sample than what the government is able to do. But EBRI’s retirement confidence survey is transparent.
It wasn’t in the initial phases sponsor proprietary for the fullest analysis. But that’s one of the changes EBRI was forced to make after I left and was part of a pivot or transition, which is they had to become far more reliant on per project funding dollars. We tried, we frankly tried to avoid that as long as we could.
It hit a wall. That wall came about interestingly because of industry consolidation. We were founded by 13 actuarial benefit consulting firms, only four of them remain in business.
We were supported by 12 very large group insurance companies that did health and retirement. Only two of them exist today. After the AT&T breakup, we had funding every year from every single AT&T breakup company.
So that was a good thing. But then it reconsolidated and now EBRI gets checks from only one of those companies. And because of some of the change in the financial services industry consolidation and the move from individual employers and to far more consolidation where essentially Vanguard Fidelity and Empower, but really now Fidelity and Empower totally dominate the record-keeping business or largely dominate the record-keeping business for both IRAs and individual retirement accounts.
There are others in the game, but that’s a shrinking universe each and every year as well. So EBRI from 1996 until 2016 had built a very large, robust 401k individual retirement account ESOP database, which was pre a lot of that consolidation. And that database was fed by most of the record keepers.
It allowed us to do something unique as we got the data from enough record keepers that as people moved from one job to another job, we could track them. And that so-called continuous participant database got up to nearly 10 million individuals, a huge sample with very, very disaggregated data. That database to a very significant degree, the continuous component of it has largely disappeared.
Many of those providers no longer are willing to participate. Why? Because they now have such as a consolidated source, they have such large individual databases themselves that they now publish once a year and they do constant analyses for their clients against the client base.
And if your client base is 22 million participants, then the fact that it’s only one provider and you happen to be a small provider, but they’ve got a couple of hundred thousand small plans from an employer perspective, you can rely on that data for your plan design decisions and other things. It does create a policymaking vacuum issue and a data access issue for many researchers, including researchers at EBRI to be blunt, because those databases, each of those entities now has pretty much created its own internal research institute. It’s hired its own PhDs.
They do the data, they publish the data, but the data is not generally available to feed into the policy process. And to be blunt, when Vanguard or Fidelity or Empower appear before a congressional committee presenting their data and presenting recommendations, I guess not unrealistically, the assumption is that they’ve calculated that what they’re advocating is what’s going to do the best for their bottom line and their assets under management, which after all is what consolidation is all about. Which stream you’re reading tells you the strength and limits of the findings and the broader the realm.
So, government data is still the critical source, including for much of EBRI’s work, most of EBRI’s work, because it is the broadest, most reliable and the research units within the government don’t get into the policy business. They don’t make policy recommendations. They leave that to others within the government.
It’s the general accountability office, the old general accounting office that does reviews of these things. And they keep publishing studies based on all of this data with very explicit policy recommendations. So, even GAO, when it goes and testifies as a government agency on the Hill, it has less cred than you will, than the same information being presented by a witness from the Social Security Administration or a witness from EBRI or the Boston Policy Center.
One other difference is the way in which many of these entities have reacted and interacted with policymakers in Capitol Hill. If you look at EBRI’s website, it has all of the old testimony. And what you’ll find is that during my 37 years, there was a very heavy focus on education of the public, education of policymakers, state, local, international.
You look at the history in the period since I left in 2017, it really very little focus on testimony, very little focus on those realms, and much more of an emphasis on short form presentation, infographics and all, which can be very valuable to reporters, can be very valuable to many public providers, many providers, employers and unions and others. But to be blunt, it’s much less useful to policymakers because they actually do generally want to dig into the data. One example of that is when Enron went bankrupt, and you’re still young enough to have remembered that.
Jeffrey Snyder, Broadcast Retirement Network
I remember Enron. Yeah, I do.
Dallas Salisbury, Consumer Policy Center
And that created an entire debate over employee stock owner plans and employer stock in these plans. And Jack Vanderhyde then of research of EBRI and for a very long period of time, the research director at EBRI, is Jack got databases, he got data from providers, he did extraordinary micro data analysis that was presented to the congressional committees on the actual role of stock ownership, the frequency, the outcomes, without any policy recommendations. But the result of that, one has to end up concluding for better or for worse, was fewer restrictions being added to plans and stock ownership plans as opposed to more being added.
So we weren’t seeking that outcome with that research. But in many cases, you can look at what didn’t happen as an indicator of how the research was, it was interpreted by the policymakers, or by others.
Jeffrey Snyder, Broadcast Retirement Network
Dallas, I want to follow up on something you mentioned. And that is, how do you distinguish between the unbiased research and maybe what I would think is quasi marketing, right? Where research is created, and it is tailored to, or at least it’s perceived that way that it is tailored to a specific outcome, because you kind of intimated that some of the bigger organizations have their own research institutes, which is not a, I’m not saying that’s good, bad or indifferent.
But a lot of times bias will creep in. So how do I as an independent consultant, or others that are independent, trying to act on the best interests of their clients, their fiduciaries, or their plan sponsor, how do you ensure that what you’re reviewing is either research, or has a hint of marketing, or could be both?
Dallas Salisbury, Consumer Policy Center
Well, let me answer it in two parts. One, the distinguish. Whose data is it?
Is it proprietary or a broad public industry dataset? In other words, is it what was and still is on a smaller scale, the EBRI 401k IRA HSA data, which is industry dataset? Or is it a specific provider?
Does the findings solution match something the sponsor sells? If I’m in the annuity business, and everything I put out says people should have annuities, you know, I have to question if it’s somebody that’s in favor of eliminating 401k plans, Teresa Gugliarci, and they put out a research study that concludes these are terrible, and you should get rid of them. But they’ve given 26 speeches saying that’s their conclusion before the research.
What do you do? The same is true on the other side of these issues. Is funding disclosed?
And did the sponsor control unflattering results? EBRI had a policy through the time I was there, that most research was published without the membership getting it first. Those funding got it last, not first.
They didn’t have a chance to comment and say you need to change this. And it was publicly available, broadly distributed. And through 19 to 2018, every single thing on the EBRI website was available without username, without password, to literally anyone, anytime.
I would get calls at 3am in the morning from some reporter or some congressional staffer when they were in a crisis mode. And I could lead them to the website to the precise issue brief and the precise data table that answered the question they had at this point to finish an article or to go back into the conference committee. That’s no longer there because of these restrictions that funding necessity, so to speak, have created for EBRI.
I don’t say that as a criticism, because I think EBRI being there is a very positive, good in an overall sense. But it’s a different world than we’re living in in that prior point. Is the methodology documented well enough to contest it or to contrast it to somebody else’s methodology?
Is the underlying data available for independent replication or is it just a narrative? Falls on a solution match, no data released together with functioning as a marketing. So I don’t give you the data.
I put very little data in the report. I just do top lines. But it happens to support a policy conclusion.
So is there a conscious line between research and marketing? Probably a good organization, yes. Is it enforced structurally, separate reporting lines, no lobbying, charters, editorial independence?
Those are all questions to ask every time you get a document. EBRI was built that way from day one. It’s not as much that today, but it’s still more than most entities.
Industry-wide, the line is inconsistently maintained rather than deliberately erased, I would argue. And most in-house teams believe they’re rigorous, but sit inside distribution and marketing functions in many cases. And all of those things tie into the evaluation.
And this is particularly true, again, when it gets to a reporter working on an article. It’s particularly true on Capitol Hill or in a government agency. When they’re trying to use that data in a deliberative way to reach a conclusion on implications, then yes, it does matter.
Jeffrey Snyder, Broadcast Retirement Network
Yeah. Look, we all have biases. And the question is, how do you control?
I have opinions, you have opinions, everyone’s got an opinion. How do you use the data? I could talk to you.
We’ll have to bring you back. But let me close out by asking about artificial intelligence, large language models, which are basically the open AIs, the ducked out AIs, et cetera. How does that change or does that change the research game today and in the future?
Because now I can access my phone. I was going to pick my phone up. I don’t know where it is.
Dallas Salisbury, Consumer Policy Center
It actually changes it fundamentally. I’ll give you an explicit example based on ABRI and the ages, so to speak. When we started ABRI in 1978, the first thing we did was create, working with the Urban Institute under contract, detailed thick compendiums of everything that was out there on health and retirement.
Because you’d have to literally go to a library or libraries and search for this stuff. And so we put it all into nice single volume. That ended up being a popular free product, but not download because at that point, everything had to be printed.
We then created the ABRI data book of employee benefits, which got up to about 500 pages, was published in several editions. And it took all these government data sources and all the private and public, put it all in one book. Again, pre-internet, you had to have it someplace, but this gave somebody a book on their shelf that could literally save somebody weeks of or paying a research assistant for months.
Then the internet came along. ABRI created its first website in the early 1990s. Very limited.
You can still go to webarchive.org and look at the very first ABRI website and every website in between. Fun fact for anybody that doesn’t readily have access to ABRI issue briefs, but would like them, go to webarchive.org and you say that you want to look at the lump sum distribution book that published in the early 80s. You can go to that web archive and you can download the book.
You can download issue briefs through about 214. The data is out there, but you got to go looking for it. Then introduced the next step in it of AI.
I put a query into AI, or anybody can put a query into AI now and say, find me the most recent data from census, social security, and private surveys on receipt of retirement income and retirement. In a matter of minutes, it will dig into the census bureau databases and the social security research reports. It will, at minimum, provide you the links to the exact tables that you need for any of this work.
I admit, I now use AI in all of my work related to writing on topics, because I know data exists so I can precisely tell it what survey I’m looking for. But instead of my having to go dig it out, it produces my direct access to that table at the bureau or the social security administration from any point in history of its publication instantly. I’m not asking it to interpret the data.
I’m not asking it to tell me meaning. I’m simply saying, find me the table. AI is being used in that way within private companies, within unions, on Capitol Hill.
Ironically, what it does is it almost takes us back to pre-EBRI, when the only real data out there was in all these government agency databases, but they were hard to get at and hard to find, and the internet didn’t exist, so you had to look at print books. Now, it has moved us to where all of that data is now instantly available. The early EBRI, in fact, if you had AI in 1978, you could argue that EBRI wouldn’t have been needed, because the people on Capitol Hill and reporters without spending weeks and months or asking agencies to pull things together, AI would have given them what they needed very quickly. An exception. Two exceptions.
The ESOP after Enron. Most of that was non-public data or in data sets that were not yet on the internet. So it was still needed.
And if it’s proprietary data, there’s still a role for consortiums and partnerships. And if an entity hires a researcher that wants to be, and quotes, a real researcher, and they segregate it from marketing, then there are organizations that can still do extraordinary research that is helpful. And AI simply enhances that dramatically.
So world of change.
Jeffrey Snyder, Broadcast Retirement Network
Yeah, absolutely. Well, Dallas, we’re going to have to leave it there. We’ll have to bring you back.
And as I said, I’m very interested in other people’s opinions as well. So I want to be very open about that. I mean, I think the world research is so vital to what we do as consultants, as business leaders in the retirement industry and the benefits industry.
Thanks so much for joining us. And we look forward to having you back on the program again very soon, sir.
Dallas Salisbury, Consumer Policy Center
Thanks so much.