Auto Enroll: 3%, 5%, 7% – What’s Optimal?

We met Dr. Mingli Zhong when she was working alongside Dr. Olivia Mitchell and team at the University of Pennsylvania’s Wharton School, focused on retirement savings research. Since then Mingli has been on the move, but her focus is steady. She asks important questions about how we can optimize the way we use automatic features in our retirement savings plans and programs. She’s also studying pandemic-related impacts on savers, going beyond top-level results to understand how this economy is differentiating between those of us that kept our jobs, and those of us who didn’t.

Mingli, you are a postdoctoral fellow at the National Bureau of Economic Research. How did you come to your current role and what are you working on?

Thank you for having me! In 2019 I was in the last year of my PhD program in applied economics at the Wharton School, University of Pennsylvania. I often found myself on the NBER website reading their latest working papers. One day I spotted a post-doc opportunity funded by the Social Security Administration where they were specifically looking for someone doing retirement policy. This felt like a really good match, because at that time I was looking for opportunities to continue doing my research full time. I applied and fortunately I got in!

You recently did some work focused on auto-enroll retirement programs where you were analyzing the optimal default savings rate. A lot of people want to know this answer! Let’s start by asking how you define “optimal.”

I love this space. I've been working on “optimal retirement policy” questions. First of all, there are a range of questions. What is the optimal initial default rate? What is the optimal automatic escalation arrangement? How do we optimally design our auto-enrollment retirement plans taking into account Social Security and other social safety net programs?

My starting point has been to understand what an optimal initial default rate is. You can see early results of this analysis in the just-published piece, Optimal Default Retirement Savings: Theory and Evidence from OregonSaves. In this research we define optimal as the rate that maximizes the total lifetime individual wellbeing. That definition includes all individuals eligible for a retirement plan. We include passive savers who stay at the program’s default rate, as well active individuals who save at a different rate or might opt out of the program altogether.

Very cool.

The way that I think that a default rate can improve individual wellbeing is through two channels. The first channel is for passive savers. Definitely some of them increase savings after they are automatically enrolled; that's how they save more.

And for active individuals who save at a different level, they can benefit from the default rate as well. Because the default rate is a cost effective policy tool to get their attention, to prompt them to start thinking about their retirement savings and their long-term savings goals. Maybe they feel the default rate is not the right rate for them. But it encourages them to start thinking about how much they should save and they will choose either a non-default rate, or maybe delay the savings decision to the next year. So that's how I think about optimality.

Can a specific default savings level prompt more saving, or less saving?

From the plan design or policy makers’ perspective, as we set the default rate we also need to prioritize.

Please say more.

If the program’s priority is to expand coverage, we want as many people to be included in the program as possible. Then the policymaker may want to set a relatively low default rate so everyone would be encouraged to stay in the program.

If we want people to build their nest eggs, to really save something, then we might want to consider setting a relatively high default rate, because that would help them save more, faster. These higher levels might cause some people to opt out of the default rate or out of the program altogether. That doesn't necessarily mean the total assets in the program would be less. Because of the high default rate, people who can afford to save would save more.

So that's a summary of how we think about the optimal default rate. I also discussed the ideas of this project at the virtual RAND Behavioral Finance Forum in 2020, in Session 3: Household Finance.

We loved that your work considered the wellbeing of the whole population. Here the optimal rate was the rate at which you capture both the greatest number of savers and the maximum amount of savings.

Exactly. I used OregonSaves data, thanks to Oregon Treasurer Read and the OregonSaves team who helped me access the data. Calibrating the model, what I found is the optimal initial default rate is roughly between 5% to 10%. In the paper I was more specific and referenced 7% as an optimal initial default rate. In another paper that I co-authored with John Chalmers, Olivia S. Mitchell, and Jonathan Reuter, we extensively documented the impact of OregonSaves including some key questions of who has opted out and who is participating.

7% -- interestingly this is higher than where the state Auto IRA programs are today. It's also higher than most 401(k) plan defaults, which are often closer to 6%.

Yes, and there are some assumptions I made. In this first paper I didn't take into account Social Security. That's what I'm working on next. If you add in Social Security and other social benefits, the optimal default rate could be slightly lower to increase the number of savers without reducing overall retirement income wealth. That's my advance guess. So, we will see!

Let’s talk about auto-escalation, what do you want to look at there?

This is a space where I am trying to get some data to complete a project. My hypothesis is that we should think about two scenarios.

In one we would set a relatively high initial default rate of 7%, and that is a way to increase early savings. In a second scenario is we would start from a relatively low default rate, maybe 3% or 4%, and then we gradually increased to 7%.

We may find that the second way could be even more welfare-improving because workers would have more time to adjust their consumption over time.

There are other questions to be answered in terms of the optimal auto-escalation:

• What is the optimal frequency with which to increase the rate

• How soon should we start auto-escalation, and

• What is the optimal date to increase auto-escalation. Is January first a good date, or should we associate auto-escalation with a pay increase?

Many plans use January 1 as an auto-escalation date. Conceptually, I'm not sure if this is a good day, because it's right after the holiday season when people spend a lot of money. Maybe we could pick the day after the tax season, after they receive their tax refunds. That's something to consider.

I would definitely would love to do a field experiment, a pilot experiment, to really test the frequency and timing that maximize the power of auto-escalation. (Readers – any takers for this natural experiment? Mingli wants to talk with you!)

The last 18 months have been very unusual. Are you seeing any interesting outcomes or lessons learned via the work that you are doing?

Yes, actually this is the second line of research I'm working on right now: understanding the impact of COVID-19 on financial wellbeing. We want to understand how COVID-19 has changed savings and borrowing behavior. How has it changed the long-term household financial security?

One thing I've seen both from recent studies and my own ongoing project is that there seem to be two opposite trends happening.

In one group, people in households still have their jobs. Because they travel less or are moving around less overall, they also spend less. And in addition to loan forbearance[1], we also see an improvement in their credit scores. They're saving more and they continue to contribute to their retirement accounts. So they are doing better than they were pre-pandemic.

On the other hand, there is a second group of people who have lost their jobs. They tend to spend their stimulus payment fairly quickly, in the exact month that they received the stimulus payment. Because they lost their jobs, they stopped contributing to their retirement accounts. They’ve also had difficulty making ends meet. They are clearly in a worse position than they were pre-pandemic.

Part of my ongoing projects is to find out who's in the second group of people and who needs recovery resources the most.

That is important work. You're doing your postdoctoral work right now but completing this phase soon. What comes next?

Yes! I will finish my post-doc at NBER in June and start working at the Urban Institute in July. I will be joining the Labor, Human Services and Population Center (LHP). It's a great center.

My team and I will work together on financial security and retirement savings. We are also thinking about the gig economy, and how we should update retirement policy to accommodate growing gig work and platforms.

We are looking forward to seeing your work and what you learn! In closing, we’re slowly transitioning out of a very challenging time. Tell us about any silver linings you've experienced during the pandemic.

It's been pretty tough on everyone, especially for families with kids. I myself have a four year old. On the bright side, I did spend more time with my family and my daughter is very happy about that. We had many chances to just have family time, we went on a camping trip together; these are good things.

In terms of career, one silver lining is that in a world of virtual engagement, I’ve been able to attend and present at many great conferences. Just a couple of weeks ago, I managed to present in a conference and in Australia in the morning, and then attend another conference in California on the same day. That could never happen pre pandemic!

And for a researcher, the virtual workplace has helped my productivity a bit – and I have been able to expand my professional network at the same time. So we get by!

Dr. Mingli Zhong, thank you so much for sharing everything that you're working on with us. Even your early findings are important and immediately useful.

Would you like to know more? You can connect directly with Mingli Zhong here. You can follow Mingli’s work at her personal website. You can also connect with Mingli on LinkedIn.

This piece was featured in the June 3, 2021 edition of Retirement Security Matters. For more fresh thinking on retirement savings innovation, check out the newsletter here.