When Cass Sunstein and I wrote the book Nudge we had two goals. First, provide a foundation for the elusive middle ground between libertarian anarchy and a tightly controlled nanny state. That foundation was based on the premise of “libertarian paternalism”: Strategies for helping people achieve their goals without forcing anyone to do anything. Our friend George Loewenstein and colleagues had a similar vision (Camerer, Issacharoff, Loewenstein, O'Donoghue, & Rabin, Reference Camerer, Issacharoff, Loewenstein, O'Donoghue and Rabin2003). Devising policies that are somewhere between the extremes in our polarized world can be essential because, by definition, official public policies require political support, either via government legislation and/or regulation.
Our second goal was to use the findings of behavioral science to inform such policies, to give the policies the best chance of achieving the desired goals. In the United States, and many other countries, the rules and regulations of governments at every level tend to be written by lawyers (including legislators) who are advised by economists. There is a Council of Economic Advisors in the White House, but no Council of Behavioral Scientists. We thought that highlighting the insights from other social sciences might help government officials to devise better policies. Much to our surprise, hundreds of governments, beginning with Britain and the United States, have created organizations of behavioral scientists that have become known colloquially as “nudge units.”
According to Chater & Loewenstein (C&L) our efforts and those of all behavioral scientists have been an enormous failure. C&L assert that behaviorally inspired interventions are not only ineffective, but also they are downright harmful. They argue (with no plausible supporting evidence, or even a single compelling example) that by trying to improve the current state of affairs in various domains we all have deterred some stronger measures that would be better.
The problem, C&L assert, is that behavioral scientists have been concentrating their efforts using something they call the i-frame instead of using the much more effective s-frame. Unfortunately, neither of these terms is clearly defined in the article. The i-frame obviously has something to do with individuals, but what constitutes an s-frame intervention is not spelled out, though in the domain of retirement savings it seems to favor mandates over policies that permit citizens to opt out.
This is an odd critique on many levels. One obvious question is: Who do C&L think is limiting themselves to the i-frame? Certainly not Cass Sunstein and me. Cass has served in various administrative roles in government where he got to design or modify actual policies, and in our book Nudge, at least three quarters of the chapters are explicitly about structural policy designs. Choice architecture, our primary tool, is policy infrastructure.
An important point to stress is that behavioral scientists, whether they are in academia or nudge units, do not have the authority to experiment with most of the rules and regulations in a given domain. No nudge unit has the ability to say, hey, let's try a carbon tax in half the country and strict emission rules in the other and see how it goes. In practice they are often limited to messaging campaigns, which are less impactful.
This implies that the range of interventions studied by behavioral scientists is truncated by what I call permission bias: You can only test what you can get the approval to try. It is wrong to infer from this fact of life that behavioral scientists are using the wrong “frame.” Rather, they face constraints! It also makes it problematic to judge the potential impact of possible behavioral policy interventions based on the set of randomized controlled experiments behavioral scientists have been allowed to run. In many cases social scientists must rely on natural quasi-experiments made possible when governments decide to change the rules or offer a new program. As I explain below, the United Kingdom has provided such an opportunity in the domain of retirement savings that demonstrates the potential effectiveness of nudging.
Retirement savings is an attractive topic for behavioral economics research because the task is “difficult” in two ways. First, the mathematical problem of how much to save and how to invest the money is hard, even for an economist. Second, saving for retirement requires exerting self-control in order to delay consumption for decades. The traditional economic approach to this problem is simple: People solve the math problem optimally, and they implement the appropriate plan. (Economic agents excel at both mathematics and willpower.)
In four decades of behavioral economic research on this topic, the focus has always been on making the system work better for humans. Isn't that the s-frame? Do C&L think we have been just going around quoting Ben Franklin's line: “A penny saved is a penny earned?”
As C&L note, the early systems devised to help households save for retirement took the form of both public and private “defined-benefit” plans that guarantee a retirement income stream which depends on years of work and level of pay. These plans are easy for participants because there are few choices to make aside from when to retire, but they are costly to administer. The plan sponsor (e.g., government, employer, or union) has to set aside large amounts of money for decades and deal with multiple risks including the returns on the portfolio and what is called “longevity risk”: The chance that participants live longer than expected. Defined-benefit plans were not perfect for all participants either. The plans were especially attractive to employees who worked with one employer (or union) for most of their career, and who had the good luck that their plan sponsor fully funded the plan, invested wisely, and did not go bankrupt. Defined-benefit plans were particularly good for high-paid workers because the pension depended on final salary. C&L seem to be afflicted by defined-benefit plan nostalgia. They probably also miss Pan Am and TWA, companies with defined-benefit plans that went bankrupt.
Historically, when it became legal to offer defined-contribution plans in the 1990s, virtually all new firms adopted this framework, and some older companies transitioned to it. As C&L acknowledge, defined-benefit plans are becoming rare, and there are virtually no new ones being created. Given this, it is hard to buy C&L's contention that the efforts by behavioral economists to make defined-contribution plans more user-friendly deterred the return of defined-benefit plans. Private sector defined-benefit plans, like typewriters and dial telephones, are obsolete technologies few people pine for. Meanwhile, governments are finding that pay-as-you-go social security systems that have a defined-benefit structure are facing funding crises in an era of increasing life expectancy and declining birth rates.
In trying to improve the growing number of defined-contribution plans, behavioral economists realized that participants needed help in three domains: Signing up for the plan, saving enough to provide for their retirement needs, and investing the money wisely. The practical solutions to these problems were automatic enrollment (make joining the default), automatic escalation or Save More Tomorrow (gradually raising the saving rate over time), and well-designed default investment products. None of these features existed in 2000, but now, two decades later, all are common in a majority of plans. Accomplishing this had to begin with convincing both regulators and legislators to make these options legal.
In the United States in 2005, the only default investment product that had a legal safe harbor was a money market fund or some equivalent low-return product. The Department of Labor had to be convinced to create new types of what they call Qualified Default Investment Alternatives. Then, in 2006 Congress passed the Pension Protection Act making automatic enrollment and automatic escalation clearly legal, and offering employers an incentive to adopt them. A law called SECURE 2.0 just passed at the end of 2022 which (in a limited way) further encourages the use of these nudges. Bottom line: Changing laws and regulations is hard and can take decades. Such work is not usually rewarded in academia.
C&L question whether these behaviorally informed innovations actually work. A relatively new program launched in Britain demonstrates that they can (Nest Insight, 2022). The National Employment Savings Trust (NEST) pension scheme was created to make sure that all employees had access to a workplace retirement savings plan even if their employer did not offer one. (A similar plan is badly needed in the United States.) Firms were required to offer a workplace plan if they didn't have one (with a government-run option available at low cost) and to automatically enroll workers who were 22 or older and made more than £8,105. Cleverly, the initial minimum savings rate was just 1%, to avoid “pay-stub shock,” but it was steadily increased to 8%: 5% from the employee and at least 3% from the employer, though most large employers actually contribute more. Opt-out rates have remained around just 8%, and nearly everyone elects the low-cost default investment fund. There are over 17 million workers in the plan now. This is a remarkable success.
C&L are critical of the NEST plan calling the Australian alternative “far superior,” though it is hard to see what criteria they are using to make this claim. Yes, the Australian plan is mandatory for workers and until recently had a slightly higher savings rate of 9% (that is now being gradually increased to 12%). But the design of the plan has distinct flaws: There are hundreds of investment options, which is too many, and fees can be high. Why are C&L such fans of the Aussie plan? The key factor seems to be inability for employees to opt out, though a bit later they refer to this as a “relatively minor feature.” Call me confused.
A proper s-frame evaluation would recognize that a required plan might be less attractive than one that allows opt out and still achieves 92% participation. Even people who do not value freedom of choice per se might be persuaded by the fact that those who do opt out of automatic enrollment tend to have what seem to be good reasons (Chalmers, Mitchell, Reuter, & Zhong, Reference Chalmers, Mitchell, Reuter and Zhong2021). Another feature of the Australian plan that C&L praise is the prohibition on workers borrowing from their accumulated savings, even in the case of an emergency. I am not sure why this is a plus. Isn't it possible that borrowing against retirement savings to finance a new furnace or medical expense is better than using a credit card or payday lender? One study in the United States concludes that 401(k) loans “are neither a blessing nor a bogeyman” (Beshears, Choi, Laibson, & Madrian, Reference Beshears, Choi, Laibson and Madrian2008).
In conclusion, the philosophy of nudging is partly based on humility. Proponents are humble about the ability of human beings to solve all problems themselves, but are also humble about the ability of an outside party to always know what is best. The requirement that citizens are able to opt out acts as an insurance policy against overreach by plan designers. Crucially, it can also sell the policy to legislators, as it did in the United Kingdom.
In contrast, C&L seem to think less highly of people and more highly of governments than I do. They are so sure that they know what is best that they want to require it, and they appear confident that governments will share their values.
To this all I can do is to channel Mick Jagger and say: “You can't always get the government you want.” If you want to advocate for more intrusive government actions, you better be confident that the government will make those choices wisely. Alas,your government may not share your “frame.”
When Cass Sunstein and I wrote the book Nudge we had two goals. First, provide a foundation for the elusive middle ground between libertarian anarchy and a tightly controlled nanny state. That foundation was based on the premise of “libertarian paternalism”: Strategies for helping people achieve their goals without forcing anyone to do anything. Our friend George Loewenstein and colleagues had a similar vision (Camerer, Issacharoff, Loewenstein, O'Donoghue, & Rabin, Reference Camerer, Issacharoff, Loewenstein, O'Donoghue and Rabin2003). Devising policies that are somewhere between the extremes in our polarized world can be essential because, by definition, official public policies require political support, either via government legislation and/or regulation.
Our second goal was to use the findings of behavioral science to inform such policies, to give the policies the best chance of achieving the desired goals. In the United States, and many other countries, the rules and regulations of governments at every level tend to be written by lawyers (including legislators) who are advised by economists. There is a Council of Economic Advisors in the White House, but no Council of Behavioral Scientists. We thought that highlighting the insights from other social sciences might help government officials to devise better policies. Much to our surprise, hundreds of governments, beginning with Britain and the United States, have created organizations of behavioral scientists that have become known colloquially as “nudge units.”
According to Chater & Loewenstein (C&L) our efforts and those of all behavioral scientists have been an enormous failure. C&L assert that behaviorally inspired interventions are not only ineffective, but also they are downright harmful. They argue (with no plausible supporting evidence, or even a single compelling example) that by trying to improve the current state of affairs in various domains we all have deterred some stronger measures that would be better.
The problem, C&L assert, is that behavioral scientists have been concentrating their efforts using something they call the i-frame instead of using the much more effective s-frame. Unfortunately, neither of these terms is clearly defined in the article. The i-frame obviously has something to do with individuals, but what constitutes an s-frame intervention is not spelled out, though in the domain of retirement savings it seems to favor mandates over policies that permit citizens to opt out.
This is an odd critique on many levels. One obvious question is: Who do C&L think is limiting themselves to the i-frame? Certainly not Cass Sunstein and me. Cass has served in various administrative roles in government where he got to design or modify actual policies, and in our book Nudge, at least three quarters of the chapters are explicitly about structural policy designs. Choice architecture, our primary tool, is policy infrastructure.
An important point to stress is that behavioral scientists, whether they are in academia or nudge units, do not have the authority to experiment with most of the rules and regulations in a given domain. No nudge unit has the ability to say, hey, let's try a carbon tax in half the country and strict emission rules in the other and see how it goes. In practice they are often limited to messaging campaigns, which are less impactful.
This implies that the range of interventions studied by behavioral scientists is truncated by what I call permission bias: You can only test what you can get the approval to try. It is wrong to infer from this fact of life that behavioral scientists are using the wrong “frame.” Rather, they face constraints! It also makes it problematic to judge the potential impact of possible behavioral policy interventions based on the set of randomized controlled experiments behavioral scientists have been allowed to run. In many cases social scientists must rely on natural quasi-experiments made possible when governments decide to change the rules or offer a new program. As I explain below, the United Kingdom has provided such an opportunity in the domain of retirement savings that demonstrates the potential effectiveness of nudging.
Retirement savings is an attractive topic for behavioral economics research because the task is “difficult” in two ways. First, the mathematical problem of how much to save and how to invest the money is hard, even for an economist. Second, saving for retirement requires exerting self-control in order to delay consumption for decades. The traditional economic approach to this problem is simple: People solve the math problem optimally, and they implement the appropriate plan. (Economic agents excel at both mathematics and willpower.)
In four decades of behavioral economic research on this topic, the focus has always been on making the system work better for humans. Isn't that the s-frame? Do C&L think we have been just going around quoting Ben Franklin's line: “A penny saved is a penny earned?”
As C&L note, the early systems devised to help households save for retirement took the form of both public and private “defined-benefit” plans that guarantee a retirement income stream which depends on years of work and level of pay. These plans are easy for participants because there are few choices to make aside from when to retire, but they are costly to administer. The plan sponsor (e.g., government, employer, or union) has to set aside large amounts of money for decades and deal with multiple risks including the returns on the portfolio and what is called “longevity risk”: The chance that participants live longer than expected. Defined-benefit plans were not perfect for all participants either. The plans were especially attractive to employees who worked with one employer (or union) for most of their career, and who had the good luck that their plan sponsor fully funded the plan, invested wisely, and did not go bankrupt. Defined-benefit plans were particularly good for high-paid workers because the pension depended on final salary. C&L seem to be afflicted by defined-benefit plan nostalgia. They probably also miss Pan Am and TWA, companies with defined-benefit plans that went bankrupt.
Historically, when it became legal to offer defined-contribution plans in the 1990s, virtually all new firms adopted this framework, and some older companies transitioned to it. As C&L acknowledge, defined-benefit plans are becoming rare, and there are virtually no new ones being created. Given this, it is hard to buy C&L's contention that the efforts by behavioral economists to make defined-contribution plans more user-friendly deterred the return of defined-benefit plans. Private sector defined-benefit plans, like typewriters and dial telephones, are obsolete technologies few people pine for. Meanwhile, governments are finding that pay-as-you-go social security systems that have a defined-benefit structure are facing funding crises in an era of increasing life expectancy and declining birth rates.
In trying to improve the growing number of defined-contribution plans, behavioral economists realized that participants needed help in three domains: Signing up for the plan, saving enough to provide for their retirement needs, and investing the money wisely. The practical solutions to these problems were automatic enrollment (make joining the default), automatic escalation or Save More Tomorrow (gradually raising the saving rate over time), and well-designed default investment products. None of these features existed in 2000, but now, two decades later, all are common in a majority of plans. Accomplishing this had to begin with convincing both regulators and legislators to make these options legal.
In the United States in 2005, the only default investment product that had a legal safe harbor was a money market fund or some equivalent low-return product. The Department of Labor had to be convinced to create new types of what they call Qualified Default Investment Alternatives. Then, in 2006 Congress passed the Pension Protection Act making automatic enrollment and automatic escalation clearly legal, and offering employers an incentive to adopt them. A law called SECURE 2.0 just passed at the end of 2022 which (in a limited way) further encourages the use of these nudges. Bottom line: Changing laws and regulations is hard and can take decades. Such work is not usually rewarded in academia.
C&L question whether these behaviorally informed innovations actually work. A relatively new program launched in Britain demonstrates that they can (Nest Insight, 2022). The National Employment Savings Trust (NEST) pension scheme was created to make sure that all employees had access to a workplace retirement savings plan even if their employer did not offer one. (A similar plan is badly needed in the United States.) Firms were required to offer a workplace plan if they didn't have one (with a government-run option available at low cost) and to automatically enroll workers who were 22 or older and made more than £8,105. Cleverly, the initial minimum savings rate was just 1%, to avoid “pay-stub shock,” but it was steadily increased to 8%: 5% from the employee and at least 3% from the employer, though most large employers actually contribute more. Opt-out rates have remained around just 8%, and nearly everyone elects the low-cost default investment fund. There are over 17 million workers in the plan now. This is a remarkable success.
C&L are critical of the NEST plan calling the Australian alternative “far superior,” though it is hard to see what criteria they are using to make this claim. Yes, the Australian plan is mandatory for workers and until recently had a slightly higher savings rate of 9% (that is now being gradually increased to 12%). But the design of the plan has distinct flaws: There are hundreds of investment options, which is too many, and fees can be high. Why are C&L such fans of the Aussie plan? The key factor seems to be inability for employees to opt out, though a bit later they refer to this as a “relatively minor feature.” Call me confused.
A proper s-frame evaluation would recognize that a required plan might be less attractive than one that allows opt out and still achieves 92% participation. Even people who do not value freedom of choice per se might be persuaded by the fact that those who do opt out of automatic enrollment tend to have what seem to be good reasons (Chalmers, Mitchell, Reuter, & Zhong, Reference Chalmers, Mitchell, Reuter and Zhong2021). Another feature of the Australian plan that C&L praise is the prohibition on workers borrowing from their accumulated savings, even in the case of an emergency. I am not sure why this is a plus. Isn't it possible that borrowing against retirement savings to finance a new furnace or medical expense is better than using a credit card or payday lender? One study in the United States concludes that 401(k) loans “are neither a blessing nor a bogeyman” (Beshears, Choi, Laibson, & Madrian, Reference Beshears, Choi, Laibson and Madrian2008).
In conclusion, the philosophy of nudging is partly based on humility. Proponents are humble about the ability of human beings to solve all problems themselves, but are also humble about the ability of an outside party to always know what is best. The requirement that citizens are able to opt out acts as an insurance policy against overreach by plan designers. Crucially, it can also sell the policy to legislators, as it did in the United Kingdom.
In contrast, C&L seem to think less highly of people and more highly of governments than I do. They are so sure that they know what is best that they want to require it, and they appear confident that governments will share their values.
To this all I can do is to channel Mick Jagger and say: “You can't always get the government you want.” If you want to advocate for more intrusive government actions, you better be confident that the government will make those choices wisely. Alas,your government may not share your “frame.”
Acknowledgments
I acknowledge helpful comments from David Laibson, Daniel Kahneman, and Cass Sunstein.
Financial support
This research received no specific grant from any funding agency, commercial, or not-for-profit sectors.
Competing interest
None.