Now from a Panel Discussion on retirement economics. Researchers presented their work during the 19th annual meeting of the Retirement Research consortium posted by Boston College in washington d. C. This Panel Discussion is 90 minutes. [inaudible conversations] [inaudible conversations] thank you. So the sixth panel is a topic would Social Security changes prompt people to change their plans in the first paper is a behavioral or consumption of Social Security changes. Melissa and our advisers will be the discussants. Thank you. Thank you all for having me. Again Social Security thank you for funding this research and all of you for listening to the speech twice but i want to thank my coauthor who worked on the paper with me and i also want to thank allen and constantine who have been allowing us to use their model and i want to thank them for that. This is really about how changes in Social Security affect two things behavior and consumption. A variety of changes. Something has to be done about that at some point and there are a variety of things that could be done revenue increases in the combinations of those things. The chief actuary at raleigh was a wide variety of changes associated with program and their impact on our program. Its also useful in what this paper does is analyze the individual behavioral effects of any change in their effects on welfare in the state through the Household Consumption through this project uses the structural model and i will explain what it isnt the next few slides to evaluate this impacts of changes associated with Social Security for the program that has similar effects on the programs bottom line. So economic models present a way to analyze these effects if you think about what are reduced for model does its a linear regression model. What a structural model does is a little bit different that it starts with assumptions on workers utility function how they feel towards Different Things and in this model towards consumption and models choice thats ended actually estimate how workers feel about those things about consumption and about working and it does that based on their observed to hate here. For example if workers in poor health he observed them retiring at age 60 before they can receive Social Security might infer from that the people that are in poor health hate to work and thats why they have low consumption so the model base looks at what people do and try to understand how they feel about the decisions they face. The really nice thing about these models and i can tell you working on these models is throwing my computer across the room but the benefit is once you do all that youve continue behavior under different constraints. For example you can imagine the effects of unstable workers doing something about delaying eligibility age and maybe you see some markers actually do work longer to avoid that longer period of low consumption whereas others may bridge that gap and retire early because they didnt like working when they are ill. Once you have peoples preferences and want to have a model you can change those constraints and see how they respond and thats basically what a structural model lets you do and thats why its a useful approach in addition to the approaches that you are used to seeing. So this example illustrates what im basically going to do this paper which is change some of the constraints people face with respect to the Social Security program. Im going to look at three different policies and all these policies can be calibrated to have a similar effect on Social Securitys bottom line depending on how their face and then im going to assume these were faced in a similar way that these different changes are at the increase in the costofliving adjusted by 0. 5 Percentage Points and we talked about that yesterday or something similar to that an increasing the payroll tax by 1. 5 so 7. 75 at all three of these changes can have a similar affect on Social Securitys bottom line. What we want to do this paper is the how it can actually affect individual behavior around retirement and weigh consumption over a lifetime. Im going to show to slides in row. Both of them have equations that are scary. They are scary but i can explain in poor people in this model to basically they maximize their lifetime utility. They look into the future and they think about whats going to happen to them. They discount the future because you care about yourself today more than you tomorrow to some extent do you discount your future because you cant seem to the future and thats the second part of the equation and you get utility from consumption. And so basically people say at the given time there looking forward to making decisions knowing what they do today is going to influence the utility of today but also the utility from things that happen in the future so thats the first equation and theres one more and it is basically whats called a budget constraint. Its very simple to understand. Your absence at any given time you are consumption that we had yesterday multiplied by any returns you copied your labor and income your spouses income in your own Social Security and income. If you decide to work today you know that will affect how much you will have tomorrow because you wont have the income if you decide not to work and you know tomorrow you wont have as much as you have today. People are making these decisions knowing how these things basically are played out so even though the actual way that this all works is relatively complicated the ed show thing thats going on here something we all deal with. Social security and pension benefits plan Important Role in the model. Individuals see Social Security as model which is a little bit simple but not far from reality as soon as they can basically once they fall below the earnings test. They have actually modified their model and away to allow that to be detached and in the model we are using their connected. People basically claims ozar security and importantly the benefits accrue as they do in the real world. As you delay your retirement your benefit goes up with adjustment and the pension is also model and accumulate space on the individuals actual plan of rules a lot of time those rules would encourage people to retire at a given time or work a given length of time. All these things are built into the model in that way. So im going to try to convince you that when the change this model we get realistic effects of what people are going to do so the first thing i would hope you want to know is if its the real word with its estimated nonof this model does. The model fits the behavior and of this graph we have 60 to 64 going left to right. The red line sexual behavior of the age cohort formed in 1931 in 1941 and behavior predicted by the structural model that i described. You can see it did fairly well. Importantly it nears an increasing complete retirement. The model doesnt enforce that in any way. Simply because the financial incentive makes sense at that time. The other thing im going to do is im going to change the environment this model was estimated on and im going to see how people respond and for you to believe that you want to believe the model can fit things that actually happened. Recall that internal validation. We checked the model to predict changes that a party happened so in an earlier paper and i always plug my own work. Size fun. Between the cold hard this model was estimated on originally and 12 years later for big thing happened. Health improves, mortality improved people are living longer, there was an increase in the fra and a shift towards d. C. Pension plans. What we did was say what if we took the fra to a and made them look like a cohort form 12 years later in terms of these four things and what we wanted to see was does it actually looked at the model and what we actually see . I wouldnt be telling you all of this if it didnt work i guess but the model predicts the responses to these changes quite well so here we have 60 to 69. The black one is the cohort in a dotted line is in the red line is the actual early report and you can see the dot the line does show a decrease in retirement of a relatively similar magnitude we saw in the data. Time will tell us as his cohort ages whether or not the model continues to predict a gap correctly but this should give you some confidence that this model can predict how a change in behavior response to some dig change in the economy during that time period. The last thing i will say about the model is this model is estimated on a Social Security environment that is different than we have today so is estimated where the fra is 65 for example. In my report i wont jump off a different point time so we are going to use a baseline policy and fra of 67 at 10 per year that starts at 67 and knowing the environment of these changes at their starting point to the baseline is scheduled under current law obviously they continue our path we reduce by 30 in 2034. The baseline is that people the situation goes on as it is right now. What i have here is a somewhat complicated table. Ask you to focus just on the three columns on your right. Those are the change in complete retirement at ages 62 to 69 for each of the three policies that describe and fra increase, a reduction in the tax increase. Can see the fra increase decreases the share of people fully retired by between basically two and 4. 5 depending on the age you look at. The tax increase is much smaller based on behavior and encourages people to retire slightly earlier. The impact is fairly small. It doesnt have a huge behavioral effects. The reason nicole has a smaller effect the fact the reduce the cola doesnt show up for quite a bit of time after that so the effect has lagged and People Discount the future so they dont necessarily take it into consideration like he do the fra increase which has an immediate effect. The effect of the production tends to be larger for low income workers. Its relying on Social Security and they have less benefits. If you look on the left three columns increasing the fra to 69 we have low middle and high Income Distribution and cola reduction. The low group the effects are between 4. 5 and 7 reduction and thats compared to the foreperson besought earlier and the highincome group the effect is a 2 reduction so the cola reduction is between three and 4 reduction for the highest 1 . The behavioral effects for the benefit reduction tend to be larger for lower income people. The timing of consumption changes differs considerably across these policies so ages 25 to 90 the black line is a tax increase and the red at cola reduction in the fra a dotted line to this tax increase results in an decrease in consumption during the working years. That affects retirement with a reduction to the fact that people less because the tax increase. The effect becomes knowledgeable agent to stop relying on savings or malign Social Security so that retirement is shrinking. The cola reduction of tax increase has little effect during the working years and it tends to pick up in the 50s and people start to recognize the benefits use and save a little bit more. Then during the retirement theres a larger reduction in consumption ranging to his seven or 8 . The cola reduction is a smaller effect at first but that affect picks up as people get older and the fact that there were long, is shrinking starts to take hold the consumption affect the summer across income for a tax release because of the flat tax increase. The benefit reduction the reduction tends to be larger relative to lowincome people and the exact same reason that they tend to rely on Social Security is the main source of income for retirement where other groups have other sources they can rely on. Its between six and eight for the fra and as low as 12 for the cola increase as the cola increase takes hold later in life. In conclusion i think the reductions in benefit will likely have a larger effect in retirement. Both in terms of behavior and in terms of consumption. Tax increases have a smaller effect on consumption in retirement but no longer period decrease consumption and working lives and to some extent that carries over into retirement to decrease savings and hopefully ive convinced you the structural models are useful way to frame the debate on what kind of changes in Social Security should we implement and with that i thank you. [applause] good morning. I would like to thank alicia, jeff and the rest of the team for inviting me to be here today i would also like to thank my colleagues joanne and kay lynn who are in the audience today, for doing the research on this project and for putting together this presentation. I have two disclaimers. One is i am not an academic researcher so one might wonder why i was invited here. And if the operations work for you jeff means that impacts on me so thats one disclaimer. The other is that we put this presentation together based on the draft of the paper. We have four different options so my presentation is going to cover all the forms so we will start there. So, what i was asked to do was look at not just the Public Policy implications of jeffs research but also the political feasibility so i think jeff i found your paper fascinating and i thought the conclusions were really interesting in terms of possibly behaviors in response to different changes as well as Consumption Patterns pay what we tried to do here to bring in some real light reallife possibilities. We look at different aspects of what happens in the real world which is we look at survey data over time of what people say they will do in response or what changes are except double to them but we also look at voting patterns over time. Thats what im going to cover today. First just to review the four different kinds of changes that were in the draft paper and who they mostly have an impact on so first is the cola adjustment and for that adjustment mostly thats going to have an impact on people who tend to live longer and those could be more highly educated higher income individuals for the most part. The opposite effect happens when you raise the retirement age as jeff would say so those people who will be the most affected by that of people who retire early and you dont have as long a Life Expectancy so lowincome individuals and minorities. Payroll taxes if you increase those they tend to affect everybody acrosstheboard from all demographic groups. Attacks which is something that jeff didnt cover here today tends to impact high income individuals for the most part. What we looked at first which we found really interesting is voting patterns. We looked at it from a number of demographics and perspectives. The first perspective was how people vote according to income. This graph shows basically that lowincome people, people who have earned less than 50,000 a year even though they are a larger share of the population, they are twice as likely not to vote as people who earned 75,000 or more. From a more political point of view one could argue that any changes that affect that demographic maybe have less of an impact from a political point of view so that was one aspect that we look at in terms of voting patterns. On this slide you can see another thing that we look at what age and the impact of age and its ideology on voting patterns. In terms of the age younger people make up the largest share those that would be most affected by Social Security trust fund running out of money. They are the ones that are going to be the most effective. They have the largest share of the vote so you would think that they would have the greatest impact in terms of a voting pattern. Interestingly enough for survey data what we find is across the ideological and demographic spectrum whether you are republican, democrat, high income, low income whatever your age is people all seem to say even if this change affects me i still think in order to save the system that we should do it. Here you see 74 of wrapup republicans feel that raising the Social Security tax even though well have the biggest impact on them potentially they are in favor of it. So what do we suggest taking together Just Research in terms of the behavioral response to it as well as all of the other aspects of voting in what people say. I guess we would propose solutions that are in two parts. The first is either raise or eliminates attacks and interestingly enough even though in any given year that only affects 6 of the population over the course of ones working lifetime 20 of the population will hit that tax gap and we think that is clearly one of the parts of the solutions that obviously has a greater impact on high income individuals. At the same time we think that raising the retirement age also should be considered in here and from a political point of view even though it impacts people who are in low for incomes they tend to vote less from a political backlash may be less as a result. And then finally is it feasible . Is impossible to actually do this and we do think it is possible. Back in 1983 the last time that they raise the retirement age and made changes in the Social Security program they raise the fica limit to 90 income limit and thats because of the increases in wages over time because of the fica tax has gone up every year. Hasnt kept up with wages so now its only 83 . Again i think this is something across the demographic spectrum people support this so i think its something that could happen. And then with regard to raising the retirement age and National Poll found that basically once people were educated about the different impacts and heard arguments on both sides of the retirement age basically eight out of 10 people were for raising the retirement age to 68 this was a few years ago. Again you have to layer and all of the different aspects, behavioral aspects as well as what people say