90s the republicans have been trying to get dynamic scoring in after the one to 94 elections. Its always been an option to have it in addition to the projections that are already provided by cbo, right . The article use dynamic scoring for literally every bill instead of the magnitude we discussed instead of the regular scoring . Or are they going to be cases where were still going to also use the regular scoring as well . I think on the basic question ill say what i know and then we can ask the jcp people later. Basically, the rule says if the bill is large enough, jcp and cbo should dynamically score it in the official score to the extent practical. And i dont think anybody knows exactly how thats going to work in practice. But its required for those big bills in the house. And i would supplement that with i think the suggestion that you provide the dynamic score, but also the information what it will look like is perfectly sensible so youll know. On the issue of debt, this is the issue of what do you do with um, offsets to increasing policies, big spending increases or tax cuts that dont balance over the long term. It does have to balance over the long term, and were proving as a country we have no interest in balancing over the long term, so im interested in at least doing it in the models. Thats not quote easy. The point is there are lots of things that are going to have to be decided on operational things like what do you do with the fed . Well put a taylor rule in and have them use the same taylor rule no matter what. What do you do with your deficits . Your 20 that you always trigger do it the same with every piece of legislation. These are things that i think are genuinely real and have to be dealt with but they dont disqualify dynamic scoring from providing some Information Congress could use. Anyone have a Quick Response before we turn it over to the next panel . I agree with doug, except i would do the deficit offset in year five. Okay. Bills going to introduce the next panel, and were going to hear there jcp, so thank you very much, guys. [applause] all right, thank you. We are delighted to have members of the jct staff here as well as the director of the jct. But in any case were going to hear from pam [inaudible] and nick ball on how what is that tom hanks movie how you do the best thing you do . [laughter] anyway, im not sure whos speaking first, but pam and nick pams speaking first for 15 minutes and then nick will speak for 15 minutes. Without further ado, let me introduce pam. Hi. Well, thank you all for coming out in this awful weather. I was a little worried. It just occurred to me at the last minute that i didnt have a joke to start with, but then len put me in mind of an oldie but goody. What would happen if all the economists and all the weather forecasters switched places . Nothing. [laughter] so i think thats where ive learned kind of the philosophy that appears to be underlying a lot of the questions that are going on here today. What nick and i are going to do and this goes to dades introduction davids introduction is were not going to talk about the same things that we usually talk about because weve all heard it a million times. Im going to be providing some history to remind people of where we are in terms of modeling and modeling research, and then nick is going to provide a lot of informationing about the detail that goes into characterizing policies. And we are going to talk about those two rules and what the differences are between them, so with respect to the discussion that just happened, that will happen at the end when nick talks. So just a reminder a model is a very simplified view of the world, and thats where a lot of the concern comes from. The modeler has to make choices about what aspects of the economy they need to make sure to model carefully and what aspects they can simplify away from, because no model can solve if youre going to try to include everything about the economy. So jct started working on deciding about modeling, on deciding what type of models to use with this symposium, and there are people in the room people on the panel that were part of symposium in 1996. And you can find that pamphlet on our web site to get. I still think it kind of sets the table for all the discussion that came later so if you want to get really in the weeds, read the pamphlet. What that symposium did was it invited nine groups of modelers to analyze same sets of proposals and to the extent possible analyze the same sets of proposals assuming the same things about the current economy. We had three overlapping generation models, three infinitely live agent models and three macroeconomic models. And the policies, this was driven by what was the interest in tax reform at the time, consumptionbased reform. So we had a unified income tax. Some people think of it as corporate inte graduation. And is we had a vat tax, and then we had variations on them with various transition relays. And these are just things from the pamphlet, i dont expect you to absorb them completely but we had a whole big range of results. We also tried to summarize parameters that went into the different models, and the takeaway we had from these was there was a huge difference between the results of the models even though they were in theory analyzing the same proposal and in theory starting with the same start ising assumption about the economy starting assumption about the economy. So in the short run in the vat it was predicted that gdp would decrease by 4. 2 all the way up to increasing by 6. 4 . 16. 4 . A lot of the models were more geared toward longer run analysis. The ones that could produce long run kind of settled down to a narrow range of 1. 77. 5 . Its kind of going back to thinking about the concern about camp macro which results in 1. 6 and 1. 5. And not all models could model the short run. Also not all models could model the long unrun. And the thing about the other side on the parameters, there was less variation in parameters per model than there was in the gdp results. So what did we take from that Going Forward for developing our models . Well, a modeling framework meaning is it an overlapping generations model is it an econometric model matters. So the choice of parameters the choice of how sensitive you assume labor and capital and various other things are to the tax matters. Some models model Monetary Policy. That matters. It was a bug deal in the con a big deal in the consumption tax. And this was more of a surprise i think academicians knew all those other things. Characterization of present law matters. A lot of we discovered this symposium met for three or four meetings before the final result, and the results were more broadly skewed. In the early meetings than they were in the end. And part of that was because of everybody not even having the same understanding what perfect law is. And the details of the proposal matter. Some modelers were very surprised to find that when they put Transition Relief in the results changed a lot. So after that initial bit of learning that we did, the jct staff went out and selected a couple models to work with and then we started working with them and presenting analysis that we did with them to a lot of different groups. The criteria that we ended up with for our models was that they should reflect, first of all to the extent possible, the state of the art of macro modeling and the academic literature. However, we had to take into account several practical considerations. First, there are time constraints for producing estimate thes. So the real estimates. So the real state of the art in the economic literature then and even today is very fancy, compute bl general equilibrium models. And the fancier they are the longer they take to solve. So the model that is the technical wizards have out there today and earlier ones had back then would take two weeks to run through one simulation. Obviously, when, you know and then, if you get weird results you have to start over for another two weeks before you even get your first result. So we cant quite be at the cutting edge because we dont have that kind of time. We also wanted to be able to produce a range of results because there is this divergence in the literature. So we had several models. And most importantly given where we are, we need to be able to make sure we have the tax sector characterized correct. So i emphasize that again models should have as much tax detail as possible. Academic models dont have to. Lets talk about house rule that weve been operating under since 2003. Its required that we provide a macroeconomic analysis of the effects of the proposal on gdp, labor, capital and revenues, basically, of any bill that comes up that is reported by the ways and Means Committee to the house floor. And so weve done that. Now, as it turns out, the vast majority of bills that get reported out of the ways and Means Committee to the house floor are very small. Theyre so small that showing gdb effects within reasonable gdp effects within reasonable rounding you get zero. So for those bills we have a statement that says results are too small to report. Now, there have been ore other proposals that we suspect would have a measurable effect but our models havent been configured to take them into account. For example, there have been some models that had a lot of attempts to reform International Tax laws. In those cases we had a Qualitative Analysis. Informed to the best we can with our models, but we dont since our models we dont think, or the we dont think the academic literature has enough research to tell us quantity we dont try to give quantities. And then finally, we have full scale [inaudible] so currently, the models that we use for macro analysis are something we call structural macro ally rib equilibrium models which we refer to as meg, and weve been working on and off with a dynamic general equilibrium model, and you can see descriptions of these models on our web site. We have a tab or a link to macroeconomic documents. The models that we use to an lose representative Tax Reform Act of 2014 were the meg model and the olg model. Im going to tell you a little more about them. They both have basic neoclassical foundations with mainstream, that comes from the mainstream of economic literature; consumption follows a life cycle pattern labor supply responds to marginal and average changes and aftertax wages, saving and consumption respond to aftertax return savings and aftertax income, Business Investment responds to the effective return on investment and to something called the aftertax cost of capital which is, um, taking taxation of capital into account. And that in part depends on the vault of savings. Both models do have Crossborder Capital flows so that net exports affect the domestic economy. There are Exchange Rate equations in them. So lets talk about the difference between the two models. In the meg model, we do have in the long run equilibrium demandadjusted supply but in the short run welcome we can allow unemployment. And that turned out to be very important because weve had to analyze bills that were short run bills. Our behavioral equations are structural meaning we use elasticities that come from empirical measurements. We divide our labor supply into four categories; high and low primary earners high and low secondary earners. The purpose is they tend to have different responses to tax changes and often different proposals affect them differently. And we, one of the things we discovered with all our experimentation is when you separate that out, you can get a very different answer relative to if you just use one tax rate. The other key thing about the meg model is that people are myopic meaning they know what the economy looks like today when theyre making their decisions, they dont know what its going to look like telephone years from now. Ten years from now. Now, what this does is it enables us to model policy changes that have a growing deficit. So the discussion we had before about what do you do about assuming the debt, fortunately in this model with we dont have to submit. One thing it can tell us because it does solve out to the future is where the economy blows up, and thats a piece of information as well. In contrast, our olg model is more kind of coming directly from the whats going on in academic departments. So its constructed on microeconomic foundation cans, it uses deep parameters, supply always has to equal demand. It models instead of income groups, age cohorts. And the people in the model have [inaudible] so they can look today and see a huge deficit in the future, and this is what youre always hearing about these models. Theres no rational thing for them to do, so is they dont and the model doesnt solve. And thats why there is a lot of discussion about needing some kind of fiscal closing assumption. So recently weve also added a specific Multinational Corporation sector in the olg model. Actually, we leased this from someone whos been working with that and the good thing about that is it gives us a better handle on those proposals that are designed to affect that. Ongoing Model Development we work all the time to keep up the literature, do the best we can to reflect changes. Right now were double checking some of the program fors in the olg multinational sector. Were doing our own e common metric work, see what we think of that. And were still building in house, our own inhouse olg model and a [inaudible] model. So the rest of the presentation is going to tell you why the development of your macro model is only half the story. Because getting both present law tax and tax policy right is a lot more complicated than anyone who hasnt tried to do it in the detail that we have to do it for jct where conventional estimates can understand, and were going to use the reform patch. So im showing you a couple pictures from my revenue page. Every one of those items we have to decide how to add together to put into our model. There are a lot of different deductions and credits and we have to decide how to treat each one, and that is what nick is going to talk about. [applause] we were already on the right page. Many people talk about dynamic analysis as though its impossible to do. Dynamic analysis is what weve been doing for a decade, and you can argue about whether you like the results we have or not but we think what weve been doing is fairly reasonable although we dont think were perfect and we welcome comments and discussion. Many others talk about dynamic analysis as though its a magical thing you just press a button, you could use this as realtime advice about somebodys proposed amendments. Well it doesnt work that way, but one speculative guess as to why people think it does is because its an example of clarks law. Clark is arthur c. Clark of 2001 a space odyssey, and his law says any advanced technology is indistinguishable from magic. Well, what were hoping is that after a few minutes of looking inside the hat, youll realize there is no magic button. Its hard work and thats what we do. So pams talked about a little bit about initial using models to choosing models, initializing to parameter values consistent with the economic literature, etc. We had a paper almost a decade ago where we looked at what is the impact of putting in really simple tax assumptions like one average tax rate for the whole economy or just an average and just a mar juneal or break marginal or breaking it into components that address different aspects of income. And we found that its really important to get it right for multiple sources of income. So in particular for individuals we compute average and marginal tax rate for wage and salaries in total and for the labor supply groups that pam mentioned, high and low, primary and secondary we compute average and marginal for interest dividends, Capital Gains, business income, individual return. So thats schedules c e and f. And then other. For corporations we compute average and marginal tax rates. If you ask our corporate estimator whats the average rate or the marginal rate, you spend two or three hours in a discussion about what that is. That we combine the individual rates and the corporate rate to get a weighted average business read that gets fed into the macro are models. And finally, both of the main models that we work with olg and meg, handle depreciation separately. So we talk with the conventional estimator about present value effects, liability evicts and then you have to effects and then you have to back out whats the applied change in capital consumption consistent with the way that the models set up. Okay. So pam showed you a little bit of the 15 pages of the camp table. This is just picking one provision almost at random can the Domestic Production reduction, and the columns are itm, im, bm. So utm is itm is if this provision is on the individual tax model, then youd have an indicator there. Individual marginal, im, Domestic Production deduction obviously has an individual effect through passthrough income thats reported on individual returns. Similarly, it has a corporate effect. So that one is sort of obvious. For each provision, you know, there are four macro estimators and 15 or so conventional estimators, so we cant know the details of every possible provision. So what i or somebody else ends up doing is Walking Around the floor and talking with the estimators about any provisions that are significant enough in terms of their score that you really want to find out how does this provision work and whats it doing. For a lot of provisions its obvious whether it has just average effects or some marginal effect. But for other provisions like lastin, firstout met of invenn tour, it gets more complicated. That has large average effects, but also its has marginal effects, and you can sort of take a look at the slide and think about it a little bit more later. So at this stage we have a good idea about the details of the proposal ideally and then the question is can the existing models handle those details or do we need to figure out how we can modify the model so that it will handle it correctly . So is, for instance the first time that we modeled repeal of the home interest reduction in meg, we had to go and tweak the cost of capital equations a little bit to make sure we were modeling that correctly. Now, some provisions you might just decide you cant model it in any reasonable way and then youre sort of stuck. But you cant have, you cant make models that can handle every possible strange thing that people come up with. Okay. So for provisions that are modeled using the individual tax model, we need to compute the effect on average and marginal