of the marginal efficiency of investment shock and I mentioned that because that proved very useful and work I did with Jonathan Eaton and Brent Nyman where we needed a kind of real shock that would move investment spending around and maybe it feels too easy and yet it was kind of striking in

your pictures where you want something that makes you suddenly you don't want to invest it's not so much that you lost capital it's that you realize right then if you invested you probably wouldn't get very much so. Okay thank you three over there.

It seems to me that you are endowing your agents with a very inefficient technology to estimate tail risk. This kernel estimation I don't think any econometrician would use a kernel estimation to estimate the tail they would rather use extreme value theory and try to estimate the decay of the rate of decay of the distribution so once you do that my

feeling especially if you incorporate data up to the Great Depression is that your estimate would be much less sensitive to the tail event and would that have an implication also for the

prediction of your macro model? Also on the Great Depression. The Great Depression must have come as a surprise to those who experienced it and must have led to an update in people's perceived

distributions. Why didn't we see the same pattern after the Great Depression if you if you plot to a Great Depression you would see that you know the the economy reverts back to trend and not only that but it even overshoots it a linear trend. Okay the gentleman behind you.

Thank you I someone already said I think what we all learned was that that's why I'm convinced that's my narrative what we learned is that that we were not on the trend we thought we were on

this very controversial by some but I think it's it's compelling case and there is no so I don't really understand why you say it couldn't be a surprise about TFP I mean I think a good strong narrative is that during the boom we overestimated TFP and then we realized that we

did and now we're on a different trend now your result your construction is excellent explaining the fact that we understand now the downside of such mistake and you see it in the asset price data but I think this interpretation is is much easier to square with other facts so I don't

understand why you say you didn't want this to be a surprise about TFP. Okay thank you very much any any more questions comments I don't see any for the moment so Laura the floor is yours.

So first I want to thank Alberto for a very thoughtful a very generous discussion I think asset prices are tough to look at because they're a mix of things for remember firms here are delevering when risk goes up they're reducing debt right and so that's going to affect equity prices it's going to make equity more valuable on the other hand you've got more risk so you

got these two things moving in opposite directions it's really hard to just look at price levels for either debt or equity and figure out what's going to happen when the mix changes the truth is it could go either way with a slightly different calibration and that's why we don't want to rely on it the VIX is low the VIX is not uncertainty the VIX is volatility volatility

is not high in this model it's high at the time of the financial crisis and then it's low again the question of do we know that we don't know no in this paper they re-estimate the distribution and then take that as truth is that ideal no it would be a lot harder model to do

and I'll tell you we did some experiments with thinking about what if tomorrow you had a different distribution what if two periods from now you had a different distribution and and sampling all of those and seeing how different the policies were and there were there were very minute differences but one thing we would miss that's important and actually this is based on work Pierre Colin Dufresne here has done which is if you know

that estimates of a model are uncertain it's a source of long-run risk so if we had a model where a long-run risk were an important determinant of asset prices that's where that assumption that we are acknowledge our uncertainty about future distributions would would bite it would introduce knowledge of a new martingale what if policy beliefs

change would that change things yes and in fact that would be the role for policy here would be to have some policy that actually changed people's beliefs about the possibility of future tail events if you could implement something that credibly removed the tail risk that would be effective here but that's a big if

thank you for the suggestion about the marginal value shocks why did the Great Depression not last I'd say World War two I don't know that I have much more to say about that there there certainly was some persistence of that shock it dragged out for a while and then

we entered a pre-war phase and then lastly why why didn't we do this for TFP so two problems with TFP one is the measured productivity shocks during this period were actually rather small so we wouldn't get a lot of kick out of it and the second is we wanted this to be consistent with asset prices because we wanted to use asset

price data and options data to as as over identifying moments as evaluations of the theory and most TFP driven models are very difficult to reconcile with asset prices this but one of the reasons that Francois Gouriot wrote this model is that these capital quality shocks allow for there to be large variations in in the returns to capital that

with with not so much macro volatility and it's very difficult I don't know whether these marginal value of capital shocks do that they may I'd have to look into that but but this was specifically constructed to have some meaningful thoughts about asset prices that being said no

paper no paper explains all the ups and downs of the asset market where that's that's an impossible barrier I will not say that this tells us why we have such as high asset prices today there are a lot of things going on in the world and this is just hopefully one of them okay thank you very much any yes another one please so coming back to my point before so

I think when you say that the estimated TFP shocks were too small to have any action is because of the way they were estimated did not really allow the possibility that we had a mistake for several years running during the credit boom misunderstanding overspending with

borrow money for something that we attribute to be actually TFP so I think if we were to redo that calculation it wouldn't be clear at all the day shocks were the surprise perhaps was we thought for a number of years we were growing because there was rising productivity actually

we're just borrowing and spending now if you recalculate the path then you come to a different estimate of what might have been a mistake I think that's but I think this is this someone should do it that's possible I did not want to invent a new way of measuring TFP we

took standard measures we took utilization adjusted those those didn't seem to work what you don't know so if we have the wrong way to measure because we have the wrong expectation and the wrong inference everybody does and everybody spends and therefore if everybody were right our measure would be right then we discovered we were wrong that's the real

discovery and that's what produced the class that's something you can take you back to your

paper Laura so um so there seemed to be sort of two thoughts as you said there was maybe a minority that thought that there may be another crisis in the making and some thinking sort of about you know Ulrike Momendier's work on depression babies and the like so could we

I know that I mean you look to there were some questions already on what happens if we were to extend the period to the Great Depression but could we use your framework to actually sort of estimate how much memory loss there is in society I think you could do that no

because it's something that doesn't treat me I think there are two camps there but maybe one is very small and the other is very large but maybe you could fit the data somehow so we did some experiments where we discount old data the behavioral foundations of that are a little questionable but you know it is a question that comes up and it seemed an exercise

worth doing the short answer is about 1% per year discounting seems to generate if we then include the fact that there was there has been a financial crisis before in the Great Depression that seemed to square with the size of the decline that we saw okay so much

about Alzheimer effect of society any any more question one more please also demographics should matter in the sense that I mean if you have a reshuffling population that is growing very faster you know the average memory should become shorter and shorter so that is related also to

why the Great Depression was showing less persistent perhaps because the growth of operation was faster at the time and the other is that prior to the Great Depression there

were other rare events that were in the memory of people that were faced with the Great Depression whereas you know the this Great Recession was preceded by the Great Moderation so people were forgetting because of the Great Moderation so I think all these features seems to be consistent thank you on the Great Depression in the equity premium

if you compute the ex-ante equity premium from 1929 on basically went up enormously just after the great new after the fall in the stock market it's between eight and ten percent at

time and then it slowly went down for about 30 years and there's a well-known saying by Paul Samuelson which is that basically you know the memories of the equity premium disappeared one death at the time and so I think it's very relevant to what you discussed okay thank

you very much I don't see any more so I think both Laura and Alberto for the paper and for the discussion and all of you who asked questions and so I think we learn how transitory even very small unlikely events but the extreme events can change beliefs and also change macro outcomes

but at least this institution as it was said as a policy institution tries to change the beliefs of the people again and via statisticians hopefully supply all the data to change the beliefs in this sense so that you don't have wrong beliefs with that we move to coffee break

there are 22 minutes till 1645 and the next session will start so enjoy your coffee and thanks again for for your questions and contributions