Inequality and Asset Prices during Sudden Stops

videoType

dynamic-media

videoDmUrl

keyFrameImage

timestamp

00:00 Crises have happened a lot.

00:03 Just in the last 30 years,

00:05 there have been more than 50 sudden stop

00:07 episodes both across emerging and advanced economies.

00:11 Now these episodes are characterized by uh declines in

00:13 asset prices which we know that affect households differently.

00:17 To give you an numbers about this,

00:20 during Mexico's 2009 crisis,

00:23 the share of high leveraged households,

00:25 these are the households that have a lot of debt relative to their assets,

00:29 decreased the most their expenditures,

00:31 around 6.2%.

00:33 While the share of non-leveraged households,

00:36 the ones that do not have any debts,

00:38 Uh,

00:39 decrease,

00:39 increase their expenditures by around 5%.

00:42 This makes sense because these are the households that are,

00:44 are least exposed to tightening financial conditions.

00:48 But the heterogeneous effects are not only from the aggregates into the,

00:52 the,

00:52 into the households.

00:53 This heterogenetic component also will play a role

00:57 in determining the aggregate effects of a crisis.

01:00 And to be specific,

01:01 there are going to be two effects that affect the

01:03 severity of a financial crisis that go in opposite directions.

01:07 We're going to have a dampening effect and an amplifying effect

01:10 which will depend on how the leverage and the wealth distributions are in,

01:15 in,

01:15 in an economy before the crisis.

01:18 So to be specific,

01:19 in this paper,

01:20 I'm gonna ask the following question.

01:22 How is the,

01:23 which are the effects that different degrees of

01:25 inequality have on the severity of crisis?

01:28 More particularly,

01:29 I'm gonna look at

01:31 the

01:31 cross-sectional dimension

01:33 of Fisher's debt deflation mechanism.

01:35 This mechanism is the standard financial friction that triggers endogenous,

01:40 uh,

01:40 endogenously sudden stops

01:42 as a equilibrium,

01:44 uh,

01:44 as equilibrium,

01:45 uh,

01:45 choices,

01:46 as equilibrium behavior in the agents.

01:49 And to be precise about this dimension,

01:52 I'm gonna,

01:53 the two main contributions of my paper are first,

01:55 I give empirical evidence that such a

01:57 dimension operates via these two opposing effects.

02:02 To,

02:02 to,

02:02 to,

02:03 in a couple of slides,

02:04 I'm gonna be more specific about how the

02:05 mechanism works and how these two dimensions operate.

02:08 But uh,

02:09 uh,

02:09 as a summary,

02:11 there,

02:11 there's gonna be this amplification

02:13 coming from the high leveraged households that have to fire sale their assets

02:18 and

02:19 amplify the drop in asset prices.

02:22 And there's gonna be a dampening effect on,

02:24 on,

02:24 on the severity of the crisis

02:26 by which wealthy,

02:28 low leveraged households

02:30 can buy

02:31 some of these assets,

02:32 accumulate more of them,

02:33 and this dampens the,

02:34 the aggregate effect.

02:37 The second question,

02:39 there's a quick clarification question.

02:40 Can you define a sudden stop crisis?

02:43 Definitely,

02:44 Pierre.

02:44 Thanks.

02:44 Uh,

02:45 so,

02:46 a,

02:46 a sudden stop is an episode in which international capital

02:49 stops flowing into an economy

02:52 and

02:52 if the domestic economy is sufficiently indebted,

02:55 this generates a domestic credit crunch.

02:58 So we can think that

02:59 an aggregate shock,

03:00 let's say an interest rate price increase or a drop in TFP

03:04 or a change in terms of trade or a combination of all of these aggregate shocks.

03:08 If the,

03:09 if the domestic economy is in a,

03:11 a state of the world that in which it's sufficiently indebted,

03:14 depends on,

03:15 on,

03:16 on foreign,

03:16 on foreign

03:17 resources on foreign capital,

03:19 this generates a,

03:20 a,

03:20 a sudden stop.

03:21 Uh,

03:22 the definition has been,

03:23 uh,

03:23 I think it was introduced by,

03:25 by Calvo,

03:26 by Guillermo Calvo,

03:26 but it's a,

03:27 a,

03:27 a fairly standard,

03:29 uh,

03:29 definition of,

03:30 of this type of crisis.

03:33 So,

03:34 the,

03:34 the second contribution of the,

03:35 of the paper is that standard models in the literature

03:39 cannot account for,

03:40 for this cross-sectional facts or the distributional effects.

03:44 Mostly because these,

03:45 these papers use representative agent frameworks.

03:48 And this is where the model of my paper enters.

03:50 I propose a small open economy framework

03:53 with a rich heterogeneity across the households.

03:56 It's gonna be a purely economy

03:58 and the households are gonna be

03:59 exempted homogeneous and they will get exposed to heterogeneity

04:03 coming from uninsurable idiosyncratic shocks.

04:06 And on top of that,

04:07 I will add aggregate risk to analysis in the sense that I

04:10 will not only look at comparisons between steady states or transitions,

04:15 I will look at the full stochastic stationary equilibrium of the model with,

04:18 with risk.

04:19 So,

04:20 uh,

04:21 the main results of the model are that

04:23 it can explain key stylized facts or aggregate stylized facts on solid stocks.

04:28 But,

04:30 uh,

04:30 as a contribution compared to uh,

04:32 uh,

04:32 the representative agent version of this model.

04:36 The model with heterogeneity predicts that sudden stops are persistent.

04:40 It will take more than 3 years or periods in the model to

04:44 the current account reversal and the drops in

04:46 consumption to come back to their pre-crisis levels.

04:50 And this is something that is precisely driven by 380.

04:54 When,

04:54 when the crisis hits,

04:56 financially constrained households will start at the leveraging process

04:59 in which they will start uh

05:02 fire selling their assets

05:03 and due to the adjustment costs in the different allocations of,

05:07 of these portfolio choices,

05:09 this process will take time.

05:10 So this introduces the persistent element.

05:13 This is something that the reb agent version cannot account for.

05:16 Precisely because when every household behaves as the average household,

05:21 then

05:22 when everyone wants to sell part of their assets,

05:24 there is no one that wants to buy that asset.

05:27 So,

05:27 this model becomes very similar to a Lucas ree type of framework in which

05:31 there is effectively no trading

05:33 and the asset price has to absorb all of the shock.

05:36 So in that way,

05:37 the model with heterogeneity

05:39 will predict that asset prices drop less,

05:41 the dampening effect will dominate.

05:44 Other consumption,

05:45 consumptions are gonna drop more

05:46 and this larger drop aggregate consumptions in

05:49 aggregate drops in consumption will come from

05:52 the

05:53 high leveraged households which have the most,

05:56 the largest marginal propensities to consume

05:59 and are the ones that are get that get

06:01 uh affected by the tightening financial conditions the most.

06:04 So that's why we get larger drops in consumption which are consistent with the data.

06:10 Lastly,

06:10 in,

06:11 in,

06:11 in the last part of the paper,

06:12 I will,

06:13 now,

06:13 I will be ready to use the model to quantitatively

06:16 assess what are the effects of different degrees of inequality.

06:19 So,

06:20 uh,

06:20 as a motivation,

06:21 I will compare

06:23 two different calibrations of the model,

06:25 one for an emerging economy in Mexico and one for an advanced economy

06:29 in which I reduce

06:30 the wealth inequality.

06:32 And I find that the model predicts that

06:34 more unequal economies experience more severe sudden stops.

06:38 So,

06:38 in this exercise,

06:40 it's a,

06:40 uh,

06:40 what I'm doing is I'm

06:42 uh changing a parameter that drives a

06:44 mean preserving spread on this inequality component

06:47 and we find that the amplification effect dominates.

06:51 The last part as a policy analysis motivated by the fact

06:54 that an economy or a government

06:57 cannot wake up one day and change the level of inequality in the economy,

07:01 I introduced a dividend tax that precisely reduces this,

07:04 uh,

07:05 this inequality component coming from the dividend risk.

07:08 And I find

07:09 that the,

07:09 the tax policy

07:11 is very effective in prevent,

07:12 in preventing consumption drops,

07:14 mostly because the government is ensuring

07:17 uh this highly leveraged households.

07:20 But it's ineffective and it is completely ineffective to prevent

07:25 asset prices drops

07:26 precisely because these leveraged households are still fire selling.

07:30 So they,

07:30 they get the benefits in terms of consumption

07:32 but they are still fire selling their assets.

07:36 So now,

07:36 uh,

07:37 let me,

07:38 let me describe to you the,

07:39 in a nutshell,

07:40 which is this Fisherian debt deflation mechanism that triggers so stocks.

07:45 Imagine,

07:46 let's take an open economy,

07:48 Mexico for example,

07:49 and suddenly the United States increases their interest rates.

07:52 This is an aggregate negative shock.

07:54 The share of financially constrained households in Mexico

07:58 will,

07:58 would like to smooth their consumption

08:00 with debt,

08:01 but they can't because they are up to their limits.

08:03 So they will start fire selling their assets.

08:06 Now,

08:06 this behavior

08:08 introduces additional downward pressure on the asset price

08:11 and

08:12 uh uh the collateral constraint of everyone starts to tighten.

08:14 So financial conditions tighten for everyone.

08:17 Now,

08:18 more agents will start to fire sale of more assets.

08:22 And this feedbacks as additional,

08:24 this,

08:24 this feedback loop

08:26 gives additional downward version asset prices.

08:28 So that,

08:28 that is the mechanism.

08:30 Now,

08:31 This has been well studied in the literature in a representative agent version,

08:36 but there are going to be two cross-sectional effects.

08:39 I,

08:39 I already previewed them,

08:40 but let me be more specific here.

08:42 The first one is a dampening.

08:45 When we think that in this uh leverage distribution that any economy,

08:49 leverage distribution across the households that any economy has.

08:53 There's gonna be a dampening effect coming

08:55 from the share of unconstrained wealthy agents

08:58 that

08:59 have the resources to bear more risk

09:01 and take advantage of the depressed prices,

09:04 so they start accumulating these assets

09:06 and this relieves the downward pressure.

09:09 On the other end,

09:11 there's gonna be the share of financially constrained households

09:14 that as the mechanism,

09:15 the debt deflation mechanism is triggered,

09:18 they end up fire selling more assets

09:21 and more households end up constrained.

09:23 So this creates kind of a snowball effect

09:25 in which

09:26 the,

09:26 the,

09:26 the first households that are constrained

09:28 fire sale their assets.

09:31 Asset prices are dropping,

09:32 so more households end up constrained and so on.

09:36 This strengthens the downward pressure.

09:38 So as a graphical intuition,

09:39 let's think that

09:41 there are two economies

09:42 which have different,

09:43 uh,

09:43 different distributions on the leverage of their households.

09:47 The first one is the solid blue line,

09:50 which is,

09:50 uh,

09:51 which is represented here and then a second economy represented as the dashed line

09:56 which is more unequal.

09:57 There are,

09:58 there are fatter tales

09:59 on the constrained households,

10:01 but there are also fatter tails on the wealthy households.

10:05 Ex ante,

10:06 it's ambiguous which of these two effects will

10:08 dominate in the sense that we don't know

10:10 which of,

10:11 in which of these two economies,

10:12 the crises are going to be more severe.

10:14 So I will propose this model to as a,

10:17 as a

10:17 measurement device

10:19 to quantify the,

10:19 the effects.

10:22 So now,

10:23 uh,

10:24 let me,

10:25 uh,

10:25 in the interest of time,

10:26 let me skip,

10:27 uh,

10:28 Sergio,

10:28 we have one question.

10:29 Maybe that's a good time to take the first question,

10:31 then I have one for you too.

10:32 So Alvaro,

10:33 please go ahead.

10:35 Yeah,

10:35 thank you.

10:36 So just a quick question,

10:37 in,

10:38 you're thinking of fire sales of assets,

10:40 but in most,

10:41 most countries,

10:42 the largest asset of households are,

10:45 you know,

10:45 housing actually.

10:46 So

10:47 how do we think about that here?

10:49 I mean,

10:49 uh,

10:50 You,

10:50 you told us about the data in Mexico where households that are higher in debt

10:55 in debited

10:56 lower consump lower consumption,

10:58 but do they actually go out and sell the house or their biggest asset or like how do we,

11:02 we think about the,

11:03 the mechanism here in terms of the asset?

11:05 What,

11:06 what,

11:06 what asset do you have in mind?

11:08 Yes,

11:08 Alvaro,

11:09 perfect.

11:09 Uh,

11:09 so that,

11:10 thank you for your question,

11:11 Alvaro.

11:11 That's a,

11:11 that's a very,

11:12 very important point.

11:14 Uh,

11:15 here I am talking about assets in a general way,

11:18 but when I,

11:19 when,

11:20 as you well noted,

11:21 most of the asset

11:23 portfolio,

11:24 the,

11:25 the,

11:25 the,

11:25 most of the asset wealth in terms of,

11:27 in the households

11:29 corresponds to real estate.

11:31 And what I'm going to show you in the microdata that I have for the Mexican economy

11:36 is that in fact,

11:37 this,

11:37 uh,

11:38 depending on the group,

11:39 or which the households correspond in terms of leverage and wealth,

11:43 we're going to see that in fact they are reducing their

11:46 asset positions in terms of real estate.

11:48 So,

11:49 uh,

11:50 This,

11:50 this,

11:50 this means that they are either selling the

11:53 part of the real estate,

11:54 they are selling the current state and,

11:56 and moving to a smaller house

11:58 given the financial conditions,

11:59 tightening financial conditions,

12:01 but,

12:01 uh,

12:02 but you're right.

12:03 Uh what,

12:03 what I'm gonna show you is for real estate

12:06 assets.

12:07 We,

12:08 but also if,

12:08 if I,

12:09 uh,

12:10 account for

12:10 the whole,

12:11 uh,

12:11 portfolio of,

12:13 of the households,

12:14 I include financial,

12:16 financial assets or I include

12:17 durable consumption goods,

12:19 the,

12:19 the,

12:20 the,

12:20 the results are gonna be robust.

12:22 But

12:22 I think this will be more clear in,

12:24 in the next in the next two slides where I show you the,

12:27 uh,

12:28 the,

12:28 the,

12:28 the,

12:28 the empirical evidence and in,

12:30 in particular the data that I am using.

12:33 But that's a good question.

12:35 Thank you.

12:36 Uh,

12:36 uh,

12:36 Art,

12:37 uh,

12:37 go ahead,

12:37 please with your question.

12:40 Yeah,

12:40 hi,

12:40 Sergio.

12:40 Um,

12:41 I,

12:41 I had two questions.

12:42 One was actually a little bit of follow up on what,

12:44 um,

12:45 uh,

12:45 Alvaro asked a moment ago,

12:46 which is that if you take his point even further,

12:48 in poorer countries,

12:49 the only

12:50 sort of asset of a lot of households is actually people's human capital.

12:54 And then

12:54 does that completely erase this mechanism,

12:56 or how do we think about it,

12:58 uh,

12:58 because there's obvious limits on,

13:00 you know.

13:01 Uh,

13:02 trade in that case.

13:03 And the other one is that in your introduction,

13:05 you referred to how the

13:07 initial level of inequality essentially governs the,

13:10 um,

13:10 you know,

13:10 whether the dampening effect or the amplifying effect

13:13 sort of dominates,

13:14 and I see the intuition for that.

13:15 But

13:16 does your model also have implications for how,

13:19 um,

13:20 uh,

13:20 asset inequality and consumption inequality evolves following the shock?

13:24 Um,

13:25 I mean,

13:25 it seems clear from this mechanism through the dampening

13:27 effect at least that you would get increased.

13:29 Asset inequality.

13:31 It's less clear what's happening with consumption inequality.

13:34 I see.

13:34 No,

13:35 uh,

13:35 thank you,

13:35 Art,

13:35 for the question.

13:36 Um,

13:37 so yeah,

13:38 you,

13:38 you're right in that,

13:39 uh,

13:39 I,

13:40 I.

13:41 I,

13:41 I completely agree that in,

13:42 in,

13:43 in some economies,

13:44 uh,

13:44 human capital would be the,

13:46 the,

13:46 the,

13:46 the largest uh assets for some part of the household distribution.

13:50 And

13:51 I haven't thought about that,

13:53 but as,

13:53 as a first thought,

13:54 I would say that

13:56 There,

13:57 there are no fire sales in human capital.

13:59 As you will say,

14:00 there,

14:00 there,

14:00 there's no trade,

14:01 there's no market for that type of,

14:03 of asset

14:04 or,

14:04 or for that

14:05 type of

14:06 human capital.

14:06 So,

14:07 um,

14:08 I,

14:09 I,

14:09 I,

14:10 I would say that

14:11 This,

14:12 this,

14:12 this,

14:12 this will only,

14:13 this will,

14:13 this will be general in terms of assets that can be traded within the column.

14:18 So that,

14:18 that will be the first part.

14:19 For the second,

14:21 uh,

14:21 you,

14:21 the,

14:22 the,

14:22 the implications on,

14:24 on,

14:24 on other types of inequality measures like

14:26 in consumption or assets after the crisis

14:29 are,

14:29 are super interesting.

14:30 In particular,

14:31 I,

14:32 I will document

14:33 how the Gini coefficient,

14:35 which measures how unequal uh the economy is in terms of the consumption.

14:39 Uh,

14:40 is behaves before,

14:42 during,

14:43 and after the crisis,

14:44 and the model will have,

14:45 uh,

14:46 will,

14:46 will be able to,

14:47 uh,

14:47 rationalize these quali qua qualitative movements

14:50 on the inequality of consumption.

14:52 Quick preview.

14:54 During the crisis,

14:55 consumption inequality dropped.

14:58 And this drop is mainly driven by the right tail of the distribution.

15:02 Uh,

15:02 excessive lux luxurious,

15:04 uh,

15:04 consumption coming from these highly leveraged households that

15:07 are the ones that reduce the most consumption,

15:10 uh,

15:11 translates into lower inequalities in terms of consumption.

15:14 And that is something that I will capture with the most.

15:18 One more question from uh Carmen.

15:20 Carmen,

15:20 please go ahead.

15:26 A

15:27 couple of,

15:28 a,

15:28 a,

15:28 a question and a comment on the dampening effect.

15:33 Suppose indeed

15:35 Abstracting from Art's question,

15:37 suppose indeed that households

15:39 do sell

15:40 in the middle of a fire sale.

15:43 Their homes.

15:45 Uh,

15:45 which to some degree

15:47 happened quite a bit in the 1995 crisis in Mexico.

15:51 OK,

15:51 but suppose on the buying side,

15:55 The dampening effect is muted not because

15:59 the high income households.

16:01 do not necessarily want

16:03 the fire sale assets.

16:05 They want,

16:05 say,

16:07 Uh,

16:08 assets

16:10 outside the country,

16:11 you know,

16:12 namely,

16:12 you know,

16:13 you,

16:13 you're in a sudden stop mode,

16:16 um,

16:16 US assets,

16:17 so

16:19 Uh,

16:20 the wealthy households

16:22 choose not to buy the domestic fire sale assets but opt to buy

16:27 foreign,

16:28 uh,

16:28 some,

16:29 any kind of,

16:29 uh,

16:30 foreign asset.

16:31 In that case,

16:33 your dampening effect

16:34 isn't,

16:36 you know,

16:36 the initial conditions may be there in terms of inequality,

16:39 but

16:40 you don't necessarily get the kind of dampening effect.

16:45 Uh,

16:45 that you're suggesting.

16:48 And this is,

16:48 this is the question.

16:50 So

16:51 what if,

16:52 uh,

16:53 that is the case?

16:54 Secondly,

16:55 wouldn't the dampening effect not be

16:58 needed?

16:59 Uh,

17:00 to be rooted in inequality if

17:03 a central bank.

17:05 Like

17:06 the asset purchase program of the Fed in some,

17:10 of course,

17:10 different scale,

17:11 whatever.

17:12 But

17:13 if a,

17:14 a,

17:14 a public sector player.

17:17 Come in and does

17:20 the actual purchase

17:22 of fire sale

17:24 uh assets.

17:26 In that case,

17:28 Even if you have very limited

17:30 inequality to start with,

17:31 you still get a dampening effect

17:33 through

17:36 a third party.

17:37 In this case,

17:38 a central bank or,

17:39 or some sort of government agency that is,

17:42 is,

17:42 is buying the distressed assets.

17:45 I see.

17:46 Uh,

17:46 thank you,

17:46 Carmen,

17:47 for,

17:47 for the questions,

17:48 and,

17:48 and I think

17:49 both of them are,

17:50 uh,

17:50 or,

17:51 or,

17:51 it's one in which the core,

17:53 the core question is about

17:56 Who is buying this,

17:57 this,

17:57 uh

17:59 Depressed price assets.

18:01 In,

18:01 in,

18:01 in the first part,

18:02 if,

18:03 if the wealthy own constrained households are

18:06 interested in buying foreign assets,

18:08 then

18:09 we will lose part of this dampening effect.

18:11 The,

18:12 the,

18:12 the,

18:12 the,

18:13 the story I think here,

18:14 what,

18:14 what is relevant is that

18:16 these domestic households

18:18 that are risky and unleveraged.

18:20 know

18:21 that

18:22 the price that they are observing right now on real estate

18:26 is due to

18:28 fire sales.

18:28 So it's,

18:29 it's,

18:29 they know that at some point,

18:32 their wealth is gonna increase

18:34 by buying cheap right now.

18:36 So

18:36 this is the story of

18:38 when is a good time to buy assets

18:40 during a crisis because they are cheap.

18:41 So that,

18:42 that is,

18:42 that is what I'm thinking when

18:44 I say that unconstrained wealthy households will buy these domestic assets

18:48 because they know that right now they are very depressed.

18:51 And their

18:53 future prices or fundamental prices are going to be larger.

18:56 Now,

18:56 as,

18:57 as you will say or,

18:58 or proposed

18:59 an,

18:59 uh,

18:59 uh,

19:00 asset,

19:01 asset purchase program from central banks

19:04 would be a,

19:05 a,

19:05 a,

19:05 a very effective policy

19:07 in,

19:07 in,

19:07 in,

19:08 in,

19:08 for future uh research on,

19:10 on,

19:10 on,

19:11 on,

19:11 on analyzing the effects that

19:13 this policy can,

19:14 can

19:15 uh prevent this deflationary process

19:18 and,

19:19 and counter and,

19:20 and in,

19:20 in a way stop

19:22 the,

19:23 the,

19:23 the fishery and deflation mechanism.

19:25 Uh,

19:26 so,

19:26 so indeed,

19:27 if,

19:28 if it,

19:28 if,

19:28 if the

19:29 dampening effect could come from these wealthy unconstrained households or it

19:33 could come from another player that is the central bank.

19:37 Uh,

19:37 that is,

19:37 uh,

19:38 uh,

19:39 for the research that I am,

19:40 I'm thinking about,

19:41 but it's out of the scope of,

19:42 of this current paper.

19:44 Here,

19:44 I'm,

19:44 I'm only focusing on,

19:45 on,

19:46 on the effects

19:47 coming from the same households.

19:51 Great.

19:52 Uh,

19:52 so,

19:53 now,

19:53 let,

19:53 let me move on.

19:54 Uh,

19:54 um,

19:55 so

19:56 with,

19:56 uh,

19:57 in the interest of time,

19:58 I,

19:59 we'll skip the literature review.

20:01 It's,

20:01 I'll just summarize it in the sense that this paper contributes to three strands.

20:06 Uh,

20:06 the,

20:06 the first one is about

20:08 prices

20:08 triggered by debt deflation mechanisms in representative frameworks.

20:13 Then,

20:13 uh,

20:13 the second literature on asset price determination and equity premiums

20:17 with uh

20:19 individual incompleteness market,

20:21 individual incomplete markets,

20:22 it means that

20:23 households are going to be heterogeneous.

20:25 And the third one is the one that merges these two strands

20:29 in and analyzes the effects of inequality or

20:33 the effects of heterogeneity into a macroeconom.

20:36 So now,

20:37 uh,

20:37 before going into the model,

20:39 let me spend two slides showing you the empirical evidence.

20:43 So,

20:44 Mexico in 2009 had a severe crisis.

20:48 Uh,

20:48 the MP spread that measures the,

20:50 uh,

20:50 sovereign risk on the interest rate increased 200 basis points

20:54 and aggregate consumption decreased 5%.

20:56 So even for an emerging economy like Mexico,

20:58 these are big movements.

21:00 This was a,

21:00 a,

21:00 a severe,

21:01 it's not the standard real business cycle

21:03 uh movement.

21:04 It's a,

21:04 it,

21:05 it was a severe so stop.

21:07 But when we look at the microdata,

21:09 we see differences depending on the composition

21:11 of the balance sheet of the households.

21:13 And to show you this,

21:15 what I'm going to do is I'm going to use a panel survey

21:18 for households

21:19 for Mexico in which I'm going to follow them

21:22 before and after the 2009 crisis.

21:25 And I'm gonna sort the households according to their leverage ratio

21:29 in the uh here in,

21:30 in the rows

21:32 in deciles

21:33 and according to their net wealth

21:35 in columns.

21:36 And I'm going to see how their asset positions change before and after.

21:41 And what we see are 3 things.

21:43 First,

21:44 All of the action is happening on the top decile,

21:47 uh,

21:48 uh,

21:49 top decile in terms of wealth.

21:51 So this is expected as some of the questions already noted in the introduction.

21:56 Uh,

21:56 wealthy,

21:57 some of,

21:58 some who have low wealth households do not have much assets,

22:02 and it's not that they do not have much of them.

22:04 There's not much trade.

22:06 So,

22:07 uh,

22:08 all of the action happens in the top wealthy,

22:10 uh,

22:12 deci in terms of the households.

22:14 And then the second two things that I would like to highlight is that

22:17 If we focus on the very high leveraged households,

22:20 these are the top 10,

22:22 uh,

22:22 the 10th decile in terms of leverage.

22:25 We see that these are the households that decrease the most their asset positions.

22:28 These are going to be proxies to

22:30 precise quantities.

22:32 It's not,

22:32 these are not gonna be,

22:33 these are going to be deflated by an aggregate index,

22:36 a house price.

22:37 So this,

22:38 this will reflect changes in quantities.

22:40 So they fire sale the most their assets.

22:43 Then moving one

22:44 deci above

22:46 to the 9th deci,

22:47 you see that.

22:49 This group also decreased their assets.

22:51 So

22:52 as as financial conditions were tightening,

22:55 they ended up constrained,

22:57 so they have to file a sale.

22:59 And when we moved to the,

23:00 to the other end of the distribution,

23:02 the low leveraged households are the ones that bought these,

23:05 these assets.

23:07 Then,

23:08 so this is about the asset composition,

23:10 the portfolio choice.

23:11 When we,

23:12 but

23:12 ultimately,

23:14 households care about their utilities,

23:15 so they care about consumption in a way.

23:18 And to,

23:19 to show you the terogeneous consumption dynamics,

23:21 what I'm gonna do here is only

23:23 uh in a one dimensional uh

23:26 Table,

23:27 sort the

23:28 the households according to their leverage ratio

23:31 and see how much they change their consumption before and after the crisis.

23:35 And here we see that uh

23:37 the more leveraged the household is,

23:40 the more uh it has to

23:43 reduce or adjust their consumption.

23:44 So this is consistent with

23:46 uh

23:47 The high leveraged households having this high marginal propensities to consume,

23:52 and they are the ones that are the most exposed

23:55 to,

23:55 to tightening financial conditions.

23:57 So even small changes in asset prices

24:00 that tighten their,

24:01 their collateral constraints

24:03 have

24:04 large effects on their

24:06 individual consumption

24:07 and they correspond,

24:08 and they contribute largely to the aggregate movements in consumption.

24:12 So that's why we get

24:13 uh

24:14 large

24:15 drops in,

24:16 in aggregate consumption.

24:19 we're ready to go to the model,

24:20 so I'll stop here in case there are any,

24:23 any more questions.

24:24 I've got one for you and then Clement has one.

24:26 Is there a little word in what you showed us about mean reversion that is,

24:30 you know,

24:30 even outside of a,

24:32 of a crisis period,

24:33 you'd see that the low leverage,

24:35 you know,

24:35 would increase a bit consumption of the assets.

24:37 I mean,

24:38 the low leverage because they have few assets to begin with and they might want to buy

24:41 a house,

24:41 right?

24:42 Yes,

24:43 yes.

24:43 So,

24:44 uh,

24:44 so I,

24:45 that's a,

24:45 that's a very good point,

24:46 Pierre,

24:47 uh.

24:48 I,

24:48 I would say in,

24:49 in mean reversing in the sense that

24:51 they're gonna be idiosyncratic shocks.

24:53 So,

24:54 some households are gonna

24:56 be

24:56 having

24:57 a good luck,

24:58 having good,

24:59 good shocks in terms of their income or in terms of their dividends.

25:03 And

25:04 probably this is what is driving this,

25:06 uh,

25:06 well,

25:06 no,

25:06 it's,

25:07 this is what is driving the enterogenetic component.

25:10 So even,

25:11 uh,

25:12 outside of the crisis or during the crisis,

25:14 these idiosyncratic shocks are gonna be driving some of the,

25:18 some of the uh movements.

25:20 Uh,

25:21 it would be

25:22 awesome to have,

25:23 uh,

25:24 a larger,

25:25 more number of waves in this panel.

25:27 I only have 3,

25:28 2002,

25:28 2005,

25:29 and 2009.

25:31 So,

25:31 uh,

25:31 a more detailed empirical analysis on how to,

25:33 how,

25:34 how this,

25:34 uh,

25:35 idiosyncratic shocks affect the households uh with the available data would,

25:40 would be,

25:41 would be great to do.

25:42 With the,

25:42 the available data,

25:43 I,

25:43 I cannot tell you more,

25:44 more about uh

25:45 uh this mean reversing type of effects.

25:52 Hi,

25:52 Sergio.

25:53 Um,

25:53 I was also wondering about uh

25:57 But more in terms of uh measurement.

26:00 So,

26:00 you know,

26:01 since you're relying on like the 10th decile of the

26:03 distributions of assets and assets are typically measured with error,

26:07 um,

26:08 can you tell us a little bit about that,

26:09 you know,

26:10 and how you would avoid that,

26:11 that concern?

26:13 I see.

26:13 Uh,

26:14 thank you,

26:14 Clem Clement,

26:15 Clement for,

26:15 for the question.

26:16 Um,

26:17 so the household survey corresponds,

26:19 is,

26:20 is,

26:20 is composed of 8,

26:21 around 8000 observations and its representative at the national level.

26:26 As,

26:27 as,

26:27 as you all said,

26:28 when,

26:28 when we start,

26:29 uh,

26:31 grouping or,

26:31 or looking at the top wealthy or the top even 1%,

26:35 we start to lose,

26:36 uh,

26:37 a representative or,

26:38 or we start,

26:39 we start to have a lot of measurement errors.

26:41 So that's why I'm focusing on the 10th person on,

26:43 on only,

26:44 on only on only the sales.

26:47 Uh,

26:47 with respect to other,

26:48 uh,

26:48 to try to avoid other type of,

26:50 uh,

26:50 errors,

26:51 I am using median,

26:52 median changes.

26:53 So,

26:54 uh,

26:54 in a way,

26:55 there,

26:57 the media,

26:57 uh,

26:58 we,

26:58 we know that it's less,

26:59 less biased against uh the extremes.

27:03 Uh,

27:03 so that,

27:03 that,

27:04 that is what I,

27:04 I'm trying to,

27:05 to use to avoid these,

27:07 these,

27:07 uh,

27:09 possible errors.

27:10 Besides that,

27:10 uh,

27:11 I,

27:11 uh,

27:11 there is no,

27:12 no other strategy that I am currently doing.

27:16 So we,

27:17 we've got many questions still.

27:18 So,

27:19 so David,

27:19 Steven,

27:20 and then Sergio,

27:20 please try to keep the question short

27:23 so Sergio can finish.

27:24 Go ahead,

27:24 David.

27:25 My,

27:26 my question is just whether we,

27:28 we think of these sudden stops as

27:29 all alike and all exogenous or,

27:32 you know,

27:33 whether,

27:33 in particular,

27:34 if they're accompanied by inflation,

27:36 you would want to be holding debt and,

27:38 and inflating away the debt,

27:39 and so,

27:40 you know,

27:40 whether this Mexican case is,

27:42 is

27:44 representative of how we should always think of these.

27:46 Mm.

27:47 Yeah,

27:48 so thank you,

27:48 David,

27:49 for the question.

27:49 Um,

27:50 I think the,

27:51 the literature has,

27:52 has done a very good job on documenting how these sudden stops happen

27:57 and,

27:57 and they happen in,

27:59 in episodes

28:00 characterized by

28:01 uh the,

28:02 the debt deflation process.

28:04 So,

28:05 so.

28:06 Uh,

28:06 there are

28:07 two possible strands here to analyze the sort of stuff.

28:10 We could think about a real exchange rates

28:13 in the sense that there is a non-tradable good,

28:15 so that,

28:16 that,

28:16 that suddenly depreciates,

28:18 uh,

28:19 this like a real exchange depreciation

28:21 and,

28:21 uh,

28:22 domestic,

28:23 domestic income,

28:24 uh,

28:25 gets tightened

28:26 and this serves as collateral or this serves as a

28:30 debt to income type of,

28:31 of ratio.

28:32 So

28:33 the,

28:33 the households end up more constrained.

28:35 In,

28:36 in my setup here,

28:37 I am thinking about the depreciation,

28:39 the,

28:39 the deflation

28:41 on the uh

28:42 productive asset,

28:43 the deflation on the collateral

28:46 uh asset that,

28:47 that on the asset that serves as collateral.

28:48 And I think it's very,

28:50 uh,

28:50 the,

28:50 the,

28:51 the,

28:51 the literature has done a good job uh documenting in,

28:54 in,

28:55 in uh

28:56 stylized facts about

28:57 these episodes that are characterized by declines in,

29:00 in prices.

29:01 So,

29:02 Uh,

29:02 as you will say,

29:03 if,

29:03 if we were in,

29:04 in a case where

29:05 there was high inflation,

29:07 then it would be good to be indebted so you can,

29:09 uh,

29:10 you can in real terms,

29:11 uh,

29:11 end up.

29:13 With lower real,

29:14 real debt,

29:15 but that's,

29:15 that's not the case in these other stocks.

29:22 OK,

29:23 great.

29:23 So,

29:23 um,

29:25 So

29:26 now let me move on and go

29:29 straight uh so Sergio,

29:29 we still have a few questions.

29:32 Uh,

29:33 let's go with uh with Steven and Sergio and Carmen.

29:36 Maybe we'll let them Sergio move a little bit forward and then take your question.

29:42 Hi,

29:42 um,

29:42 so I have a,

29:43 a question about this sort of application to the housing market.

29:47 We're talking about leverage,

29:49 um,

29:49 a lot of this is household leverage,

29:52 but when you have a fall in,

29:53 um,

29:54 in house prices,

29:56 you usually don't have a margin call.

29:58 Um,

30:00 so I was wondering how,

30:01 um,

30:01 sort of this mechanism

30:03 would work a bit more mechanically.

30:06 Um,

30:06 and also you mentioned the increase in,

30:08 in interest rates,

30:09 but,

30:09 um,

30:10 I guess the Bank of Mexico would have dropped interest rates,

30:12 um,

30:13 substantially at that time as well.

30:17 Uh,

30:17 great.

30:18 So,

30:18 thank you,

30:19 Steven,

30:19 uh,

30:20 uh,

30:20 for the questions.

30:21 Regarding the second one.

30:23 in fact,

30:24 yeah,

30:24 the,

30:25 the,

30:25 uh,

30:26 the policy

30:29 interest rate target drops or there,

30:31 there is an easing in the monetary policy,

30:33 so that's,

30:34 that,

30:34 that,

30:34 that happens.

30:35 But when you look at the,

30:37 uh,

30:38 effective interest rates on,

30:39 on,

30:39 on house,

30:40 on households,

30:41 there was a,

30:42 uh,

30:42 an increase and,

30:43 and this,

30:44 uh,

30:45 this has been

30:46 Uh,

30:47 documented and,

30:47 and that is uh,

30:49 a common,

30:49 a common trait on,

30:50 on these

30:52 aggregate shocks like an interest rate

30:54 hike

30:55 that affects small open economies.

30:58 Uh,

30:59 so that,

30:59 that's about the real,

31:00 the real rate,

31:01 not the policy,

31:02 the policy instrumented.

31:04 Uh,

31:04 for the,

31:05 the,

31:05 the margin calls on,

31:07 on housing,

31:08 I,

31:08 I,

31:08 I,

31:09 I see your point and,

31:10 and,

31:11 uh,

31:12 it's,

31:12 it's gonna be

31:14 key for the uh fire selling of these,

31:17 of these agents.

31:18 Uh,

31:20 I,

31:20 I,

31:21 I do not have um

31:24 Like

31:25 I,

31:25 I would say that there,

31:27 there are indeed this uh margin calls

31:29 in,

31:30 in an economy like Mexico where the

31:32 uh

31:33 large part of the employment is,

31:35 is informal,

31:36 around 50 to 60% of employment is informal.

31:39 So we could think that the self-employed workers.

31:42 Use part of their collateral,

31:44 use their real estate as collateral

31:46 to,

31:47 uh,

31:48 to,

31:48 to,

31:48 to,

31:49 to get uh

31:50 funds for their productive,

31:52 uh,

31:52 fund productive process.

31:54 And there,

31:55 uh,

31:55 I would say that they,

31:56 we,

31:57 it's not,

31:58 it's not uh

32:00 unthinkable to think that there,

32:01 there's,

32:02 there are these margin calls in,

32:03 in that type of,

32:04 of.

32:05 Of contracts of debt and collateral.

32:08 But,

32:09 but uh yeah,

32:09 so that,

32:10 that,

32:10 that would be my answer.

32:14 Uh,

32:14 Sergio,

32:14 one,

32:15 quick question.

32:16 One prominent feature of,

32:18 um,

32:19 science stops is,

32:20 uh,

32:20 sharp movements in the exchange rate,

32:22 and that can produce

32:24 also the significant transfers,

32:27 uh,

32:27 between households that linked to the tradeable and non-tradable sectors.

32:31 Or households that uh have assets in foreign currency

32:34 versus the ones that don't have a foreign currency.

32:37 I was wondering whether you consider uh that mechanism as well,

32:40 but I guess not,

32:41 but that's also important.

32:44 Uh,

32:45 thank you,

32:45 Sergio.

32:45 Yes,

32:46 that is an excellent point.

32:48 Uh,

32:48 here in,

32:48 in the data,

32:49 I am not

32:50 considering anything about the real exchange rate or currency risk.

32:54 We,

32:55 when I,

32:55 when I go over the model,

32:56 you will see that I am not using a tradables and

32:59 non-tradables type of model in which uh the relative price of,

33:03 of the,

33:04 of the

33:04 non-tradable serves as the real exchange rate that depreciates,

33:08 which is the,

33:08 the

33:09 The movement that you're talking about.

33:11 But uh in a,

33:12 in a,

33:12 in a

33:14 related way

33:15 uh in my model,

33:16 I will have a

33:18 domestic,

33:19 uh,

33:19 a domestic asset which is not traded within foreign to foreigners,

33:24 which has a domestic price,

33:26 uh equilibrium price.

33:27 And when this price depreciate when this price deflates or when this price drops.

33:33 It is in a way analogous to uh uh

33:36 the deflationary process on the exchange rate,

33:40 but explicitly,

33:41 it's not taking into account the,

33:42 the currency,

33:43 the currency movements.

33:53 There's one last question by Carmen,

33:55 Carmen,

33:55 do you want to try to keep 16 so then Sergio can,

33:58 can move,

33:59 uh,

33:59 and then we'll let you move forward.

34:00 If there's more questions,

34:01 we'll hold them,

34:01 OK?

34:03 Very quickly,

34:05 to follow up on uh David McKenzie's uh question,

34:09 not all sudden stops are created equal.

34:11 You are focused here on the

34:14 global financial crisis sudden stop,

34:16 which was really an exogenous shock,

34:19 a problem in the US and in other advanced economies in Mexico.

34:22 This is relevant for your offset and it goes back to my previous question.

34:27 Have you

34:28 looked or considered looking,

34:29 I don't even know if it's possible,

34:32 uh,

34:32 at calibrating,

34:34 uh,

34:34 the exercise to the 1995

34:37 Mexican crisis because

34:40 uh it's,

34:40 it's pertinent to the earlier point of the offset,

34:44 OK?

34:45 If

34:45 the

34:46 shock is outside,

34:48 yes,

34:48 you,

34:49 you,

34:49 the,

34:50 you know,

34:50 from the vantage point of the high-income

34:53 households,

34:54 it

34:55 This is a temporary decline.

34:56 This is a great buying opportunity.

34:59 Uh,

35:00 if the shock originates domestically and there's greater uncertainty

35:05 whether real estate prices will recover because

35:08 there are all kinds of domestic problems.

35:10 This is the 1994,

35:11 1995 crisis.

35:14 Um,

35:15 I am less confident,

35:16 as I mentioned earlier on the offset part.

35:18 I

35:19 actually

35:20 buy

35:21 very much

35:22 the,

35:22 the effects you highlight

35:24 on the,

35:25 uh,

35:27 on the low income households and their

35:29 constraints on,

35:30 on,

35:31 on consumption,

35:32 but,

35:33 uh,

35:33 on the offsetting side,

35:34 I think there are differences,

35:37 uh,

35:37 across those two,

35:39 those two sudden stop.

35:41 Uh,

35:42 episodes that are more generic,

35:43 that is,

35:44 you know,

35:45 sudden stops originating with domestic problems

35:48 versus sudden stops that are originating.

35:51 From a problem somewhere else.

35:54 I see.

35:54 Great.

35:55 Uh thanks,

35:55 Carmen,

35:56 for,

35:56 for,

35:56 for the uh clarification.

35:58 Yeah,

35:59 so here.

36:00 As,

36:00 as,

36:01 as you will say,

36:01 if the,

36:02 if the shock

36:03 originates outside,

36:04 we

36:05 precisely rely on the domestic agents to absorb

36:08 this dampening effect.

36:10 If,

36:10 if the combination of,

36:11 of,

36:12 of shocks is domestic,

36:14 then,

36:15 and maybe domestic and maybe with a more permanent structure,

36:18 then

36:19 uh maybe as you will say,

36:20 it's clear that

36:22 domestic agents are not going to be the ones

36:24 that are going to be buying these assets because uh

36:27 Their,

36:27 their,

36:28 their new price is gonna be low for long.

36:30 So they will try to,

36:31 to,

36:32 to get uh other type of assets or diversify with,

36:35 with foreign type of assets.

36:37 So in,

36:38 in that sense,

36:38 yes,

36:39 here I am focusing on,

36:41 on the,

36:42 the,

36:42 the,

36:42 the,

36:42 the,

36:43 the,

36:44 the events

36:45 or the,

36:45 the episodes

36:47 in which

36:48 domestically there,

36:49 there was no change.

36:50 The,

36:50 the,

36:50 all of the change comes from the,

36:52 the interest rate shock or the foreigners.

36:54 And uh uh those are the,

36:56 the ones that I,

36:56 that I will focus on.

36:58 Or in a way,

36:59 what I,

36:59 what I wanted,

37:00 what I meant to say when,

37:02 uh

37:03 when I was talking about

37:05 all of a sudden stops being equal is that

37:08 I am focusing on

37:10 Oh,

37:10 the domestic,

37:11 the domestic risky part of,

37:13 of the soil and stuff is that

37:14 the domestic economy has to be sufficiently indebted.

37:17 So,

37:17 so we have to end up in,

37:19 in,

37:19 in a state of the world in which

37:21 uh

37:21 the,

37:22 the,

37:22 the,

37:22 the domestic economy is sufficiently exposed

37:26 to foreigners

37:27 and then as uh an,

37:28 an interest,

37:29 an interest rate

37:30 increase that would be otherwise uh

37:32 A real business cycle type of movement triggers these,

37:35 these sudden stops.

37:36 But uh

37:37 I'm,

37:38 I'm gonna focus only on one aggregate shock,

37:40 which is the interest rate.

37:41 So in that sense,

37:42 This,

37:43 it wouldn't be a,

37:44 a model to,

37:46 to model the '94,

37:47 95 crisis.

37:48 It will be more a model to,

37:49 to,

37:50 to,

37:50 to calibrate for the 2009 episodes.

37:55 Perfect.

37:56 So,

37:57 so now,

37:57 uh,

37:58 I'll,

37:58 I'll move on and describe to you very quickly the,

38:01 the model framework.

38:03 So I'm proposing a small open economy Bly model.

38:05 The Bly component is that

38:07 the households are gonna be heterogeneous.

38:10 There's going to be a representative production firm

38:13 that with given all of the other assumptions,

38:15 it will behave basically as an endowment economy

38:18 in which all of the

38:19 pie

38:19 of the production

38:20 is divided between wages and dividends.

38:23 And then we have the rest of the world

38:25 with which uh households trade their international bonds.

38:30 With respect to the the financial markets,

38:32 there are going to be two types of assets,

38:34 the international risk-free bond,

38:36 which the households buy and sell with the foreigners,

38:39 and a domestic illiquid risky asset.

38:42 Which is illiquid because there is an adjustment cost associated to it.

38:46 It is risky because

38:48 uh the domestic asset

38:49 price uh is,

38:50 is,

38:50 is stochastic

38:52 and,

38:53 and will only be traded within the corner.

38:55 This is a,

38:56 uh,

38:56 an assumption that could easily be uh

38:59 uh

39:00 relaxed if we introduce an an exogenous demand for this domestic assets

39:05 which um

39:06 Uh,

39:07 could,

39:08 could,

39:08 could be like a,

39:09 uh,

39:09 foreigners buying part of,

39:11 of,

39:11 of the domestic asset or,

39:12 or leaving the economy in,

39:14 in turbulent times.

39:16 Then the financial friction that I am using is a loan to value credit constraint,

39:20 which means that the international debt that the households have cannot

39:24 be more negative than a fraction of the market value assets.

39:29 And lastly,

39:30 the aggregate shock is the shocks to the interest rate.

39:33 So,

39:34 very quickly,

39:35 let the,

39:35 let me describe to you the domestic agents.

39:37 The households have a standard utility function over consumption.

39:41 This means that they do not care about

39:44 their labor,

39:45 so they will supply one unit of labor inelastically

39:48 and they will be heterogeneous in their

39:50 bond,

39:51 international bond positions B,

39:53 their domestic asset positions A.

39:57 The shocks,

39:58 the idiosyncratic shocks that they get

40:00 on their wages.

40:01 This is the standard Ayagari type of idiosyncratic shock

40:05 and a dividend productivity shock

40:07 that affects the returns on their assets.

40:10 And this shock is uh analogous to a wealth,

40:13 uh,

40:13 uh heterogeneous wealth returns or heterogeneous capital returns

40:17 that recent literature has shown

40:19 that are the main drivers of inequality in an economy.

40:24 With respect to production,

40:25 a standard,

40:26 uh,

40:27 constant return to scale production function that takes,

40:30 that demands aggregate labor which is going to be fixed

40:33 and and rents aggregate assets,

40:35 which is also going to be in

40:36 fixed supply.

40:38 So in that sense,

40:39 the,

40:39 the profit maximization problem becomes static

40:43 and profits are going to be zero in equilibrium.

40:44 So all of the,

40:45 all of the production is divided between wages and dividends.

40:49 Finally,

40:50 to close the economy,

40:51 the market clearing condition means that

40:54 the demand

40:55 of all of the assets traded by the households

40:58 has to be equal to the fixed supply of these of these assets.

41:02 And the demand

41:04 depends on the whole distribution of households across these individual states.

41:09 So,

41:09 now we're ready to

41:11 Uh,

41:12 formulate the recursive problem of the household.

41:14 So in a way,

41:15 it's a very simple model in the sense that

41:17 households have some idiosyncratic states.

41:20 There is a,

41:21 an aggregate shock to interest rate,

41:23 but the last part is that uh

41:26 the,

41:26 the,

41:27 the whole distribution of households is gonna be also a state.

41:30 So this adds a lot of complexity.

41:32 But

41:33 we,

41:34 if we think about the,

41:35 the fundamentals of the model,

41:36 it's,

41:36 it's kind of a simple,

41:38 simple representation.

41:39 The house funds will maximize their utility function and future value functions

41:45 subject to

41:46 a standard budget constraint.

41:49 In which on the expenditure side,

41:50 they buy consumption goods,

41:52 they buy

41:54 bond holdings for tomorrow at an exogenous price.

41:57 They buy assets for tomorrow at an endogenous

42:00 Q price that depends on the whole distribution.

42:04 And with respect to their income side,

42:06 they get wages,

42:07 they get uh

42:08 the dividends of the shares that they own

42:12 and they

42:12 get the income from the current bond positions they bring into the period.

42:18 Now,

42:18 secondly,

42:19 the maximization is subject to this loan to value constraint

42:22 and this is gonna be key.

42:23 Here,

42:24 uh,

42:25 I,

42:25 I'm highlighting in,

42:26 in red

42:27 precisely where the debt deflation happens.

42:31 The amount of assets or the amount of debts that

42:33 the households own cannot be more negative than this fraction kappa

42:37 of the market value of their assets.

42:40 So,

42:40 the,

42:41 the market value of their assets has 22 components,

42:44 quantities A and prices Q.

42:47 So this,

42:48 this Q here introduces this pecuniary externality in the sense that

42:52 when a household

42:54 sells or fire sales their assets A,

42:57 they do not internalize how this behavior.

43:00 will affect

43:01 the,

43:01 the market clearing price.

43:03 And if they are selling,

43:05 this price will drop.

43:07 So they do not internalize how this

43:08 behavior tightens their own collateral constraint,

43:11 but also tightens the,

43:12 the collateral constraint of everyone in the economy.

43:15 So that's the pecuniary externality.

43:19 Lastly,

43:19 uh,

43:19 the adjustment cost function or the liquidity component of

43:22 the asset is quadratic as standard in the literature.

43:26 So,

43:26 let me

43:27 briefly describe to you

43:29 the marketing completeness because this,

43:30 this is gonna introduce a trade-off that is very interesting.

43:33 With respect to the uh the idiosyncratic shocks on wages,

43:37 given that the,

43:38 that the household supplies inelastically their labor,

43:42 the exposure to the earnings risk is fixed.

43:44 It doesn't matter what the household does if,

43:46 if,

43:47 if,

43:48 it sells or,

43:49 or buys assets,

43:51 the exposure to the earnings risk is fixed.

43:54 But when we think about the dividend risk,

43:56 which is persistent,

43:58 this uh risk component combined with the loan to value debt constraint generates,

44:03 generates an asset wealth trade-off.

44:05 And what do I mean by this?

44:06 Imagine that a household gets one additional unit of the asset.

44:10 This in a way some,

44:11 gives something good to the household.

44:13 It gets larger debt capacity.

44:16 It loosens the,

44:17 the collateral constraint.

44:19 And that allows for better consumption smoothing.

44:21 So,

44:22 consumption volatility is reduced,

44:23 something desired.

44:25 However,

44:25 on the other end,

44:26 this household that gets this additional unit of the asset

44:30 gets additional future exposure to the dividend risk.

44:34 To give you the intuition,

44:35 let's think of a household that has zero assets.

44:37 It doesn't matter if the dividend,

44:39 if the dividend is high or low.

44:41 The disposable income of the household wouldn't change.

44:45 However,

44:46 if the household relies a lot

44:48 on,

44:48 on the asset,

44:49 on the risky asset

44:50 uh component,

44:52 then it,

44:53 depending on the,

44:53 on the dividend risk,

44:55 there's going to be extra uh volatility in the disposable income.

44:59 So in the end,

44:59 what we're gonna get is that the 2nd bullet point dominates

45:04 and asset-rich households

45:07 will end up

45:08 getting less debt

45:09 or even getting positive

45:12 international bonds.

45:13 So in a way,

45:13 endogenously here the model

45:15 generates that asset wealth,

45:17 uh,

45:17 asset-rich households end up with a more diversified portfolio.

45:21 They end up holding

45:22 both

45:22 international bonds

45:24 and domestic assets.

45:26 And this behavior generates the unconstrained wealthy agents

45:29 which are gonna be key for the dampening

45:31 uh uh

45:33 effect.

45:35 And I'm gonna skip this equity premium decomposition that I,

45:38 that I do in the paper,

45:39 but basically it shows how

45:41 both

45:42 the dampening effect

45:44 coming from the unconstrained wealthy agents and the

45:46 amplifying effect coming from the fire sales,

45:49 uh,

45:50 interact on the equity premium,

45:52 uh,

45:52 the uh

45:54 uh the equity premium uh

45:57 result on the equity premium of the,

45:59 of the economy.

46:02 So now,

46:03 uh we're ready to use the,

46:04 the model uh to quantitatively analyze solid stocks.

46:07 So the roadmap is going to be the following.

46:09 I'm gonna show you first how the calibration of the stationary model,

46:13 uh,

46:14 the,

46:14 the calibration of the stationary model

46:16 in which I keep the interest rate at the long run mean.

46:19 I'm gonna match some moment,

46:21 some data moments,

46:22 and I'll show you policy functions and we're gonna be able to see clearly how

46:26 the fire sales occur

46:27 and how this asset wealth trade-off works.

46:31 Then uh we're going to specify the stochastic process on the interest rate

46:36 and we're going to solve,

46:37 I'm going to solve and simulate the,

46:38 the full aggregate risk model.

46:41 After a long simulation of,

46:42 of,

46:42 of periods,

46:44 I'm going to identify sudden stop events as

46:47 the periods in which the current account reversal

46:50 is two standard deviations above the historical mean.

46:53 So these are large outflows of capital.

46:56 And we're gonna,

46:57 and uh look at the,

46:59 uh,

46:59 and the,

47:00 in the end,

47:00 we're gonna look at how

47:02 the severities of the crisis

47:04 change depending on the decrease of inequality.

47:08 So very briefly,

47:08 the calibration in a way is a parsimonious model.

47:11 I uh take calibrate,

47:13 I calibrate exogenously the set of parameters that

47:16 are either common in the literature or,

47:18 or,

47:19 or do not require simulation of the model.

47:21 So risk aversion equal to 2 standard

47:24 interest rate,

47:25 long run interest rate equal to 3%.

47:28 The capital factor share of 13.

47:31 And uh lastly,

47:32 the,

47:32 the CA debt fraction collateral I calibrated to equal,

47:36 to be equal to the average leverage in,

47:39 in,

47:39 in Mexico in 2005.

47:42 Then for the,

47:43 the set of

47:43 parameters

47:44 uh calibrated by simulation,

47:46 uh discount factor of 0.90,

47:49 which is on the lower end of standard uh literature

47:52 values,

47:53 but it's,

47:53 uh,

47:54 but it allows me to match the net foreign asset position equal to 40% of,

47:59 of GDP.

48:01 And the last part which is gonna be key

48:03 is the trading cost parameter.

48:05 Here,

48:05 uh,

48:05 I do not have that data available data on transaction costs,

48:09 but I'm,

48:09 I'm gonna

48:11 fix.

48:12 To 5%,

48:13 average 5% of transaction costs

48:16 that is in,

48:17 in the bounds of uh

48:19 data or in the bounds of estimates obtained by Ayacari and Gertlet for the US.

48:25 Then

48:26 we have Sergio.

48:27 We've got 1 question,

48:28 2 questions actually.

48:29 Is that a good time?

48:30 OK,

48:30 let's go.

48:30 So Tuan and then Alvaro.

48:33 Uh,

48:34 hi,

48:34 Shay.

48:34 So,

48:35 um,

48:37 So in this model,

48:38 you know,

48:38 all the households are the same.

48:41 And uh it's very practical because uh predictions on,

48:44 on,

48:44 on the investment and,

48:46 and consumption

48:48 uh choices are easy,

48:49 but

48:50 so you,

48:50 you have to assume that

48:52 the inequality in assets is kind of an exogenous

48:55 shock.

48:56 But if we think that the initial inequality is driven by

48:59 heterogeneity in,

49:01 let's say risk preferences.

49:04 Then,

49:04 then how would you want to think about

49:07 Uh,

49:07 uh,

49:08 this,

49:08 this identification problem.

49:10 Uh,

49:11 in the context of your model,

49:13 so

49:13 thanks Juan for,

49:14 for the question.

49:15 Uh,

49:16 so in,

49:16 in fact,

49:16 in,

49:17 in this model every household is,

49:18 is the same.

49:20 But they are exposed to the,

49:21 to these idiosyncratic shocks in terms of their

49:23 wages and in terms of their effective dividends.

49:26 So they end up

49:27 uh

49:29 optimally choosing different portfolio choices.

49:31 So they end up choosing,

49:33 some of them end up choosing more assets,

49:35 some of them end up choosing

49:36 more debt.

49:38 And,

49:38 and we end up with a,

49:39 with a distribution of leverage.

49:41 So in that way,

49:42 uh,

49:43 Uh,

49:43 they,

49:44 they,

49:44 they,

49:44 it's not that they are ex ante

49:46 identical in terms of their assets

49:49 or,

49:49 or,

49:49 or the distribution of assets is gonna be fixed.

49:52 This is something that,

49:53 uh,

49:54 the,

49:54 the,

49:54 the households are gonna be moving along these distributions.

49:58 So that's,

49:59 that's,

49:59 that's,

50:00 that,

50:00 that,

50:00 that's the point.

50:03 Uh,

50:03 Sergio,

50:04 this is Alvaro.

50:04 So quick question and feel free to defer if I'm getting ahead of myself.

50:07 So you said earlier that you were gonna do some estimation also for a

50:11 high-income country.

50:12 I see that you can do this,

50:14 uh,

50:15 away from the serial lower bound,

50:16 but you're more like if you,

50:18 when you do it for high income,

50:19 like how do you get away from the multiplicity of equilibrium

50:21 if you were to be like close to the CLD?

50:25 I see.

50:25 OK.

50:25 So,

50:26 uh,

50:26 great.

50:27 Thanks,

50:27 Alvaro,

50:28 uh,

50:28 here.

50:29 Um,

50:31 so,

50:32 I,

50:34 uh,

50:34 at first sight,

50:34 I would tell you this is applicable to,

50:37 to an,

50:37 to an advanced economy

50:38 that is far from the zero lower bound.

50:40 Here I am not taking into account any uh

50:44 uh effective lower bound considerations in terms of the,

50:46 of the interest rate.

50:48 If you think that the interest rate is exogenous even for an advanced economy and uh

50:53 uh

50:54 you,

50:54 you can model it exogenously how the zero lower bound affects this,

50:58 uh,

50:59 the,

50:59 the,

50:59 the sort of stocks,

51:00 but my,

51:01 my,

51:01 my answer would be like.

51:02 Uh,

51:04 In a way,

51:04 I am,

51:06 uh,

51:06 if,

51:07 if I would like to compare an advanced economy,

51:09 which is something that I do in one of my uh counterfactuals.

51:13 Uh,

51:13 I am abstracting from,

51:15 from monetary policy restrictions.

51:20 Great.

51:21 So now,

51:22 um,

51:23 So now,

51:24 let me go briefly and tell you about the other two

51:27 sources,

51:28 well,

51:28 the two sources of,

51:29 of heterogeneity in the model are first,

51:32 the earnings process.

51:33 Here,

51:34 there is a very vast literature on how to properly estimate

51:37 the wage risk on,

51:38 on,

51:39 on households.

51:40 So I,

51:40 I,

51:41 I take,

51:42 uh,

51:43 some,

51:43 I,

51:43 I,

51:44 I follow one strand of this literature in,

51:46 in the sense that I

51:47 Um,

51:48 estimate a stationary process with a persistent

51:51 and a and a transitory component.

51:54 This,

51:54 uh,

51:54 in fact,

51:55 I,

51:55 I would say that it is one of the very few papers that

51:58 has done this for an emerging economy and in particular for Mexico.

52:02 Uh,

52:02 so,

52:02 so in,

52:02 in a way,

52:03 it's also a contribution to compare

52:05 uh the estimates of this earnings risk

52:08 to the estimates obtained for the US.

52:10 So what I do is using a quarterly

52:12 Uh,

52:13 household

52:14 employment and labor,

52:15 uh,

52:16 employment and wages

52:17 survey,

52:17 the ENOE for Mexico

52:21 with,

52:21 with data from 2005 to 2014.

52:24 I estimate this earnings risk

52:27 and I find that the auto correlation coefficient

52:29 is smaller than the estimates for the US

52:32 and the variance is a little bit larger.

52:34 It's,

52:34 it's on the upper end.

52:36 Uh,

52:36 this is,

52:37 uh,

52:37 possibly,

52:38 uh,

52:39 Uh,

52:40 this,

52:40 this results from a combination of a large

52:42 informal sector market which is more flexible,

52:45 so this allows for lower persistence on,

52:47 on,

52:47 on shocks.

52:49 And,

52:49 uh,

52:50 the lack of safety nets or unemployment insurance also allows for this,

52:54 uh,

52:54 or contributes to these larger risks.

52:56 So to test this hypothesis,

52:58 then I estimate the same model

53:00 restricting only for a formal employment and in fact,

53:03 we see that

53:03 the auto correlation uh increases.

53:06 But there's still a difference between the estimates

53:08 for the US and for the American economy.

53:11 Then the second source is about the dividend risk.

53:14 Here,

53:14 there is uh,

53:15 I,

53:16 I lack the available data to properly estimated the process,

53:19 but what I do to discipline the model is

53:21 to match the leverage distribution of the households.

53:24 In particular,

53:24 I match some moments of this distribution.

53:26 So in particular,

53:27 I match the,

53:27 the percentage of savers,

53:29 which is around 5%.

53:31 The percentage of indebted households that have less than

53:35 the average debt,

53:36 the average leverage,

53:37 which is around 60%.

53:39 And then the financially constrained households which are 18%.

53:44 So now we're ready to look into the policy functions.

53:47 Uh,

53:48 I have 20

53:49 I have 20 minutes.

53:50 So let me show you,

53:51 uh,

53:51 very quickly some of these policy functions.

53:54 Remember,

53:55 the two choices of the households are

53:57 next period bond holdings or B

54:00 and next period asset holdings or A.

54:03 And these functions are a function of the current bond positions,

54:07 the current

54:08 asset positions A,

54:10 the current shock on the wages,

54:12 and the current shock on the dividends.

54:14 So,

54:15 in this,

54:15 in this plots,

54:16 I am

54:16 fixing the amount of assets that the household owns

54:19 and I am varying in the X axis the bonds,

54:22 the,

54:22 the,

54:22 the current bond.

54:24 And here,

54:25 uh,

54:26 the,

54:26 the plot on the left is a standard plot on debt.

54:29 We have uh seen these

54:30 plots before in the sense that

54:32 households get more debt,

54:34 more debt until they are up to their limits.

54:36 But when,

54:37 if we look at the right uh figure,

54:39 here we can see precisely how the fire sales work.

54:43 When the household is far from the constraint,

54:46 their adjustments in terms of uh,

54:48 of,

54:48 uh,

54:49 of,

54:49 of assets

54:50 is fairly constant.

54:51 Either they sell if they get the low dividend,

54:54 which is the red line,

54:55 or they buy more assets if they get the high dividend,

54:58 which is the blue.

54:59 But when they,

55:00 they get close to the collateral constraint,

55:03 we see a large increase in their,

55:05 uh,

55:05 in this,

55:06 in the slopes of these policy functions

55:08 which reflect the fact that they are fire selling their assets.

55:13 Then we can do the same exercise but holding now constant the value of,

55:17 of the bond position

55:19 and moving around in the X axis,

55:21 the A prime,

55:22 the,

55:22 the,

55:22 the current assets.

55:24 Here,

55:24 uh,

55:24 if we focus on the,

55:25 on the left panel,

55:27 the

55:27 bond policy function,

55:29 we show that we,

55:30 we,

55:30 we see that

55:32 if we focus on the,

55:33 on the blue line.

55:35 As the households get a little bit

55:37 richer,

55:38 they decide to choose as much debt as possible,

55:41 so they are constrained.

55:42 Then there is a point where they become a little bit more richer,

55:46 so they start saving.

55:47 They are still increasing their debts,

55:49 but they are leaving part of of their available credit.

55:52 And then,

55:53 there's another threshold by which they become sufficiently rich

55:57 in the sense that they start decreasing their debts

56:00 and even become savers.

56:02 So that is what I mean by the asset wealth

56:04 trade-off in the sense that these households become more diversified.

56:07 They end up saving in foreign bonds and saving in domestic assets.

56:13 So the solution of the stationary model very,

56:15 uh,

56:15 it's uh

56:17 very interesting in the sense that uh the

56:18 model is able to capture non-targeted inequality measures.

56:22 The wealth Gini is,

56:23 is around 60.68 and the observed in the data is 0.73.

56:28 And we also get large consumption inequality around a Gini of around 0.30.

56:34 So now,

56:35 uh,

56:36 going back to the roadmap,

56:37 so far we have looked at the stationary model and

56:40 now we're ready to look at the full uh aggregate,

56:42 uh,

56:43 aggregate risk framework.

56:45 So to,

56:46 to be able to do that,

56:47 I,

56:47 uh,

56:48 first,

56:48 let's specify the interest rate uh stochastic process.

56:51 It will follow a two-state Markov process in

56:54 which the interest rate will take values of

56:56 1%

56:58 and 5% with some persistence.

57:00 And to be able to solve the model I will

57:02 uh build on Cruzel and Smith's non-trivial market clearing algorithm

57:07 and I will adapt it to a small open economy framework.

57:10 So this requires uh,

57:12 uh,

57:12 it's computationally intensive and it requires,

57:14 uh,

57:15 it's not,

57:15 it's called non-trivial because every period you simulate,

57:19 you have to solve um a fixed point iteration

57:22 problem on the market clearing price.

57:25 So,

57:25 but

57:26 after doing a long simulation of,

57:27 of the economy,

57:29 I identify the sudden stops as the moments in which the international capital

57:34 or the current account reversals are two standard deviations above their mean

57:38 and I average across them

57:40 to create these event study plots.

57:43 Uh,

57:44 so,

57:44 I,

57:45 I,

57:45 uh,

57:46 centered the plots around zero,

57:47 which is the period of the crisis,

57:49 and I,

57:49 I'm showing you

57:50 plus and minus 3 periods.

57:53 And first,

57:54 in,

57:54 in solid lines,

57:55 these are the uh the data simulated from the model and in

57:58 dotted lines are the average of Mexico's 95 and 2009 crisis.

58:03 So two things.

58:04 First,

58:04 uh,

58:05 one thing that I highlight is

58:07 All of the sudden stops are,

58:09 uh,

58:10 caused

58:10 by interest rate increases.

58:12 That is something expected because in the model,

58:14 the only aggregate risk is the interest rate.

58:18 But not all of the interest rate hikes.

58:21 Uh,

58:21 end up being a sudden stop.

58:23 Only the ones in which they,

58:24 uh,

58:25 there is sufficiently debt or there is a sufficiently large share

58:28 of,

58:29 of constrained households.

58:31 Then uh the model can capture these

58:33 uh current account reversals that are persistent.

58:36 As we see in the data.

58:38 And in terms of asset prices and consumption,

58:40 the,

58:40 the,

58:41 the model captures that consumption drops are larger than the asset prices drops.

58:47 And although here due to the uh

58:49 uh

58:50 scale of the plot,

58:52 it seems that the model cannot capture this boom

58:55 prior,

58:55 boom in terms of asset prices prior to the crisis.

58:59 Indeed,

59:00 we get,

59:00 we get an increasing,

59:02 increasing uh

59:04 trajectory in terms of the assets,

59:06 although it's insignificant compared to,

59:08 to what we see in the data.

59:10 And in the last,

59:11 in the,

59:11 in the last

59:12 panel,

59:13 I am showing you uh one of,

59:15 related to one of the questions that we got at the beginning about uh

59:18 future consumption inequality.

59:20 And we see that the model can capture that during the crisis,

59:23 uh,

59:24 inequality in consumption decreased

59:26 and then after that it starts to increase.

59:30 So now,

59:31 let me,

59:31 let me uh tell you how the model

59:33 rationalizes the asset and consumption dynamics during crisis.

59:36 And to do that,

59:37 I am reproducing precisely the,

59:39 the,

59:39 the same uh statistics that I showed you uh in,

59:42 from the empirical evidence

59:44 in parenthesis.

59:45 And now

59:46 outside of the parenthesis,

59:47 the,

59:47 the moments coming from data simulated from the model.

59:51 So,

59:51 uh,

59:52 in a way,

59:52 not in,

59:53 in,

59:53 in,

59:53 in levels or in qua qualitatively can capture that

59:57 the

59:57 unconstrained wealthy agents are precisely the ones that are buying these assets.

1:00:03 But more importantly,

1:00:04 that the very high leveraged households and wealthy

1:00:07 are the ones that fire sale their assets.

1:00:10 Now,

1:00:10 with respect to consumption.

1:00:12 In,

1:00:12 in the,

1:00:13 in the,

1:00:13 in the right side,

1:00:14 we see that

1:00:15 the model can also rationalize

1:00:17 the fact that

1:00:18 more

1:00:19 as we increase the leverage of the households,

1:00:21 they react more

1:00:23 to this uh tightening financial conditions.

1:00:25 Again,

1:00:26 this is because the high leveraged households have the

1:00:30 large NPCs

1:00:31 and they adjust the most their consumption.

1:00:36 So now,

1:00:37 uh,

1:00:37 in the,

1:00:37 in this last part,

1:00:38 let me,

1:00:39 uh,

1:00:39 we're ready to take the model and,

1:00:41 and analyze how different levels of inequality affect the severity of crisis.

1:00:45 So to do that,

1:00:47 uh,

1:00:47 there is this

1:00:48 recent literature done uh by Alberto Bisin and,

1:00:51 and Jess Benaviv in which they show that this dividend risk

1:00:56 or heterogeneous capital returns

1:00:58 are the main drivers or are the drivers

1:01:01 of wealth inequality and in particular the right tail of the wealth inequality.

1:01:06 So,

1:01:06 with this in mind,

1:01:07 I,

1:01:08 I reduced

1:01:09 the variance of the dividend risk by 150%

1:01:12 to proxy an advanced economy.

1:01:15 Also motivated by the fact that in

1:01:17 advanced economies there is lower confiscation and

1:01:19 lower informal uh high returns coming from corruption and better rule of law.

1:01:26 And what I find is that the model predicts that

1:01:28 low,

1:01:29 this,

1:01:29 this reduction in idiosyncratic risk

1:01:33 reduces

1:01:34 uh uh the precautionary motive,

1:01:36 uh,

1:01:36 motives of the households.

1:01:38 So this allows the economy to have

1:01:41 larger debt positions.

1:01:42 So net foreign debt is twice as large.

1:01:45 This is something that we see in the data.

1:01:47 Advanced economies have on average

1:01:49 larger uh

1:01:51 net foreign asset positions with respect to their GDPs than emerging economies.

1:01:55 But surprisingly,

1:01:56 this,

1:01:57 we,

1:01:57 we would expect them to be more exposed,

1:01:59 but

1:02:00 the model predicts that sudden stops are less severe.

1:02:03 Consumption and asset price,

1:02:05 aggregate consumption and the asset price drops

1:02:08 30% less

1:02:10 with,

1:02:10 with half of the dividend,

1:02:12 with,

1:02:12 with,

1:02:12 in an economy with half dividend risk.

1:02:16 Now,

1:02:16 as a sanity check,

1:02:17 oh well,

1:02:18 first,

1:02:18 let me show you the,

1:02:19 the,

1:02:19 the event story of these two economies.

1:02:22 Now in dashed lines,

1:02:23 I am showing you

1:02:24 data simulated for the advanced economy.

1:02:27 First,

1:02:27 we,

1:02:27 we get the same persistence uh properties

1:02:31 and the,

1:02:31 the only

1:02:32 changes are in terms of the magnitudes of the drop.

1:02:35 Consumption and asset prices dropped 30 30% less.

1:02:40 And now as a sanity check,

1:02:41 uh,

1:02:41 I,

1:02:42 I,

1:02:42 I

1:02:43 took

1:02:43 inequality measures

1:02:45 for in particular the Gini coefficient from data from the World Bank precisely,

1:02:49 and I matched these uh coefficients to sort and top episodes

1:02:53 identified in a recent paper by Javier Bianchi and Enrique Mendoza.

1:02:58 And I match them to,

1:03:00 to the episodes and to the drops in consumption and,

1:03:02 and GDP.

1:03:03 And very simple correlation analysis.

1:03:06 In fact,

1:03:06 are these two scattered plots.

1:03:08 I,

1:03:09 uh,

1:03:09 I show you,

1:03:10 I am showing you in triangles,

1:03:12 data for advanced economies and in circles,

1:03:14 data for emerging economies.

1:03:16 We see two things.

1:03:17 First,

1:03:18 Emerging economies are in fact more unequal on average than advanced.

1:03:22 These are the,

1:03:23 the,

1:03:23 the circles,

1:03:24 the red circles compared to the blue triangles.

1:03:27 And also the second thing is that we see this negative relation in the sense that

1:03:31 a higher

1:03:32 income genus correspond to uh or associated with consumption,

1:03:37 larger consumption drops.

1:03:39 So this,

1:03:40 uh,

1:03:40 the,

1:03:40 the,

1:03:40 in a way the predictions of the model,

1:03:42 this,

1:03:43 this is just a sanity check for

1:03:45 it,

1:03:45 in,

1:03:45 in

1:03:46 the predictions of the model are consistent with this uh correlation

1:03:50 that we see in the data.

1:03:54 And,

1:03:54 uh,

1:03:55 lastly,

1:03:55 in,

1:03:55 in the last 10 minutes of,

1:03:57 of,

1:03:57 of the talk,

1:03:58 let me tell you about this,

1:03:59 uh,

1:03:59 redistributive dividend tax.

1:04:01 So Sergio,

1:04:02 we've got a question from Carmen,

1:04:03 I think no,

1:04:04 and then we'll let you finish.

1:04:06 Go ahead,

1:04:06 Carmen.

1:04:08 Uh,

1:04:09 very quickly,

1:04:10 I,

1:04:10 I think

1:04:11 the definition of sudden stops

1:04:14 based on the current account do hide

1:04:16 important differences between advanced and emerging.

1:04:19 Uh,

1:04:19 one of the reasons for the less,

1:04:21 lesser fluctuations in the current account balances

1:04:25 in advanced is the offset

1:04:27 by the public sector

1:04:29 of what the private side does,

1:04:32 which,

1:04:33 so,

1:04:33 so.

1:04:35 I,

1:04:35 I think

1:04:36 to get at some of the issues you want,

1:04:38 you actually want to look at that.

1:04:40 You want to look at,

1:04:42 if you will,

1:04:43 S minus I,

1:04:44 but for,

1:04:45 for the private sector because I think when you're doing the,

1:04:48 if you're doing just,

1:04:50 you know,

1:04:50 EMs,

1:04:51 it's,

1:04:51 it's much more comparable,

1:04:52 but when you bring in the advanced economy dimension.

1:04:56 Uh,

1:04:56 during

1:04:57 sudden stop episodes,

1:04:59 typically

1:05:00 public sector saves much less,

1:05:03 which,

1:05:04 you know,

1:05:05 offsets.

1:05:06 Seemingly,

1:05:08 uh,

1:05:08 the,

1:05:09 the,

1:05:09 the big reversal that you get in the private,

1:05:12 uh,

1:05:13 um,

1:05:14 Guillermo Calvo and I have some ongoing work on that.

1:05:18 I see.

1:05:19 Uh,

1:05:19 great point,

1:05:19 Carmen.

1:05:20 Thank you.

1:05:20 So,

1:05:21 yeah,

1:05:21 so further,

1:05:22 uh,

1:05:23 more rigorous analysis on the comparison between advanced and EEM economies.

1:05:27 Uh,

1:05:28 I would look only at the,

1:05:29 at the private,

1:05:30 if,

1:05:30 if you,

1:05:31 if,

1:05:31 if we can call it

1:05:32 a,

1:05:32 a

1:05:33 private current account or,

1:05:34 or private balance of payment uh flows.

1:05:37 So that uh I'll,

1:05:38 I'll

1:05:39 check that.

1:05:40 Uh,

1:05:40 to,

1:05:40 to be fair here,

1:05:41 uh,

1:05:41 what I'm,

1:05:42 what I'm doing is I am

1:05:43 taking

1:05:44 the,

1:05:44 the sudden stocks already identified

1:05:47 by,

1:05:47 by,

1:05:47 by,

1:05:48 by the literature.

1:05:49 So,

1:05:49 but,

1:05:50 but it,

1:05:50 for,

1:05:51 for further,

1:05:52 um,

1:05:53 In

1:05:54 research on,

1:05:54 on this comparison,

1:05:55 I'll,

1:05:56 I'll take,

1:05:56 I'll,

1:05:56 I'll take that into account.

1:05:59 We've got a question from Burke.

1:06:01 Burke.

1:06:02 Yeah,

1:06:03 hi,

1:06:03 thanks.

1:06:04 Uh,

1:06:04 Sergio,

1:06:04 um,

1:06:05 you talk about the Gini,

1:06:07 and that's kind of sensitive to most of the kind of

1:06:11 movements in the middle of the distribution.

1:06:13 So am I right in thinking that if you were thinking about,

1:06:15 let's say,

1:06:16 a general entropy class where it's sensitive to The bottom,

1:06:19 you may actually be overestimating the effects of inequality,

1:06:22 and then if you are thinking,

1:06:23 let's say top 5%

1:06:25 or something like that,

1:06:26 you may actually be underestimating the kind of like

1:06:29 the importance of inequality in these sudden stuffs.

1:06:32 Am I right in thinking that or am I missing something?

1:06:35 Uh,

1:06:36 so,

1:06:36 I,

1:06:37 I think you're totally right in the sense that

1:06:40 I am going to quantify.

1:06:42 Two effects

1:06:44 That depend on,

1:06:46 in a way on the tails of the distributions.

1:06:49 So if I look only on,

1:06:51 on,

1:06:51 on,

1:06:51 on a statistic that measures inequality that focuses on,

1:06:54 on the bottom.

1:06:56 It,

1:06:56 it,

1:06:57 I,

1:06:57 I,

1:06:57 I,

1:06:57 I will,

1:06:58 uh,

1:06:59 yeah,

1:06:59 and if the amplification effect is stronger,

1:07:02 then I will,

1:07:02 I will be overestimating it.

1:07:05 Oh,

1:07:05 and on the other end,

1:07:06 if,

1:07:06 if I take the,

1:07:07 the

1:07:08 The top 10,

1:07:09 uh,

1:07:10 share or,

1:07:11 or,

1:07:11 or these estimates,

1:07:12 I,

1:07:12 I will be underestimating you.

1:07:14 But uh,

1:07:14 yeah,

1:07:14 that's,

1:07:15 that's true.

1:07:17 In,

1:07:17 in a way,

1:07:17 I am using the Gini,

1:07:19 uh,

1:07:20 probably in,

1:07:20 in future iterations I could look at the

1:07:23 shares by deciles or,

1:07:25 or,

1:07:25 or a more decomposed or uh

1:07:28 statistics.

1:07:31 Great.

1:07:32 So,

1:07:33 uh,

1:07:34 so very quickly,

1:07:35 I,

1:07:35 in,

1:07:35 in,

1:07:35 in the next

1:07:37 3 minutes,

1:07:37 let me tell you about this redistributive dividend tax that I,

1:07:40 that I analysis that I do.

1:07:42 Uh,

1:07:42 government cannot wake up one day and

1:07:45 choose

1:07:46 to reduce inequality or choose to reduce the dividend risk by a decreed,

1:07:50 by a decreed or by a law.

1:07:52 So what I,

1:07:53 what I do is to introduce a

1:07:56 Constant redistributive dividend tax

1:07:58 are constant across households and constant across time

1:08:02 and to return the revenues in a lump sum transfer.

1:08:05 So this is redistributive in that sense.

1:08:08 And I find one,

1:08:09 something very interesting that is

1:08:11 with,

1:08:11 with this reduction in the dividend risk,

1:08:13 again,

1:08:14 we get lower idiosyncratic dividend risk

1:08:16 which

1:08:17 uh

1:08:18 gives us a less potent precautionary savings motive.

1:08:21 So this has two effects.

1:08:22 First,

1:08:23 wealthy households would want to demand less bonds or if they have

1:08:28 debt,

1:08:29 they would want to get more debt.

1:08:31 So here uh in blue,

1:08:33 I am showing you the,

1:08:34 the policy functions of,

1:08:35 of bonds.

1:08:36 Uh,

1:08:37 for the,

1:08:37 with the model without a tax and in red with tax.

1:08:40 So we see that in fact,

1:08:41 they want to get more debt,

1:08:44 but

1:08:44 in,

1:08:44 in the same sense,

1:08:45 they want to get less of the risky asset.

1:08:48 So this has a general equilibrium effect in asset prices which get

1:08:51 uh lower.

1:08:52 They,

1:08:53 they end up being 11% smaller.

1:08:55 So this in a way

1:08:56 uh

1:08:57 uh affects.

1:08:59 And tightens the collateral constraint for everyone.

1:09:01 So the,

1:09:02 the,

1:09:02 the households that are

1:09:04 with low,

1:09:04 low wealthy or,

1:09:06 or are constrained today.

1:09:08 End up more constrained.

1:09:09 So we end up with this shift or depending on the distribution of the households.

1:09:14 Uh,

1:09:15 In the end,

1:09:16 uh,

1:09:16 what I find is that

1:09:18 uh the,

1:09:19 the,

1:09:19 the tax

1:09:20 is ineffective

1:09:21 in,

1:09:22 in preventing asset prices drops,

1:09:24 mostly because

1:09:26 highly leveraged households are still fire selling their assets.

1:09:29 They do not internalize that they will get the revenues back.

1:09:34 But the,

1:09:34 the,

1:09:34 the tax is very effective in reducing or preventing aggregate consumption drops

1:09:39 precisely because

1:09:41 the,

1:09:41 the,

1:09:41 the tax is the most effective or the most beneficial

1:09:45 to this

1:09:46 uh share of high leveraged households

1:09:49 that

1:09:49 are taking a lot of risk,

1:09:51 are super indebted.

1:09:52 And once the tight,

1:09:54 the tightening conditions start to affect them,

1:09:56 they can adjust,

1:09:57 they,

1:09:57 they do not have to adjust their consumption because they,

1:10:00 they,

1:10:00 they get the lump sum transfers.

1:10:02 So that's,

1:10:02 that's why

1:10:03 aggregate consumption drops less with this tax.

1:10:08 So to,

1:10:09 to conclude,

1:10:10 uh,

1:10:11 uh,

1:10:12 uh,

1:10:12 as I,

1:10:12 as I,

1:10:13 as I started

1:10:14 in this paper,

1:10:14 I want to study how different levels of inequality affect

1:10:18 the severity of financial crisis of the sudden stop time.

1:10:21 I focus on these cross-sectional effects on the debt deflation mechanism.

1:10:26 And I document that this effect,

1:10:28 there are two reasons why inequality will matter for the crisis.

1:10:32 The,

1:10:32 the amplification side is that

1:10:34 financially vulnerable households will fire sale their assets and this

1:10:38 uh generates more downward pressure on the asset price.

1:10:41 And on the dampening side,

1:10:43 you have these wealthy unconstrained households

1:10:45 or in a different setup could be uh the central banks

1:10:48 that buy these,

1:10:49 these uh

1:10:51 depressed

1:10:52 priced assets

1:10:53 and relieve part of the downward pressure.

1:10:56 So to,

1:10:56 to account for this,

1:10:57 uh,

1:10:58 for this cross-sectional facts,

1:10:59 standard model,

1:11:00 models in the literature cannot do so.

1:11:02 So I propose a small open economy

1:11:04 model with rich heterogeneity across the households.

1:11:07 And I find that the model predicts,

1:11:09 uh,

1:11:10 well,

1:11:10 first,

1:11:10 it can explain stylized facts about certain stocks

1:11:13 and

1:11:14 generate persistent current account reversal or persistent crisis.

1:11:18 Uh,

1:11:19 with respect to the initial dampening and amplifying effects and

1:11:22 the comparison between a heterogenic and a representative economies.

1:11:27 The dampening effect dominates in the sense that

1:11:29 asset prices will drop less when we have this heterogeneity.

1:11:34 Coming,

1:11:34 the,

1:11:34 the dampening effect comes from the,

1:11:36 the wealthy households that are buying these assets.

1:11:39 But we get larger aggregate consumption drops

1:11:43 coming from the high leveraged households that have

1:11:46 these high and larger NPCs and adjust their consumption.

1:11:50 And this is something that we see in the data.

1:11:53 And then lastly,

1:11:54 I,

1:11:54 I did a counterfactual analysis in which I compared

1:11:57 models calibrated to an advanced and a model calibrated to an emerging economy.

1:12:02 And I find that in the advanced economy,

1:12:04 uh larger debt positions are supported

1:12:06 twice,

1:12:07 almost twice as large if we reduce inequality by 15.

1:12:11 And these crises are less severe.

1:12:13 Something that is something that we also observe in the data.

1:12:19 And that,

1:12:19 and,

1:12:20 and that's,

1:12:20 that's it.

1:12:21 Thank you very much.

showAllTimestamps

no

transcript

Crises have happened a lot. Just in the last 30 years, there have been more than 50 sudden stop episodes both across emerging and advanced economies. Now these episodes are characterized by uh declines in asset prices which we know that affect households differently. To give you an numbers about this, during Mexico's 2009 crisis, the share of high leveraged households, these are the households that have a lot of debt relative to their assets, decreased the most their expenditures, around 6.2%. While the share of non-leveraged households, the ones that do not have any debts, Uh, decrease, increase their expenditures by around 5%. This makes sense because these are the households that are, are least exposed to tightening financial conditions. But the heterogeneous effects are not only from the aggregates into the, the, into the households. This heterogenetic component also will play a role in determining the aggregate effects of a crisis. And to be specific, there are going to be two effects that affect the severity of a financial crisis that go in opposite directions. We're going to have a dampening effect and an amplifying effect which will depend on how the leverage and the wealth distributions are in, in, in an economy before the crisis. So to be specific, in this paper, I'm gonna ask the following question. How is the, which are the effects that different degrees of inequality have on the severity of crisis? More particularly, I'm gonna look at the cross-sectional dimension of Fisher's debt deflation mechanism. This mechanism is the standard financial friction that triggers endogenous, uh, endogenously sudden stops as a equilibrium, uh, as equilibrium, uh, choices, as equilibrium behavior in the agents. And to be precise about this dimension, I'm gonna, the two main contributions of my paper are first, I give empirical evidence that such a dimension operates via these two opposing effects. To, to, to, in a couple of slides, I'm gonna be more specific about how the mechanism works and how these two dimensions operate. But uh, uh, as a summary, there, there's gonna be this amplification coming from the high leveraged households that have to fire sale their assets and amplify the drop in asset prices. And there's gonna be a dampening effect on, on, on the severity of the crisis by which wealthy, low leveraged households can buy some of these assets, accumulate more of them, and this dampens the, the aggregate effect. The second question, there's a quick clarification question. Can you define a sudden stop crisis? Definitely, Pierre. Thanks. Uh, so, a, a sudden stop is an episode in which international capital stops flowing into an economy and if the domestic economy is sufficiently indebted, this generates a domestic credit crunch. So we can think that an aggregate shock, let's say an interest rate price increase or a drop in TFP or a change in terms of trade or a combination of all of these aggregate shocks. If the, if the domestic economy is in a, a state of the world that in which it's sufficiently indebted, depends on, on, on foreign, on foreign resources on foreign capital, this generates a, a, a sudden stop. Uh, the definition has been, uh, I think it was introduced by, by Calvo, by Guillermo Calvo, but it's a, a, a fairly standard, uh, definition of, of this type of crisis. So, the, the second contribution of the, of the paper is that standard models in the literature cannot account for, for this cross-sectional facts or the distributional effects. Mostly because these, these papers use representative agent frameworks. And this is where the model of my paper enters. I propose a small open economy framework with a rich heterogeneity across the households. It's gonna be a purely economy and the households are gonna be exempted homogeneous and they will get exposed to heterogeneity coming from uninsurable idiosyncratic shocks. And on top of that, I will add aggregate risk to analysis in the sense that I will not only look at comparisons between steady states or transitions, I will look at the full stochastic stationary equilibrium of the model with, with risk. So, uh, the main results of the model are that it can explain key stylized facts or aggregate stylized facts on solid stocks. But, uh, as a contribution compared to uh, uh, the representative agent version of this model. The model with heterogeneity predicts that sudden stops are persistent. It will take more than 3 years or periods in the model to the current account reversal and the drops in consumption to come back to their pre-crisis levels. And this is something that is precisely driven by 380. When, when the crisis hits, financially constrained households will start at the leveraging process in which they will start uh fire selling their assets and due to the adjustment costs in the different allocations of, of these portfolio choices, this process will take time. So this introduces the persistent element. This is something that the reb agent version cannot account for. Precisely because when every household behaves as the average household, then when everyone wants to sell part of their assets, there is no one that wants to buy that asset. So, this model becomes very similar to a Lucas ree type of framework in which there is effectively no trading and the asset price has to absorb all of the shock. So in that way, the model with heterogeneity will predict that asset prices drop less, the dampening effect will dominate. Other consumption, consumptions are gonna drop more and this larger drop aggregate consumptions in aggregate drops in consumption will come from the high leveraged households which have the most, the largest marginal propensities to consume and are the ones that are get that get uh affected by the tightening financial conditions the most. So that's why we get larger drops in consumption which are consistent with the data. Lastly, in, in, in the last part of the paper, I will, now, I will be ready to use the model to quantitatively assess what are the effects of different degrees of inequality. So, uh, as a motivation, I will compare two different calibrations of the model, one for an emerging economy in Mexico and one for an advanced economy in which I reduce the wealth inequality. And I find that the model predicts that more unequal economies experience more severe sudden stops. So, in this exercise, it's a, uh, what I'm doing is I'm uh changing a parameter that drives a mean preserving spread on this inequality component and we find that the amplification effect dominates. The last part as a policy analysis motivated by the fact that an economy or a government cannot wake up one day and change the level of inequality in the economy, I introduced a dividend tax that precisely reduces this, uh, this inequality component coming from the dividend risk. And I find that the, the tax policy is very effective in prevent, in preventing consumption drops, mostly because the government is ensuring uh this highly leveraged households. But it's ineffective and it is completely ineffective to prevent asset prices drops precisely because these leveraged households are still fire selling. So they, they get the benefits in terms of consumption but they are still fire selling their assets. So now, uh, let me, let me describe to you the, in a nutshell, which is this Fisherian debt deflation mechanism that triggers so stocks. Imagine, let's take an open economy, Mexico for example, and suddenly the United States increases their interest rates. This is an aggregate negative shock. The share of financially constrained households in Mexico will, would like to smooth their consumption with debt, but they can't because they are up to their limits. So they will start fire selling their assets. Now, this behavior introduces additional downward pressure on the asset price and uh uh the collateral constraint of everyone starts to tighten. So financial conditions tighten for everyone. Now, more agents will start to fire sale of more assets. And this feedbacks as additional, this, this feedback loop gives additional downward version asset prices. So that, that is the mechanism. Now, This has been well studied in the literature in a representative agent version, but there are going to be two cross-sectional effects. I, I already previewed them, but let me be more specific here. The first one is a dampening. When we think that in this uh leverage distribution that any economy, leverage distribution across the households that any economy has. There's gonna be a dampening effect coming from the share of unconstrained wealthy agents that have the resources to bear more risk and take advantage of the depressed prices, so they start accumulating these assets and this relieves the downward pressure. On the other end, there's gonna be the share of financially constrained households that as the mechanism, the debt deflation mechanism is triggered, they end up fire selling more assets and more households end up constrained. So this creates kind of a snowball effect in which the, the, the first households that are constrained fire sale their assets. Asset prices are dropping, so more households end up constrained and so on. This strengthens the downward pressure. So as a graphical intuition, let's think that there are two economies which have different, uh, different distributions on the leverage of their households. The first one is the solid blue line, which is, uh, which is represented here and then a second economy represented as the dashed line which is more unequal. There are, there are fatter tales on the constrained households, but there are also fatter tails on the wealthy households. Ex ante, it's ambiguous which of these two effects will dominate in the sense that we don't know which of, in which of these two economies, the crises are going to be more severe. So I will propose this model to as a, as a measurement device to quantify the, the effects. So now, uh, let me, uh, in the interest of time, let me skip, uh, Sergio, we have one question. Maybe that's a good time to take the first question, then I have one for you too. So Alvaro, please go ahead. Yeah, thank you. So just a quick question, in, you're thinking of fire sales of assets, but in most, most countries, the largest asset of households are, you know, housing actually. So how do we think about that here? I mean, uh, You, you told us about the data in Mexico where households that are higher in debt in debited lower consump lower consumption, but do they actually go out and sell the house or their biggest asset or like how do we, we think about the, the mechanism here in terms of the asset? What, what, what asset do you have in mind? Yes, Alvaro, perfect. Uh, so that, thank you for your question, Alvaro. That's a, that's a very, very important point. Uh, here I am talking about assets in a general way, but when I, when, as you well noted, most of the asset portfolio, the, the, the, most of the asset wealth in terms of, in the households corresponds to real estate. And what I'm going to show you in the microdata that I have for the Mexican economy is that in fact, this, uh, depending on the group, or which the households correspond in terms of leverage and wealth, we're going to see that in fact they are reducing their asset positions in terms of real estate. So, uh, This, this, this means that they are either selling the part of the real estate, they are selling the current state and, and moving to a smaller house given the financial conditions, tightening financial conditions, but, uh, but you're right. Uh what, what I'm gonna show you is for real estate assets. We, but also if, if I, uh, account for the whole, uh, portfolio of, of the households, I include financial, financial assets or I include durable consumption goods, the, the, the, the results are gonna be robust. But I think this will be more clear in, in the next in the next two slides where I show you the, uh, the, the, the, the empirical evidence and in, in particular the data that I am using. But that's a good question. Thank you. Uh, uh, Art, uh, go ahead, please with your question. Yeah, hi, Sergio. Um, I, I had two questions. One was actually a little bit of follow up on what, um, uh, Alvaro asked a moment ago, which is that if you take his point even further, in poorer countries, the only sort of asset of a lot of households is actually people's human capital. And then does that completely erase this mechanism, or how do we think about it, uh, because there's obvious limits on, you know. Uh, trade in that case. And the other one is that in your introduction, you referred to how the initial level of inequality essentially governs the, um, you know, whether the dampening effect or the amplifying effect sort of dominates, and I see the intuition for that. But does your model also have implications for how, um, uh, asset inequality and consumption inequality evolves following the shock? Um, I mean, it seems clear from this mechanism through the dampening effect at least that you would get increased. Asset inequality. It's less clear what's happening with consumption inequality. I see. No, uh, thank you, Art, for the question. Um, so yeah, you, you're right in that, uh, I, I. I, I completely agree that in, in, in some economies, uh, human capital would be the, the, the, the largest uh assets for some part of the household distribution. And I haven't thought about that, but as, as a first thought, I would say that There, there are no fire sales in human capital. As you will say, there, there, there's no trade, there's no market for that type of, of asset or, or for that type of human capital. So, um, I, I, I, I would say that This, this, this, this will only, this will, this will be general in terms of assets that can be traded within the column. So that, that will be the first part. For the second, uh, you, the, the, the implications on, on, on other types of inequality measures like in consumption or assets after the crisis are, are super interesting. In particular, I, I will document how the Gini coefficient, which measures how unequal uh the economy is in terms of the consumption. Uh, is behaves before, during, and after the crisis, and the model will have, uh, will, will be able to, uh, rationalize these quali qua qualitative movements on the inequality of consumption. Quick preview. During the crisis, consumption inequality dropped. And this drop is mainly driven by the right tail of the distribution. Uh, excessive lux luxurious, uh, consumption coming from these highly leveraged households that are the ones that reduce the most consumption, uh, translates into lower inequalities in terms of consumption. And that is something that I will capture with the most. One more question from uh Carmen. Carmen, please go ahead. A couple of, a, a, a question and a comment on the dampening effect. Suppose indeed Abstracting from Art's question, suppose indeed that households do sell in the middle of a fire sale. Their homes. Uh, which to some degree happened quite a bit in the 1995 crisis in Mexico. OK, but suppose on the buying side, The dampening effect is muted not because the high income households. do not necessarily want the fire sale assets. They want, say, Uh, assets outside the country, you know, namely, you know, you, you're in a sudden stop mode, um, US assets, so Uh, the wealthy households choose not to buy the domestic fire sale assets but opt to buy foreign, uh, some, any kind of, uh, foreign asset. In that case, your dampening effect isn't, you know, the initial conditions may be there in terms of inequality, but you don't necessarily get the kind of dampening effect. Uh, that you're suggesting. And this is, this is the question. So what if, uh, that is the case? Secondly, wouldn't the dampening effect not be needed? Uh, to be rooted in inequality if a central bank. Like the asset purchase program of the Fed in some, of course, different scale, whatever. But if a, a, a public sector player. Come in and does the actual purchase of fire sale uh assets. In that case, Even if you have very limited inequality to start with, you still get a dampening effect through a third party. In this case, a central bank or, or some sort of government agency that is, is, is buying the distressed assets. I see. Uh, thank you, Carmen, for, for the questions, and, and I think both of them are, uh, or, or, it's one in which the core, the core question is about Who is buying this, this, uh Depressed price assets. In, in, in the first part, if, if the wealthy own constrained households are interested in buying foreign assets, then we will lose part of this dampening effect. The, the, the, the, the story I think here, what, what is relevant is that these domestic households that are risky and unleveraged. know that the price that they are observing right now on real estate is due to fire sales. So it's, it's, they know that at some point, their wealth is gonna increase by buying cheap right now. So this is the story of when is a good time to buy assets during a crisis because they are cheap. So that, that is, that is what I'm thinking when I say that unconstrained wealthy households will buy these domestic assets because they know that right now they are very depressed. And their future prices or fundamental prices are going to be larger. Now, as, as you will say or, or proposed an, uh, uh, asset, asset purchase program from central banks would be a, a, a, a very effective policy in, in, in, in, for future uh research on, on, on, on, on analyzing the effects that this policy can, can uh prevent this deflationary process and, and counter and, and in, in a way stop the, the, the fishery and deflation mechanism. Uh, so, so indeed, if, if it, if, if the dampening effect could come from these wealthy unconstrained households or it could come from another player that is the central bank. Uh, that is, uh, uh, for the research that I am, I'm thinking about, but it's out of the scope of, of this current paper. Here, I'm, I'm only focusing on, on, on the effects coming from the same households. Great. Uh, so, now, let, let me move on. Uh, um, so with, uh, in the interest of time, I, we'll skip the literature review. It's, I'll just summarize it in the sense that this paper contributes to three strands. Uh, the, the first one is about prices triggered by debt deflation mechanisms in representative frameworks. Then, uh, the second literature on asset price determination and equity premiums with uh individual incompleteness market, individual incomplete markets, it means that households are going to be heterogeneous. And the third one is the one that merges these two strands in and analyzes the effects of inequality or the effects of heterogeneity into a macroeconom. So now, uh, before going into the model, let me spend two slides showing you the empirical evidence. So, Mexico in 2009 had a severe crisis. Uh, the MP spread that measures the, uh, sovereign risk on the interest rate increased 200 basis points and aggregate consumption decreased 5%. So even for an emerging economy like Mexico, these are big movements. This was a, a, a severe, it's not the standard real business cycle uh movement. It's a, it, it was a severe so stop. But when we look at the microdata, we see differences depending on the composition of the balance sheet of the households. And to show you this, what I'm going to do is I'm going to use a panel survey for households for Mexico in which I'm going to follow them before and after the 2009 crisis. And I'm gonna sort the households according to their leverage ratio in the uh here in, in the rows in deciles and according to their net wealth in columns. And I'm going to see how their asset positions change before and after. And what we see are 3 things. First, All of the action is happening on the top decile, uh, uh, top decile in terms of wealth. So this is expected as some of the questions already noted in the introduction. Uh, wealthy, some of, some who have low wealth households do not have much assets, and it's not that they do not have much of them. There's not much trade. So, uh, all of the action happens in the top wealthy, uh, deci in terms of the households. And then the second two things that I would like to highlight is that If we focus on the very high leveraged households, these are the top 10, uh, the 10th decile in terms of leverage. We see that these are the households that decrease the most their asset positions. These are going to be proxies to precise quantities. It's not, these are not gonna be, these are going to be deflated by an aggregate index, a house price. So this, this will reflect changes in quantities. So they fire sale the most their assets. Then moving one deci above to the 9th deci, you see that. This group also decreased their assets. So as as financial conditions were tightening, they ended up constrained, so they have to file a sale. And when we moved to the, to the other end of the distribution, the low leveraged households are the ones that bought these, these assets. Then, so this is about the asset composition, the portfolio choice. When we, but ultimately, households care about their utilities, so they care about consumption in a way. And to, to show you the terogeneous consumption dynamics, what I'm gonna do here is only uh in a one dimensional uh Table, sort the the households according to their leverage ratio and see how much they change their consumption before and after the crisis. And here we see that uh the more leveraged the household is, the more uh it has to reduce or adjust their consumption. So this is consistent with uh The high leveraged households having this high marginal propensities to consume, and they are the ones that are the most exposed to, to tightening financial conditions. So even small changes in asset prices that tighten their, their collateral constraints have large effects on their individual consumption and they correspond, and they contribute largely to the aggregate movements in consumption. So that's why we get uh large drops in, in aggregate consumption. we're ready to go to the model, so I'll stop here in case there are any, any more questions. I've got one for you and then Clement has one. Is there a little word in what you showed us about mean reversion that is, you know, even outside of a, of a crisis period, you'd see that the low leverage, you know, would increase a bit consumption of the assets. I mean, the low leverage because they have few assets to begin with and they might want to buy a house, right? Yes, yes. So, uh, so I, that's a, that's a very good point, Pierre, uh. I, I would say in, in mean reversing in the sense that they're gonna be idiosyncratic shocks. So, some households are gonna be having a good luck, having good, good shocks in terms of their income or in terms of their dividends. And probably this is what is driving this, uh, well, no, it's, this is what is driving the enterogenetic component. So even, uh, outside of the crisis or during the crisis, these idiosyncratic shocks are gonna be driving some of the, some of the uh movements. Uh, it would be awesome to have, uh, a larger, more number of waves in this panel. I only have 3, 2002, 2005, and 2009. So, uh, a more detailed empirical analysis on how to, how, how this, uh, idiosyncratic shocks affect the households uh with the available data would, would be, would be great to do. With the, the available data, I, I cannot tell you more, more about uh uh this mean reversing type of effects. Hi, Sergio. Um, I was also wondering about uh But more in terms of uh measurement. So, you know, since you're relying on like the 10th decile of the distributions of assets and assets are typically measured with error, um, can you tell us a little bit about that, you know, and how you would avoid that, that concern? I see. Uh, thank you, Clem Clement, Clement for, for the question. Um, so the household survey corresponds, is, is, is composed of 8, around 8000 observations and its representative at the national level. As, as, as you all said, when, when we start, uh, grouping or, or looking at the top wealthy or the top even 1%, we start to lose, uh, a representative or, or we start, we start to have a lot of measurement errors. So that's why I'm focusing on the 10th person on, on only, on only on only the sales. Uh, with respect to other, uh, to try to avoid other type of, uh, errors, I am using median, median changes. So, uh, in a way, there, the media, uh, we, we know that it's less, less biased against uh the extremes. Uh, so that, that, that is what I, I'm trying to, to use to avoid these, these, uh, possible errors. Besides that, uh, I, uh, there is no, no other strategy that I am currently doing. So we, we've got many questions still. So, so David, Steven, and then Sergio, please try to keep the question short so Sergio can finish. Go ahead, David. My, my question is just whether we, we think of these sudden stops as all alike and all exogenous or, you know, whether, in particular, if they're accompanied by inflation, you would want to be holding debt and, and inflating away the debt, and so, you know, whether this Mexican case is, is representative of how we should always think of these. Mm. Yeah, so thank you, David, for the question. Um, I think the, the literature has, has done a very good job on documenting how these sudden stops happen and, and they happen in, in episodes characterized by uh the, the debt deflation process. So, so. Uh, there are two possible strands here to analyze the sort of stuff. We could think about a real exchange rates in the sense that there is a non-tradable good, so that, that, that suddenly depreciates, uh, this like a real exchange depreciation and, uh, domestic, domestic income, uh, gets tightened and this serves as collateral or this serves as a debt to income type of, of ratio. So the, the households end up more constrained. In, in my setup here, I am thinking about the depreciation, the, the deflation on the uh productive asset, the deflation on the collateral uh asset that, that on the asset that serves as collateral. And I think it's very, uh, the, the, the, the literature has done a good job uh documenting in, in, in uh stylized facts about these episodes that are characterized by declines in, in prices. So, Uh, as you will say, if, if we were in, in a case where there was high inflation, then it would be good to be indebted so you can, uh, you can in real terms, uh, end up. With lower real, real debt, but that's, that's not the case in these other stocks. OK, great. So, um, So now let me move on and go straight uh so Sergio, we still have a few questions. Uh, let's go with uh with Steven and Sergio and Carmen. Maybe we'll let them Sergio move a little bit forward and then take your question. Hi, um, so I have a, a question about this sort of application to the housing market. We're talking about leverage, um, a lot of this is household leverage, but when you have a fall in, um, in house prices, you usually don't have a margin call. Um, so I was wondering how, um, sort of this mechanism would work a bit more mechanically. Um, and also you mentioned the increase in, in interest rates, but, um, I guess the Bank of Mexico would have dropped interest rates, um, substantially at that time as well. Uh, great. So, thank you, Steven, uh, uh, for the questions. Regarding the second one. in fact, yeah, the, the, uh, the policy interest rate target drops or there, there is an easing in the monetary policy, so that's, that, that, that happens. But when you look at the, uh, effective interest rates on, on, on house, on households, there was a, uh, an increase and, and this, uh, this has been Uh, documented and, and that is uh, a common, a common trait on, on these aggregate shocks like an interest rate hike that affects small open economies. Uh, so that, that's about the real, the real rate, not the policy, the policy instrumented. Uh, for the, the, the margin calls on, on housing, I, I, I, I see your point and, and, uh, it's, it's gonna be key for the uh fire selling of these, of these agents. Uh, I, I, I do not have um Like I, I would say that there, there are indeed this uh margin calls in, in an economy like Mexico where the uh large part of the employment is, is informal, around 50 to 60% of employment is informal. So we could think that the self-employed workers. Use part of their collateral, use their real estate as collateral to, uh, to, to, to, to get uh funds for their productive, uh, fund productive process. And there, uh, I would say that they, we, it's not, it's not uh unthinkable to think that there, there's, there are these margin calls in, in that type of, of. Of contracts of debt and collateral. But, but uh yeah, so that, that, that would be my answer. Uh, Sergio, one, quick question. One prominent feature of, um, science stops is, uh, sharp movements in the exchange rate, and that can produce also the significant transfers, uh, between households that linked to the tradeable and non-tradable sectors. Or households that uh have assets in foreign currency versus the ones that don't have a foreign currency. I was wondering whether you consider uh that mechanism as well, but I guess not, but that's also important. Uh, thank you, Sergio. Yes, that is an excellent point. Uh, here in, in the data, I am not considering anything about the real exchange rate or currency risk. We, when I, when I go over the model, you will see that I am not using a tradables and non-tradables type of model in which uh the relative price of, of the, of the non-tradable serves as the real exchange rate that depreciates, which is the, the The movement that you're talking about. But uh in a, in a, in a related way uh in my model, I will have a domestic, uh, a domestic asset which is not traded within foreign to foreigners, which has a domestic price, uh equilibrium price. And when this price depreciate when this price deflates or when this price drops. It is in a way analogous to uh uh the deflationary process on the exchange rate, but explicitly, it's not taking into account the, the currency, the currency movements. There's one last question by Carmen, Carmen, do you want to try to keep 16 so then Sergio can, can move, uh, and then we'll let you move forward. If there's more questions, we'll hold them, OK? Very quickly, to follow up on uh David McKenzie's uh question, not all sudden stops are created equal. You are focused here on the global financial crisis sudden stop, which was really an exogenous shock, a problem in the US and in other advanced economies in Mexico. This is relevant for your offset and it goes back to my previous question. Have you looked or considered looking, I don't even know if it's possible, uh, at calibrating, uh, the exercise to the 1995 Mexican crisis because uh it's, it's pertinent to the earlier point of the offset, OK? If the shock is outside, yes, you, you, the, you know, from the vantage point of the high-income households, it This is a temporary decline. This is a great buying opportunity. Uh, if the shock originates domestically and there's greater uncertainty whether real estate prices will recover because there are all kinds of domestic problems. This is the 1994, 1995 crisis. Um, I am less confident, as I mentioned earlier on the offset part. I actually buy very much the, the effects you highlight on the, uh, on the low income households and their constraints on, on, on consumption, but, uh, on the offsetting side, I think there are differences, uh, across those two, those two sudden stop. Uh, episodes that are more generic, that is, you know, sudden stops originating with domestic problems versus sudden stops that are originating. From a problem somewhere else. I see. Great. Uh thanks, Carmen, for, for, for the uh clarification. Yeah, so here. As, as, as you will say, if the, if the shock originates outside, we precisely rely on the domestic agents to absorb this dampening effect. If, if the combination of, of, of shocks is domestic, then, and maybe domestic and maybe with a more permanent structure, then uh maybe as you will say, it's clear that domestic agents are not going to be the ones that are going to be buying these assets because uh Their, their, their new price is gonna be low for long. So they will try to, to, to get uh other type of assets or diversify with, with foreign type of assets. So in, in that sense, yes, here I am focusing on, on the, the, the, the, the, the, the events or the, the episodes in which domestically there, there was no change. The, the, all of the change comes from the, the interest rate shock or the foreigners. And uh uh those are the, the ones that I, that I will focus on. Or in a way, what I, what I wanted, what I meant to say when, uh when I was talking about all of a sudden stops being equal is that I am focusing on Oh, the domestic, the domestic risky part of, of the soil and stuff is that the domestic economy has to be sufficiently indebted. So, so we have to end up in, in, in a state of the world in which uh the, the, the, the domestic economy is sufficiently exposed to foreigners and then as uh an, an interest, an interest rate increase that would be otherwise uh A real business cycle type of movement triggers these, these sudden stops. But uh I'm, I'm gonna focus only on one aggregate shock, which is the interest rate. So in that sense, This, it wouldn't be a, a model to, to model the '94, 95 crisis. It will be more a model to, to, to, to calibrate for the 2009 episodes. Perfect. So, so now, uh, I'll, I'll move on and describe to you very quickly the, the model framework. So I'm proposing a small open economy Bly model. The Bly component is that the households are gonna be heterogeneous. There's going to be a representative production firm that with given all of the other assumptions, it will behave basically as an endowment economy in which all of the pie of the production is divided between wages and dividends. And then we have the rest of the world with which uh households trade their international bonds. With respect to the the financial markets, there are going to be two types of assets, the international risk-free bond, which the households buy and sell with the foreigners, and a domestic illiquid risky asset. Which is illiquid because there is an adjustment cost associated to it. It is risky because uh the domestic asset price uh is, is, is stochastic and, and will only be traded within the corner. This is a, uh, an assumption that could easily be uh uh relaxed if we introduce an an exogenous demand for this domestic assets which um Uh, could, could, could be like a, uh, foreigners buying part of, of, of the domestic asset or, or leaving the economy in, in turbulent times. Then the financial friction that I am using is a loan to value credit constraint, which means that the international debt that the households have cannot be more negative than a fraction of the market value assets. And lastly, the aggregate shock is the shocks to the interest rate. So, very quickly, let the, let me describe to you the domestic agents. The households have a standard utility function over consumption. This means that they do not care about their labor, so they will supply one unit of labor inelastically and they will be heterogeneous in their bond, international bond positions B, their domestic asset positions A. The shocks, the idiosyncratic shocks that they get on their wages. This is the standard Ayagari type of idiosyncratic shock and a dividend productivity shock that affects the returns on their assets. And this shock is uh analogous to a wealth, uh, uh heterogeneous wealth returns or heterogeneous capital returns that recent literature has shown that are the main drivers of inequality in an economy. With respect to production, a standard, uh, constant return to scale production function that takes, that demands aggregate labor which is going to be fixed and and rents aggregate assets, which is also going to be in fixed supply. So in that sense, the, the profit maximization problem becomes static and profits are going to be zero in equilibrium. So all of the, all of the production is divided between wages and dividends. Finally, to close the economy, the market clearing condition means that the demand of all of the assets traded by the households has to be equal to the fixed supply of these of these assets. And the demand depends on the whole distribution of households across these individual states. So, now we're ready to Uh, formulate the recursive problem of the household. So in a way, it's a very simple model in the sense that households have some idiosyncratic states. There is a, an aggregate shock to interest rate, but the last part is that uh the, the, the whole distribution of households is gonna be also a state. So this adds a lot of complexity. But we, if we think about the, the fundamentals of the model, it's, it's kind of a simple, simple representation. The house funds will maximize their utility function and future value functions subject to a standard budget constraint. In which on the expenditure side, they buy consumption goods, they buy bond holdings for tomorrow at an exogenous price. They buy assets for tomorrow at an endogenous Q price that depends on the whole distribution. And with respect to their income side, they get wages, they get uh the dividends of the shares that they own and they get the income from the current bond positions they bring into the period. Now, secondly, the maximization is subject to this loan to value constraint and this is gonna be key. Here, uh, I, I'm highlighting in, in red precisely where the debt deflation happens. The amount of assets or the amount of debts that the households own cannot be more negative than this fraction kappa of the market value of their assets. So, the, the market value of their assets has 22 components, quantities A and prices Q. So this, this Q here introduces this pecuniary externality in the sense that when a household sells or fire sales their assets A, they do not internalize how this behavior. will affect the, the market clearing price. And if they are selling, this price will drop. So they do not internalize how this behavior tightens their own collateral constraint, but also tightens the, the collateral constraint of everyone in the economy. So that's the pecuniary externality. Lastly, uh, the adjustment cost function or the liquidity component of the asset is quadratic as standard in the literature. So, let me briefly describe to you the marketing completeness because this, this is gonna introduce a trade-off that is very interesting. With respect to the uh the idiosyncratic shocks on wages, given that the, that the household supplies inelastically their labor, the exposure to the earnings risk is fixed. It doesn't matter what the household does if, if, if, it sells or, or buys assets, the exposure to the earnings risk is fixed. But when we think about the dividend risk, which is persistent, this uh risk component combined with the loan to value debt constraint generates, generates an asset wealth trade-off. And what do I mean by this? Imagine that a household gets one additional unit of the asset. This in a way some, gives something good to the household. It gets larger debt capacity. It loosens the, the collateral constraint. And that allows for better consumption smoothing. So, consumption volatility is reduced, something desired. However, on the other end, this household that gets this additional unit of the asset gets additional future exposure to the dividend risk. To give you the intuition, let's think of a household that has zero assets. It doesn't matter if the dividend, if the dividend is high or low. The disposable income of the household wouldn't change. However, if the household relies a lot on, on the asset, on the risky asset uh component, then it, depending on the, on the dividend risk, there's going to be extra uh volatility in the disposable income. So in the end, what we're gonna get is that the 2nd bullet point dominates and asset-rich households will end up getting less debt or even getting positive international bonds. So in a way, endogenously here the model generates that asset wealth, uh, asset-rich households end up with a more diversified portfolio. They end up holding both international bonds and domestic assets. And this behavior generates the unconstrained wealthy agents which are gonna be key for the dampening uh uh effect. And I'm gonna skip this equity premium decomposition that I, that I do in the paper, but basically it shows how both the dampening effect coming from the unconstrained wealthy agents and the amplifying effect coming from the fire sales, uh, interact on the equity premium, uh, the uh uh the equity premium uh result on the equity premium of the, of the economy. So now, uh we're ready to use the, the model uh to quantitatively analyze solid stocks. So the roadmap is going to be the following. I'm gonna show you first how the calibration of the stationary model, uh, the, the calibration of the stationary model in which I keep the interest rate at the long run mean. I'm gonna match some moment, some data moments, and I'll show you policy functions and we're gonna be able to see clearly how the fire sales occur and how this asset wealth trade-off works. Then uh we're going to specify the stochastic process on the interest rate and we're going to solve, I'm going to solve and simulate the, the full aggregate risk model. After a long simulation of, of, of periods, I'm going to identify sudden stop events as the periods in which the current account reversal is two standard deviations above the historical mean. So these are large outflows of capital. And we're gonna, and uh look at the, uh, and the, in the end, we're gonna look at how the severities of the crisis change depending on the decrease of inequality. So very briefly, the calibration in a way is a parsimonious model. I uh take calibrate, I calibrate exogenously the set of parameters that are either common in the literature or, or, or do not require simulation of the model. So risk aversion equal to 2 standard interest rate, long run interest rate equal to 3%. The capital factor share of 13. And uh lastly, the, the CA debt fraction collateral I calibrated to equal, to be equal to the average leverage in, in, in Mexico in 2005. Then for the, the set of parameters uh calibrated by simulation, uh discount factor of 0.90, which is on the lower end of standard uh literature values, but it's, uh, but it allows me to match the net foreign asset position equal to 40% of, of GDP. And the last part which is gonna be key is the trading cost parameter. Here, uh, I do not have that data available data on transaction costs, but I'm, I'm gonna fix. To 5%, average 5% of transaction costs that is in, in the bounds of uh data or in the bounds of estimates obtained by Ayacari and Gertlet for the US. Then we have Sergio. We've got 1 question, 2 questions actually. Is that a good time? OK, let's go. So Tuan and then Alvaro. Uh, hi, Shay. So, um, So in this model, you know, all the households are the same. And uh it's very practical because uh predictions on, on, on the investment and, and consumption uh choices are easy, but so you, you have to assume that the inequality in assets is kind of an exogenous shock. But if we think that the initial inequality is driven by heterogeneity in, let's say risk preferences. Then, then how would you want to think about Uh, uh, this, this identification problem. Uh, in the context of your model, so thanks Juan for, for the question. Uh, so in, in fact, in, in this model every household is, is the same. But they are exposed to the, to these idiosyncratic shocks in terms of their wages and in terms of their effective dividends. So they end up uh optimally choosing different portfolio choices. So they end up choosing, some of them end up choosing more assets, some of them end up choosing more debt. And, and we end up with a, with a distribution of leverage. So in that way, uh, Uh, they, they, they, it's not that they are ex ante identical in terms of their assets or, or, or the distribution of assets is gonna be fixed. This is something that, uh, the, the, the households are gonna be moving along these distributions. So that's, that's, that's, that, that, that's the point. Uh, Sergio, this is Alvaro. So quick question and feel free to defer if I'm getting ahead of myself. So you said earlier that you were gonna do some estimation also for a high-income country. I see that you can do this, uh, away from the serial lower bound, but you're more like if you, when you do it for high income, like how do you get away from the multiplicity of equilibrium if you were to be like close to the CLD? I see. OK. So, uh, great. Thanks, Alvaro, uh, here. Um, so, I, uh, at first sight, I would tell you this is applicable to, to an, to an advanced economy that is far from the zero lower bound. Here I am not taking into account any uh uh effective lower bound considerations in terms of the, of the interest rate. If you think that the interest rate is exogenous even for an advanced economy and uh uh you, you can model it exogenously how the zero lower bound affects this, uh, the, the, the sort of stocks, but my, my, my answer would be like. Uh, In a way, I am, uh, if, if I would like to compare an advanced economy, which is something that I do in one of my uh counterfactuals. Uh, I am abstracting from, from monetary policy restrictions. Great. So now, um, So now, let me go briefly and tell you about the other two sources, well, the two sources of, of heterogeneity in the model are first, the earnings process. Here, there is a very vast literature on how to properly estimate the wage risk on, on, on households. So I, I, I take, uh, some, I, I, I follow one strand of this literature in, in the sense that I Um, estimate a stationary process with a persistent and a and a transitory component. This, uh, in fact, I, I would say that it is one of the very few papers that has done this for an emerging economy and in particular for Mexico. Uh, so, so in, in a way, it's also a contribution to compare uh the estimates of this earnings risk to the estimates obtained for the US. So what I do is using a quarterly Uh, household employment and labor, uh, employment and wages survey, the ENOE for Mexico with, with data from 2005 to 2014. I estimate this earnings risk and I find that the auto correlation coefficient is smaller than the estimates for the US and the variance is a little bit larger. It's, it's on the upper end. Uh, this is, uh, possibly, uh, Uh, this, this results from a combination of a large informal sector market which is more flexible, so this allows for lower persistence on, on, on shocks. And, uh, the lack of safety nets or unemployment insurance also allows for this, uh, or contributes to these larger risks. So to test this hypothesis, then I estimate the same model restricting only for a formal employment and in fact, we see that the auto correlation uh increases. But there's still a difference between the estimates for the US and for the American economy. Then the second source is about the dividend risk. Here, there is uh, I, I lack the available data to properly estimated the process, but what I do to discipline the model is to match the leverage distribution of the households. In particular, I match some moments of this distribution. So in particular, I match the, the percentage of savers, which is around 5%. The percentage of indebted households that have less than the average debt, the average leverage, which is around 60%. And then the financially constrained households which are 18%. So now we're ready to look into the policy functions. Uh, I have 20 I have 20 minutes. So let me show you, uh, very quickly some of these policy functions. Remember, the two choices of the households are next period bond holdings or B and next period asset holdings or A. And these functions are a function of the current bond positions, the current asset positions A, the current shock on the wages, and the current shock on the dividends. So, in this, in this plots, I am fixing the amount of assets that the household owns and I am varying in the X axis the bonds, the, the, the current bond. And here, uh, the, the plot on the left is a standard plot on debt. We have uh seen these plots before in the sense that households get more debt, more debt until they are up to their limits. But when, if we look at the right uh figure, here we can see precisely how the fire sales work. When the household is far from the constraint, their adjustments in terms of uh, of, uh, of, of assets is fairly constant. Either they sell if they get the low dividend, which is the red line, or they buy more assets if they get the high dividend, which is the blue. But when they, they get close to the collateral constraint, we see a large increase in their, uh, in this, in the slopes of these policy functions which reflect the fact that they are fire selling their assets. Then we can do the same exercise but holding now constant the value of, of the bond position and moving around in the X axis, the A prime, the, the, the current assets. Here, uh, if we focus on the, on the left panel, the bond policy function, we show that we, we, we see that if we focus on the, on the blue line. As the households get a little bit richer, they decide to choose as much debt as possible, so they are constrained. Then there is a point where they become a little bit more richer, so they start saving. They are still increasing their debts, but they are leaving part of of their available credit. And then, there's another threshold by which they become sufficiently rich in the sense that they start decreasing their debts and even become savers. So that is what I mean by the asset wealth trade-off in the sense that these households become more diversified. They end up saving in foreign bonds and saving in domestic assets. So the solution of the stationary model very, uh, it's uh very interesting in the sense that uh the model is able to capture non-targeted inequality measures. The wealth Gini is, is around 60.68 and the observed in the data is 0.73. And we also get large consumption inequality around a Gini of around 0.30. So now, uh, going back to the roadmap, so far we have looked at the stationary model and now we're ready to look at the full uh aggregate, uh, aggregate risk framework. So to, to be able to do that, I, uh, first, let's specify the interest rate uh stochastic process. It will follow a two-state Markov process in which the interest rate will take values of 1% and 5% with some persistence. And to be able to solve the model I will uh build on Cruzel and Smith's non-trivial market clearing algorithm and I will adapt it to a small open economy framework. So this requires uh, uh, it's computationally intensive and it requires, uh, it's not, it's called non-trivial because every period you simulate, you have to solve um a fixed point iteration problem on the market clearing price. So, but after doing a long simulation of, of the economy, I identify the sudden stops as the moments in which the international capital or the current account reversals are two standard deviations above their mean and I average across them to create these event study plots. Uh, so, I, I, uh, centered the plots around zero, which is the period of the crisis, and I, I'm showing you plus and minus 3 periods. And first, in, in solid lines, these are the uh the data simulated from the model and in dotted lines are the average of Mexico's 95 and 2009 crisis. So two things. First, uh, one thing that I highlight is All of the sudden stops are, uh, caused by interest rate increases. That is something expected because in the model, the only aggregate risk is the interest rate. But not all of the interest rate hikes. Uh, end up being a sudden stop. Only the ones in which they, uh, there is sufficiently debt or there is a sufficiently large share of, of constrained households. Then uh the model can capture these uh current account reversals that are persistent. As we see in the data. And in terms of asset prices and consumption, the, the, the model captures that consumption drops are larger than the asset prices drops. And although here due to the uh uh scale of the plot, it seems that the model cannot capture this boom prior, boom in terms of asset prices prior to the crisis. Indeed, we get, we get an increasing, increasing uh trajectory in terms of the assets, although it's insignificant compared to, to what we see in the data. And in the last, in the, in the last panel, I am showing you uh one of, related to one of the questions that we got at the beginning about uh future consumption inequality. And we see that the model can capture that during the crisis, uh, inequality in consumption decreased and then after that it starts to increase. So now, let me, let me uh tell you how the model rationalizes the asset and consumption dynamics during crisis. And to do that, I am reproducing precisely the, the, the same uh statistics that I showed you uh in, from the empirical evidence in parenthesis. And now outside of the parenthesis, the, the moments coming from data simulated from the model. So, uh, in a way, not in, in, in, in levels or in qua qualitatively can capture that the unconstrained wealthy agents are precisely the ones that are buying these assets. But more importantly, that the very high leveraged households and wealthy are the ones that fire sale their assets. Now, with respect to consumption. In, in the, in the, in the right side, we see that the model can also rationalize the fact that more as we increase the leverage of the households, they react more to this uh tightening financial conditions. Again, this is because the high leveraged households have the large NPCs and they adjust the most their consumption. So now, uh, in the, in this last part, let me, uh, we're ready to take the model and, and analyze how different levels of inequality affect the severity of crisis. So to do that, uh, there is this recent literature done uh by Alberto Bisin and, and Jess Benaviv in which they show that this dividend risk or heterogeneous capital returns are the main drivers or are the drivers of wealth inequality and in particular the right tail of the wealth inequality. So, with this in mind, I, I reduced the variance of the dividend risk by 150% to proxy an advanced economy. Also motivated by the fact that in advanced economies there is lower confiscation and lower informal uh high returns coming from corruption and better rule of law. And what I find is that the model predicts that low, this, this reduction in idiosyncratic risk reduces uh uh the precautionary motive, uh, motives of the households. So this allows the economy to have larger debt positions. So net foreign debt is twice as large. This is something that we see in the data. Advanced economies have on average larger uh net foreign asset positions with respect to their GDPs than emerging economies. But surprisingly, this, we, we would expect them to be more exposed, but the model predicts that sudden stops are less severe. Consumption and asset price, aggregate consumption and the asset price drops 30% less with, with half of the dividend, with, with, in an economy with half dividend risk. Now, as a sanity check, oh well, first, let me show you the, the, the event story of these two economies. Now in dashed lines, I am showing you data simulated for the advanced economy. First, we, we get the same persistence uh properties and the, the only changes are in terms of the magnitudes of the drop. Consumption and asset prices dropped 30 30% less. And now as a sanity check, uh, I, I, I took inequality measures for in particular the Gini coefficient from data from the World Bank precisely, and I matched these uh coefficients to sort and top episodes identified in a recent paper by Javier Bianchi and Enrique Mendoza. And I match them to, to the episodes and to the drops in consumption and, and GDP. And very simple correlation analysis. In fact, are these two scattered plots. I, uh, I show you, I am showing you in triangles, data for advanced economies and in circles, data for emerging economies. We see two things. First, Emerging economies are in fact more unequal on average than advanced. These are the, the, the circles, the red circles compared to the blue triangles. And also the second thing is that we see this negative relation in the sense that a higher income genus correspond to uh or associated with consumption, larger consumption drops. So this, uh, the, the, in a way the predictions of the model, this, this is just a sanity check for it, in, in the predictions of the model are consistent with this uh correlation that we see in the data. And, uh, lastly, in, in the last 10 minutes of, of, of the talk, let me tell you about this, uh, redistributive dividend tax. So Sergio, we've got a question from Carmen, I think no, and then we'll let you finish. Go ahead, Carmen. Uh, very quickly, I, I think the definition of sudden stops based on the current account do hide important differences between advanced and emerging. Uh, one of the reasons for the less, lesser fluctuations in the current account balances in advanced is the offset by the public sector of what the private side does, which, so, so. I, I think to get at some of the issues you want, you actually want to look at that. You want to look at, if you will, S minus I, but for, for the private sector because I think when you're doing the, if you're doing just, you know, EMs, it's, it's much more comparable, but when you bring in the advanced economy dimension. Uh, during sudden stop episodes, typically public sector saves much less, which, you know, offsets. Seemingly, uh, the, the, the big reversal that you get in the private, uh, um, Guillermo Calvo and I have some ongoing work on that. I see. Uh, great point, Carmen. Thank you. So, yeah, so further, uh, more rigorous analysis on the comparison between advanced and EEM economies. Uh, I would look only at the, at the private, if, if you, if, if we can call it a, a private current account or, or private balance of payment uh flows. So that uh I'll, I'll check that. Uh, to, to be fair here, uh, what I'm, what I'm doing is I am taking the, the sudden stocks already identified by, by, by, by the literature. So, but, but it, for, for further, um, In research on, on this comparison, I'll, I'll take, I'll, I'll take that into account. We've got a question from Burke. Burke. Yeah, hi, thanks. Uh, Sergio, um, you talk about the Gini, and that's kind of sensitive to most of the kind of movements in the middle of the distribution. So am I right in thinking that if you were thinking about, let's say, a general entropy class where it's sensitive to The bottom, you may actually be overestimating the effects of inequality, and then if you are thinking, let's say top 5% or something like that, you may actually be underestimating the kind of like the importance of inequality in these sudden stuffs. Am I right in thinking that or am I missing something? Uh, so, I, I think you're totally right in the sense that I am going to quantify. Two effects That depend on, in a way on the tails of the distributions. So if I look only on, on, on, on a statistic that measures inequality that focuses on, on the bottom. It, it, I, I, I, I will, uh, yeah, and if the amplification effect is stronger, then I will, I will be overestimating it. Oh, and on the other end, if, if I take the, the The top 10, uh, share or, or, or these estimates, I, I will be underestimating you. But uh, yeah, that's, that's true. In, in a way, I am using the Gini, uh, probably in, in future iterations I could look at the shares by deciles or, or, or a more decomposed or uh statistics. Great. So, uh, so very quickly, I, in, in, in the next 3 minutes, let me tell you about this redistributive dividend tax that I, that I analysis that I do. Uh, government cannot wake up one day and choose to reduce inequality or choose to reduce the dividend risk by a decreed, by a decreed or by a law. So what I, what I do is to introduce a Constant redistributive dividend tax are constant across households and constant across time and to return the revenues in a lump sum transfer. So this is redistributive in that sense. And I find one, something very interesting that is with, with this reduction in the dividend risk, again, we get lower idiosyncratic dividend risk which uh gives us a less potent precautionary savings motive. So this has two effects. First, wealthy households would want to demand less bonds or if they have debt, they would want to get more debt. So here uh in blue, I am showing you the, the policy functions of, of bonds. Uh, for the, with the model without a tax and in red with tax. So we see that in fact, they want to get more debt, but in, in the same sense, they want to get less of the risky asset. So this has a general equilibrium effect in asset prices which get uh lower. They, they end up being 11% smaller. So this in a way uh uh affects. And tightens the collateral constraint for everyone. So the, the, the households that are with low, low wealthy or, or are constrained today. End up more constrained. So we end up with this shift or depending on the distribution of the households. Uh, In the end, uh, what I find is that uh the, the, the tax is ineffective in, in preventing asset prices drops, mostly because highly leveraged households are still fire selling their assets. They do not internalize that they will get the revenues back. But the, the, the tax is very effective in reducing or preventing aggregate consumption drops precisely because the, the, the tax is the most effective or the most beneficial to this uh share of high leveraged households that are taking a lot of risk, are super indebted. And once the tight, the tightening conditions start to affect them, they can adjust, they, they do not have to adjust their consumption because they, they, they get the lump sum transfers. So that's, that's why aggregate consumption drops less with this tax. So to, to conclude, uh, uh, uh, as I, as I, as I started in this paper, I want to study how different levels of inequality affect the severity of financial crisis of the sudden stop time. I focus on these cross-sectional effects on the debt deflation mechanism. And I document that this effect, there are two reasons why inequality will matter for the crisis. The, the amplification side is that financially vulnerable households will fire sale their assets and this uh generates more downward pressure on the asset price. And on the dampening side, you have these wealthy unconstrained households or in a different setup could be uh the central banks that buy these, these uh depressed priced assets and relieve part of the downward pressure. So to, to account for this, uh, for this cross-sectional facts, standard model, models in the literature cannot do so. So I propose a small open economy model with rich heterogeneity across the households. And I find that the model predicts, uh, well, first, it can explain stylized facts about certain stocks and generate persistent current account reversal or persistent crisis. Uh, with respect to the initial dampening and amplifying effects and the comparison between a heterogenic and a representative economies. The dampening effect dominates in the sense that asset prices will drop less when we have this heterogeneity. Coming, the, the dampening effect comes from the, the wealthy households that are buying these assets. But we get larger aggregate consumption drops coming from the high leveraged households that have these high and larger NPCs and adjust their consumption. And this is something that we see in the data. And then lastly, I, I did a counterfactual analysis in which I compared models calibrated to an advanced and a model calibrated to an emerging economy. And I find that in the advanced economy, uh larger debt positions are supported twice, almost twice as large if we reduce inequality by 15. And these crises are less severe. Something that is something that we also observe in the data. And that, and, and that's, that's it. Thank you very much.

showAllTranscripts

no

duration

PT1H12M24S

scene7File

worldbank/SergioVillalvazoSuddenStops

scene7Domain

scene7FileAvs

worldbank/SergioVillalvazoSuddenStops-AVS

title

SergioVillalvazoSuddenStops

description

SergioVillalvazoSuddenStops

showTimestampAndTranscript

yes