 So, welcome back. So, let's start the last session of the day. It's going to be multi-sector models and inflation after COVID-19. So, the setup is the same as before. It's Sebastian, he's going to present the first paper. The paper is about the import competition and labor market in the US. The floor is yours. You will have 25 minutes. Okay. Thanks very much for putting this paper on the program. This is joint work with Maria Mithy, Fatih Karahan and Aishwil Shahin and I work with the Fed, so the usual disclaimer applies. Inflation has recently searched in many countries around the world. In the US, it peaked at 8.9% in June of 2022, the highest reading since November 1981. And this behavior of inflation in the current expansion has been very different from inflation behavior in other more recent expansions as the following couple of diagrams try to make clear. So, here, what we're plotting is the growth of core CPI over the course of an expansion. Well, on the x-axis, we have the number of quarters past since unemployment peaked in the preceding recession. So, what this line is showing is that in the 1970s expansion, core CPI grew by about 17% or so after 10 quarters. And then in the following expansions, prices grew slower and slower. So, here, the 80s expansion, the 90s expansion, early 2000s expansion and finally the expansion starting in 2009, where core CPI had grown by just about 5% after 10 quarters. The current expansion is very different in that price growth has been much more rapid and has been tracking more like the 80s expansion than any of the more recent expansions. This strong price growth was led in particular by the goods sector. So, here, we're showing the same diagram just into core goods and core services. And you can see on the left side that the growth of core goods prices surpassed even the growth in the 1970s expansion and has only recently started to moderate where services price growth initially was relatively weak and only recently accelerated. So, what has contributed to this particular behavior of inflation? Well, first, as we've talked about a lot today, they've been supply chain disruptions and they have led to the surge of the prices of imported inputs. Second, there's been a decline in the willingness to work in the US, so the reservation wage has actually risen by about 20%. And as a result of these developments, we've seen simultaneous growth in input prices and in wages. So, on this table, I'm showing you annualized growth rates of different variables in the last four expansions. So, in the first row, we see that import prices have been growing at an annualized rate of about 6% in the current expansion. And if we focus on import prices of so-called industrial supplies, these are inputs into the production process such as rubber, metals, chemicals and so on. These have been growing at an annualized rate of 20% in the current expansion. In the third row, we see that wages have also grown rapidly. So, wage growth has come at about 4 to 5%, about 1 to 2 percentage points faster than in previous expansion. So, in addition to this kind of simultaneous growth of wages and input prices, there's been a shift in the good share in consumer expenditures. So, the share that's going to goods has risen from 36% to 38%. And finally, as we all know, the monetary policy in the US has been very accommodative throughout 2021, as this was already playing out and only started to raise interest rates in 2022. There's obviously been a lot of work on the current inflation episodes. Some of it was presented here earlier today by Beninio and Eggerson. What we bring to the table here is that we examine the surge in inflation to the lens of a two-sector nuke engine model. We're going to focus on, particularly on 2021, though the lessons we draw are more broadly applicable. And we do four key things. So, first we examine the impact of the combined shocks or the wage pressures, supply chain bottlenecks, shift in consumer expenditures, monetary policy. And we show that this combined shock generates about four percentage points of inflation, accounting for the entire pickup of core CPI that we saw in the US in 2021. Second, we examine the role of the different shocks and we find quite intriguingly that the interaction of import price and labor supply shocks amplifies the individual effect of the shocks by about 0.7 percentage points relative to a world in which these shocks happen separately. In addition, the shift of consumption towards goods added another percentage point to inflation and included some additional amplification. Third, we turn to the role of monetary policy and we show that controlling inflation would have required a very aggressive monetary policy at the expense of a deep recession. And then finally we present some regression evidence using aggregate and industry level data and we show that this evidence is actually consistent with this amplification channel that we emphasize here under 0.2. Okay, so let me go and just describe a little bit the model that we have. Given the time, this is going to be relatively high level. So the model we have is the standard UK engine DSGE model with wage and price rigidity with kind of four key departures from the plain vanilla textbook model. So first, there are two sectors, goods and services, and these sectors differ in their labor share and in their consumption share. And moreover, there's competition, but only in the goods sector foreign firms compete in the output market, whereas in the services sector, we assume that all firms are domestic firms. Second, there are two inputs to production, labor and intermediate inputs, and these intermediates in turn can be either domestically produced or imported. And this production structure gives rise to rich substitution patterns. So when the price of imported intermediates goes up, firms substitute towards domestically produced intermediates and towards domestic labor, and this can have implications for wage inflation and domestic production. Third, there's a finite number of domestic and foreign producers, so this gives rise to strategic interactions and incomplete pass through. So when foreign competitors cause rise and they increase their prices, this allows domestic firms to expand their markups without losing market share to their foreign competitors. And finally, we allow for changes in the expenditure share in goods to match the data. So we calibrate this model using standard parameters from the literature. I just want to talk about three key parameters in this model. The first is the elasticity of substitution between labor and intermediates, and we set this elasticity to two in the goods sector or the manufacturing sector, and to 1.5 in services based on some work by Chan in disaggregated Danish data. But in the paper, we actually explore what happens if labor and intermediates are combined in a Leontief production function, kind of thinking about the very short term where you can't substitute between labor and intermediates, and we show that the finding is actually surprisingly similar in that scenario. Second, the elasticity of substitution between domestic and foreign intermediates, which we set to two based on some work by Feinstra that has estimated this elasticity to be in the range between one and four. And finally, we're going to assume a very persistent monetary policy rule to match that the Fed didn't really respond initially to these shocks, and we're going to set this persistence parameter to 0.97, which is just a persistence term in the Taylor rule. So we're going to solve this model in Daener using a second-order approximation, so there will be non-linear effects that we capture in this model. So let me first throw all the shocks into this model and show you that we matched kind of the pickup in core CPI, and then I'll take a step back and inspect the mechanism by looking at individual shocks and how they contribute. So first, let's combine all the shocks. So there are three shocks to capture labor supply and supply chain disruptions. First, an asymmetric shock to the disutility of labor, which is stronger in services than in goods, to capture a kind of experience during the pandemic where services were harder hit. And we calibrate the shock to match the fact that there was a 0.5 percentage point decline in labor supply as documented by one of my co-authors. Second, there's an imported input price shock of 20% to match the observed rise in imported input prices. And finally, a foreign competition shock, which is a marginal core shock of 8.1% to foreign competitors' marginal cost. And we generate this because foreign firms didn't actually gain any market share in the US during this period. So the import to GDP ratio was actually surprisingly constant throughout this period. And so foreign firms must have experienced roughly a similar shock as US firms. In addition to these three shocks, there's this consumption shift. We generate a shift in the utility function to generate a shift in the expenditure share towards goods from 35% to 37%. And finally, there's an accommodative monetary policy shock to keep the normal interest rate zero on impact to kind of match the fact that it didn't raise interest rates. Okay, so what happens when we put all of these into the model? So here I'm just plotting some impulse responses. Here, just the shocks. So we see on the left panel, the labor-dissutility shock, which is stronger in services than in goods. On the right, we see that imported input prices go up by 20%, and we see this increase in foreign firms' marginal cost. And then we have this two percentage point increase in the expenditure share on goods, and this is our kind of normal interest rate pass. So it's zero on impact. And then it's very, very sluggish. It only goes up by about 1% in this model. Okay, so what do we learn? So what happens in this model, when we feed all of these five shocks into the model simultaneously, is that there's a strong increase in domestic input demand and in domestic labor demand. And this is because faced with this strong increase in imported input prices, which we think of as supply chain bottlenecks, supply pressures, from substitute towards making stuff domestically using domestic labor, and this drives up domestic labor demand and domestic input demand. The effect is much stronger in goods than in services because demand shifts towards goods, and so it's precisely in the goods sector where this additional demand comes in. It increases labor and domestic input demand much more strongly there. As a result of these shocks and this disutility shock to labor, there's an increase in the real wage in both sectors, which drives up real marginal cost in particular in the goods sector as well. And so as a result of this, we see significant wage inflation and consumer price inflation, which is about 4%, which kind of matches the core inflation pickup that we saw in 2021. So now that we've done all these shocks simultaneously, let's go and just look at them individually to see what we can actually learn from this model. So we're going to consider these shocks one by one. We're going to recalibrate this monetary policy shock so that nothing like the interest rate path kind of looks the same. So there's no difference in monetary policy across these. So the first thing we learn is a substitution effect. An imported input price shock on its own can actually generate substantial wage and price inflation. That's what we do here. So here we just feed in this imported input price shock into the model, nothing else. And we see that this shock actually generates about 0.5% price inflation. So the intuition is that when the supply chain bottlenecks happen, imported input prices go up, firms substitute towards making stuff domestically in the U.S. And this drives up demand for domestic production factors and puts pressure on two wages. Without any additional shocks to the labor market, you already get 0.5% price inflation. The second lesson is that a shock to foreign competitors also raises inflation in the U.S. So here we just feed in this marginal core shock to foreign competitors. Generates about 1% price inflation. The intuition here is that when foreign competitors are getting hit with a shock, consumers substitute or switch towards domestic producers. They buy more from domestic firms. This again causes these firms to demand more domestic inputs and drives up wages and so on. In addition, domestic firms now have greater pricing power. They can actually increase their markups because their foreign competitors have just had to increase their prices. And so they'll take advantage of that. And so that generates about 1% inflation. And then finally, and most importantly, there's an amplification effect. So this joint supply shock, this joint effect on the labor market and on import prices, has an amplified effect on inflation. So to illustrate this, let me just show you the third of these individual shocks. This labor disutility shock, that's the blue dotted line here. And now let me just feed all three shocks into the model separately and just add up these lines. So that's just this red dash line. I'm just adding up the gray, blue and green line. Now what happens if I put these shocks into the model jointly? Well, what happens is that we find significant amplification. So wage inflation increases by 1% point more and price inflation increases by 0.7% points more. And the intuition here is that in normal times, say in normal times there's some pressure on the labor market, normal times firms can then substitute away, say from using domestic labor towards using Chinese labor, importing products directly from China, thus blunting the effect of this cost increase on their prices. So they can kind of play around with their inputs to reduce the cost pressures that they're facing. But in this environment where both import prices and wages are going up simultaneously, they can't substitute between these factors of production. So they have to take these cost pressures and therefore increase their prices by more. In addition to these shocks on the cost side, there was this shift in expenditures towards the goods sector. So how does that contribute? So let me just put the shock separately into the model by itself. It raises inflation by about 0.5 percentage points. This is because when we feed the shock into the model, there's a shift towards the goods sector and the two sectors are not symmetric. Good sector uses less labor than services. And so on net, there's actually some wage inflation and some price inflation that we generate from this. Now the blue line is just the joint supply shocks that I just showed you before, which was about 3% inflation. And then the red dash is again just adding up this consumption shift and the joint supply shocks. What happens when we feed in these shocks again together, the consumption shift and the joint supply shocks, we get some additional amplification about 0.4% extra price inflation. And this is because this expenditure shift shifts precisely towards the sector that is experiencing the strong supply chain pressure, strong wage pressures and so on. So you're shifting consumption towards the sector that is already under pressure and that generates some additional inflation as we see here. And so in total, we get about 4% inflation. Okay, so what could monetary policy have done? So the Fed was very accommodative in 2021. So we're going to run an experiment. We're going to ask, well, what if the Fed had been more aggressive? Okay, so we're going to compare our baseline policy with this very persistent Taylor rule and accommodative monetary policy shock to two alternatives. One is standard Taylor rule, which is not as persistent, which we calibrate based on literature of persistent of 0.8. And second, an aggressive policy where we put even lower persistence and a much more aggressive response to inflation. And we ask, what are the implications for output labor demand and inflation under these alternative scenarios? So here's some variables that we look at. The normal interest rate, average wage inflation, averaged across the two sectors and price inflation. So the left panel just shows basically how these policies look like. So the black line is what I showed you before as the baseline. The red is the standard Taylor rule. It increases a little bit more. The blue is a really aggressive policy. Basically, the Fed raises interest rate by 7% on impact. Interestingly, what happens is that there's not that much of a dent to wage inflation and price inflation when we do the standard Taylor rule. So these are actually still relatively high. You really need this aggressive monetary policy to really make a difference. And the reason for this is that most of this inflation is driven by supply side shocks. And so there's only that much that monetary policy can do really to bring down inflation. However, this aggressive policy comes at a cost. It generates a deep recession as we see here. So the blue line, both gross output and consumption go down significantly. However, if you want to think of labor demand as kind of the flip side of unemployment, we actually still see strong labor demand. And the reason for this is, again, that in response to this kind of foreign shocks, this supply chain shock, there's a shift towards making stuff in the US towards domestic production and that props up the domestic labor market. So actually, with our baseline policy, labor demand goes up by 12%. And even with the aggressive policy, there's still a kind of a tight labor market because more stuff is getting made in the US. And this kind of matches what we see in the data to some extent that the labor market is very tight. Okay, so in the last couple of minutes, let me try and provide some empirical evidence in particular for this amplification channel that I've been talking about. So our model generates increasing marginal cost, so in price and wages and some amplification. So can we see this in the data? So I'm going to do two exercises. First, I'm going to look at some aggregate data. I'm going to run some local projections to trace out IRFs with aggregate variables. And second, I'm going to turn to industry level data to run panel regressions for about 506-digit NAICS industries. So first, let's go to the aggregate analysis. So here on the left-hand side, I'm going to regress inflation. I'm going to use the PPI. I'm going to have finished goods producer price inflation and some quarter T plus H. And I'm going to regress this on input price inflation in orange and wage inflation in green to see how changes in wage inflation or input price inflation are correlated or passed through into producer price inflation H quarters down the line. And we're going to have some lags just to capture any persistent dynamics. So we're going to estimate this regression for H running from 0 to 20 quarters, up to 20 quarters ahead for the period from 1988 to 2022. And for both input price and wage inflation, I'm going to find that H was the highest correlation. So basically, we're going to get nice, hum-shaped input responses. I want to find the gamma H and the beta H was the highest correlation. And I'm going to see how that kind of peak pass-through varies over time. So I'm going to take that. I'm going to estimate this regression with 25-year rolling window. And I'm going to look how this gamma H at the peak H, which happens to be at five quarters, and this beta H at its peak at nine quarters, how those have evolved over time. So what do we see? Well, we see that this correlation, this pass-through, has come down over time. In particular, wage-to-price pass-through was basically zero throughout the 2000s. And I have another paper with my co-authors where we have kind of argued that this was because of input competition from China, that there was basically zero wage-to-price pass-through. But importantly, in the recent period, there has been a jump up in this pass-through from input prices and wages into producer prices. There's been kind of a reemergence of wage-to-price pass-through. And that is consistent with the story I've been trying to tell in the current period where firms are facing both input price pressures and wage pressures. They are forced to pass through these wage pressures into producer prices because they can't substitute. And the Chinese competitors are also experiencing a similar shock. Now, we can actually test this more directly. We can actually just throw in this interaction term into this regression directly and trace out this roll age. So let me just run this regression from age zero to age equals 20 and see how that input response looks like. And indeed, we find a positive and significant input response for this roll. So indeed, inflation is higher when both input prices and wages go up at the same time. Normally, this is just aggregate data. So a lot can be there in aggregate data, maybe expectations are changing and so on. So the last thing I want to do is I want to do some industry analysis. So I'm going to use my 506-digit NAICS industries. And I'm going to run a kind of a similar regression. So I'm going to have on the left-hand side four-quarter changes of the producer price index. And on the right-hand side, I'm going to have four-quarter changes in input prices, which we construct using domestic and imported input prices using the IO matrix. We're going to have wages. And we also now can control for competitors prices, foreign competitors prices, which are important. So for example, the competitors in the car industry would just be the price of imported cars. So we can just get these input prices from the data. And then importantly, we can control for changes in productivity. And we have industry fixed effects and time fixed effects, which soak up any aggregate changes in expectations and so on. But let me just estimate this regression for the good sector, where it's kind of the most, the results are the strongest, the results for the services sector and the paper. So we estimate this regression using both these terms and levels plus additional interactions for 2021 to see if things have changed in 2021. And so here's what we find. So we find that in the pre-2021 period, the pass-through from input prices into producer prices was about 24%. But in 2021, this pass-through increased from about 24% to the sum of these two that increased to about 40%. There was again this pickup of pass-through from input prices to producer prices firms are passing through more of the input price shocks as they are forced to do because they can't really substitute based on our kind of model story. For wages, we saw zero pass-through in the pre-2021 period and then a jump up to 13% wage to price pass-through. And finally, the correlation of domestic producer price changes and foreign competitors price changes has also increased kind of consistent with all firms experiencing similar shocks and US firms taking advantage of the fact that their foreign competitors are raising their prices to also raise their prices in turn. What happens if we put in this interaction term into this regression? We find a positive and significant interaction term kind of saying that the inflation effect is large when both wages and prices go up together. And importantly, once we put in this interaction term, there's no additional effect from 2021. So this entire pickup and pass-through is explained by this interaction between input prices and wages. Okay, so let me conclude here with the kind of key takeaways from this talk. So we can interpret these supply chain disruptions that we've seen as a partial reversal of the disinflationary effects of globalization that we have enjoyed over the last two decades. And so even though some of these supply chain disruptions are going away in the current period, you can think of changes in trade patterns as countries are strategically thinking about sourcing from China and so on as potentially having similar effects and keeping inflation a little bit higher than we were used to during this decade where really there was a lot of sourcing from abroad and taking advantage of competition from China. Second, we find that part of the tightening of the labor market is due to supply chain disruptions on their own as firms have substituted towards domestic production. Third, this joint input price in wage shock has amplified the inflation impact and the consumption shift has further increased inflation. And finally, kind of being kind of positive about the Fed's role in monetary policy would have needed to be very aggressive to reduce inflation in this model because of the key role that was played by supply side shocks in our framework. Thanks very much for listening. Thanks a lot, Sebastian. The discussion is Katya Paneva from the Federal Reserve Board. You have 10 minutes. Thank you. Thank you for inviting me. Thank you to the organizers. It's a pleasure to be back here. It was a pleasure to read the paper. And I just really want to emphasize the usual disclaimer that my views are my own. They don't represent any of the views of my colleague at the Federal Reserve Board. And my comments are generally going to be, you know, if I was a better round-dead economist, maybe the comments would be different, but my comments would be from the perspective of somebody who's really spent the last three years looking through every theory and explanation of what we've seen over the last few years. I've been doing this for more than three years, but the last three years were particularly exciting. So just a brief recap, no equations. What is the motivation for this paper is to construct a two-sector New Keynesian DHE model with multiple factors of production, foreign competition only in the good sector, and endogenous markups. And the model is used to evaluate the effectiveness of monetary policy in two scenarios. In the first scenario, the recently high inflation is driven by supply chain disruptions and labor market disruptions. And in the second scenario, inflation is demand-driven, mainly because of the switch from services consumption to goods consumption during the pandemic. The main findings, at least my big takeaways from the paper, were that the combination of supply chain disruptions and labor market disruption, disruptions pushes inflation more than the two shocks would individually push it than the combined effect of the two shocks separately. Why? Because the firms cannot easily substitute when the price of intermediate materials goes up in order to lower costs. They would use more labor, but labor is also more expensive because this utility in the labor market increased. And also the supply chain disruptions affect foreign competition and their prices as well, which allows the domestic companies to pass through the higher costs to consumer prices. So the model can also be calibrated to produce the same increase in inflation just using a demand shock to goods. So you can get the same outcome for inflation, but via different channels. And depending on what the different channel is, the paper says, if the increase in inflation reflects supply chain disruptions and labor market disruptions, less aggressive monetary policy is better. Otherwise, you heard the labor market too much. If the increase in inflation is the main driven, you need to act faster and raise rates early in order to avoid raising rates even more and causing a recession later on. So that's a brief recap, very intuitively. I like the paper a lot. I'm particularly partial to papers that look at different sectors. And my dissertation many, many years ago was on the goods and services sectors, but here I wanted to show a chart, which many of you have seen many times, which illustrates why one model might not explain what happened during the pandemic. And I'm only focusing on core inflation because I think energy did its own thing and it's very hard to come up with the micro-DSG model that can explain what happened to energy and core inflation or might attribute more, let's say, to supply chain disruptions only because energy went up. So again, this is only focusing on core and I've split it into core goods, housing services and core services excluding housing. And you can see how different the behavior of the red line, which was inflation in core goods, how different in terms of timing and magnitude was than what happened in services. And I'm excluding housing. It went up much earlier and it went up a little bit earlier and it went up by a lot more than services. It also has come down, which we have not yet seen it in services. So there are some differences in both the magnitude and the timing. And that's why I think looking at the paper that looks at goods and services sector separately is very helpful. The model is realistic and first I had, but complicated. I mean, it's realistic and complicated. You can't have a realistic model without it being complicated. The amplified inflation effect from the combination of simultaneous shocks is very plausible and relevant. However, some of the assumptions and the shocks are a little bit harder to understand. And I'll just start with the big picture, which applies to this paper, but I think several of the presentations today was that in reality, what happened during the pandemic was a combination of supply and demand shocks. And it's very hard to disentangle and for simplicity, we label it differently in the papers, but they are not independent. So in this model, oh, okay, so the title of the slide is what is a supply chain disruption? And in this model, it's an exogenous increase in the prices of imported inputs. And in several of the presentations today, I kind of hurt people maybe using in the same way supply chain disruption as a supply shock, but they are not the same thing. And so my colleagues and I have struggled a lot with it and we started throwing terms like bottlenecks and shortages and supply chains. And we made this simple supply demand econ 101 diagram to think about through this a little bit, just for us to clarify it. And I still find it very helpful. So let's say you start with S0 and D0 and you're at the equilibrium point A if there is a demand shock, you get to the steeper part of the supply curve and right there you start seeing pressures on the supply chains because you hit a very vertical, like very steep part of the supply curve. You can also get a supply shock, right? A supply shock would be the move from S0 to S1 and you get at B. And so my point of this slide is that in this paper it's kind of like import inputs increased for their own exogenous reason, but in reality they did increase because demand for goods increased and there were supply shocks at the same time, factories closing, microchips missing, right? It was a combination of that. So that's what I struggle with some of the papers and if we label them supply and demand and we make recommendations for monetary policies to go back and ask was the increase in import and inputs really a supply shock or was it also driven by increased demand because of accommodative monetary policy in the US? For example. And then if we are going to take a stand there what would be the implications for monetary policy being? My other struggle reading the paper was the substitution effect. So when I read these papers I always try to come up with examples in my head then so I understand what's going on and just to make sure it makes sense, right? So in the paper an increase in import in the price of imported inputs leads to an increase for domestic inputs. I totally can see this, right? It happens and which then in turn leads to increase in demand for the less domestic labor because now the inputs will be produced domestically and labor will be needed. As long as there are substitutes I can see there. There are plenty of imported inputs that have no substitution so we won't see it there. The part that I struggled with a little bit more was that firms also can substitute away from intermediate inputs towards labor and I get it, right? These are our models but I was thinking well if you make furniture and you need a specific input like there's only that much you can do. Wow, okay. To replace inputs with labor and I think Sebastian said a Leontief function I didn't see it in the paper but if such a function produces the same results I would totally go with it because again it was a struggle for me. I'll skip the next step. I think because of this substitution towards labor which particularly hits the good sector the left panel are the responses of real wages in the good sector which is the black line and the red one. It doesn't matter whether I take this panel from figure 6, 7 or 10 it doesn't matter what the shock is in the paper it's always wages in the good sector that increase by more but in reality what happened if you look at on the right in the ECI measures in the US it was wages in services that increased more during during and after COVID. So that's something to think about when matching to data and my last point because I don't have time for more would be the assumption one of the assumptions in the paper is that wages and prices are equally sluggish because of the rapid adjustment of wages in the recent period I just wanted to it was a Fed's note posted maybe two days ago so there's no way the authors could have known this but the frequency of price changes also increased markedly during the pandemic so if this is an important assumption for the results that prices are equally sticky during the pandemic and before any prices and in wages it's something for the authors to look at so that's it very nice paper I enjoyed reading it it made me think about channels and interactions that they're often using from the micro models and the globalization in general so thank you thank you then let's collect some questions from the audience I see John there Joe Hazel from the London School of Economics thanks for a great really interesting paper one question I had I was quite surprised that as I read it by itself the shift in consumption from services to goods didn't seem to have a big impact on inflation per se even though this is very very big only an interaction with the other shocks so let's go through why this isn't and hopefully you can tell me in a second I guess it's something to do with the fact that to a first order inflation is going to rise in one sector and fall by the other as a consumption shift across sectors so if that is right I was wondering if that's realistic one can imagine mechanisms where for instance wages rise in the goods sector but they don't fall in the services sector so per se the shift doesn't have big inflationary effects which is sort of missing in your model so I guess in general I'd be curious if you could comment more about why that very large effect very large shift sorry doesn't have big effects on inflation except when interacted with other things thanks great thank you very much one extra question so let me try and respond so we have a slightly newer version of the paper which I'm sorry send you so a little bit of supply versus demand shock so as you picked up and as others have picked up it's very hard to try and separate the demand and supply shock so we want to get away a little bit from that so I guess we're just trying to consider a couple of shocks that we've seen import price have gone up we feed this into the model but we don't necessarily want to take a stand on what fraction of this is due to demand versus supply we're just looking at them but we're not trying to put anything to demand versus supply on the substitution channel so so I guess from your example I guess because this is an aggregate model I'm not thinking about a firm that is making furniture and buying wood from China then suddenly using its own labor to go out and chopping wood I guess I'm thinking there in reality they're like two firms it's like a domestic firm that is producing wood in a domestic firm in Jamaica and instead of using a Chinese wood manufacturer this firm is now using the domestic supplier so it is using domestic labor but it's not within the same firm there's like two firms but they get aggregated together in the model that's how I'm thinking about it we do have a Leontief version where basically labor and intermediates are compliments and we still find a very significant inflation effect the reason is as long as there is substitution from foreign intermediates towards domestic intermediates so you're substituting those then because labor and domestic intermediates are compliments if you're using more domestic intermediates you also want to use more labor so even with compliments you still get a big boost to labor in that case finally on the wages and services which you showed go up more we actually have a version of the model where we have heterogeneous labor in low-skilled versus high-skilled labor and it turns out that most of the wage increases in services where it's a low-skilled part of the distribution at least not better reserve employees didn't really get much wage boost so for the lower-skilled labor in services we did see a much stronger wage increase and we can generate this in the model as well when we put heterogeneous labor so finally Joe's question yeah I think your intuition is right I think when we have to shift on its own there's some disinflation in services and some inflation in goods and overall the effect is relatively small I guess we we should definitely think more about you know what happens if you can't lower the wages in services and so on I guess we we haven't done that but that's a good suggestion thank you so thanks thank you very much and then now we turn to Elisa Rubo who's who's online and she's going to talk about what drives inflation lessons from disaggregated price data thank you very much for having me and thank you for letting me present online I'm trying to look for I'm trying to share my slides I'm sorry I cannot see the window yet right now there we go all right can you see my slides now yes okay great now let's see perfect okay all right so I guess I don't have to convince this audience that inflation during the covid period was a puzzle and people wondered whether the main determinant of inflation was coming from supply side such as bottlenecks and shortages or from the demand stimulus packages that were put in place in favor of the supply side story people observed that during the covid episode there was not just a large increase in aggregate inflation but there was also there were also large movements in relative prices which could somehow suggest that sector were fit in different shocks and this might have some consequence also for aggregate inflation it is kind of hard to think about how to connect relative prices and aggregate inflation in the kind of textbook models that we use to think about monetary policy and inflation like the new Keynesian framework essentially because these textbook models have a representative industry and so essentially cross sectional shocks across industries cannot be modeled and so it's hard to think about their impact on inflation so in this paper what I do is I build on my research agenda and that allows for multiple industries and in particular the key innovation of the paper is that I allow for multiple primary factors of production and I will try to convince you that adding this is really key to thinking about supply bottlenecks and what they show is that in this setting there is an inflation approach coming from relative shocks to TFP and demand across sectors and I will also show you a method to identify which component of aggregate inflation is driven by these cross sectional shocks across industries and which one is driven by aggregate shocks so plausibly aggregate demand so what's the intuition and like what's my method in a nutshell well there is I show that just by looking at price data one can use apart two drivers of aggregate inflation one is given by deviations of aggregate output from an appropriate notion of potential output so that's kind of what we tend to identify with typically demand shocks the second component is what I call an inflation output tradeoff coming from cross sectional things so changes in the desired relative prices across industries and in particular the second component which is absent from standard models I show that is actually important and quantitatively large and what's the intuition why cross sectional shocks could matter for aggregate inflation well essentially think of a setting in which there is some negative TFP shocks or a sudden increase in demand for industries that are downstream so close to final consumers or they have more flexible price or they are somewhat inelastically supplied so supply constraint so these shocks I show that actually generate aggregate inflation because they have either a large effect on prices or they are close to consumers so they have a large effect on consumer prices even though the central bank is not moving demand away from potential and once I established that this is the case then the next question is well can we use data to identify how large these effects are and how important have they been in the COVID setting what I show is that essentially the relative price movements induced by aggregate demand say monetary policy or cross sectional shocks are never collinear so basically just by looking at the behavior of relative prices which direction they moved we can understand whether the movement was coming from monetary policy or it was coming from cross sectional shocks and I applied this methodology to the inflation in the US and I find that in 2020 inflation was mostly driven by supply shocks so the early part of the pandemic it was entirely bottlenecked supply shortages shifts in preferences and so on but after that the demand component has become more and more important and I would say it accounts for about two thirds of inflation starting from 2021 onwards so let me skip the literature in the interest of time I will now briefly go through the environment so that I give you a sense of how the model works, it's related with the benchmark Nick engine model I will show you why this model is useful to think about what are next and I'll then move on to the main theory and quantitative results so here is just a schematic representation of how I think about the economy so the economy is going to have a set of primary factors that are labor and capital assets and these primary factors are going to be hired by heterogeneous industries that's the middle layer the producers and industries that sell their products to final users so there's going to be consumers, investment producers and the government so right now I will walk you through the way I model all these agents and what are the relevant dimensions of heterogeneity that will be important to generate supply-driven inflation alright so as a quick overview the dimensions of heterogeneity that will be important are heterogeneity across primary factors in particular their supply elasticity that's going to be the key thing and also their wage rigidity and I will allow also for factor specific supply shocks in the model then there's going to be heterogeneous industries in that industries use different primary factors they have different price thickness and they face heterogeneous TFP and markup shocks and finally the final users are going to play a role in the model because they have different consumption models and they face shocks in their consumption preferences like that could be a model of the shifting preferences from away from services in the beginning of the pandemics and then back to services when lockdown ended so just to recap the model will allow for things such as the state shocks but also the fact that maybe some industries had kind of a negative TFP shocks because during lockdowns people could not go to their workplace and they could not be as productive really when working remotely it could account for inflation so kind of an increase in desired markups it could also account for factor supply shock like people just don't want to go to work because they're afraid of getting infected so with this setup let me just formalize a little bit how I model this setup I will have these many heterogeneous households their preferences are relatively standard so like in the plain New Keynesian model they will derive utility from consumption and have some visibility from labor supply the key thing is that the household consumption preferences are subject to these relative preference shocks I also have a model of a stylized model of investment and capital utilization this is essentially because to model bottlenecks it turns out it will be very important to have specific capital factors and this utilization model basically allows me to get to a supply curve for these capital factors that is very simple it's kind of like the consumption leisure trade-off for consumers so the quantity of capital that is available has some inverse elasticity phi with respect to the real weight of the capital so the rental rate over the cost of utilization firms as I anticipated I allow for a general input-output network so firms will hire different primary factors in intermediate inputs and they will also be subject to input demand shifters that are kind of like the consumption demand shifters is kind of a minor point at the industry level aggregation is tricky when you have a terogeneous agents model but I define it in the same way as the national accounts so in particular I will focus on the connection between inflation which in the model is going to be measured as the GDP deflator like aggregate inflation and a notion of the aggregate real output aggregate real GDP is going to be the equal to the income weighted share of changes in consumption and investment across final users which is also the equal weighted share of changes in employment of primary factors so the output gap definition is going to be based on this real GDP and kind of the main inflation measure that I'm going to look at in the model is going to be the GDP deflator in the data I look at consumer prices as well great so to close the model one needs to specify monetary policy to keep things simple I assume that monetary policy the monetary policy rules pins down aggregate nominal GDP through a cash in advance constraint this is basically without loss of generality as you will see monetary policy enters the model only through the output gap so it doesn't really matter whether it's determined by a money supply rule or a tailor rule for inflation the composition that I show great equilibrium is defined in a standard way so now that I have brought you all this complicated setup let me instead move to explain why I think having all this richness is important to understand the role of supply bottlenecks and to do so I will show you kind of a minimal model of a bottleneck so what is this minimal model of a bottleneck like essentially it's a model where there is an economy with two primary factors of production one primary factor you can think of as labor that is like a more elastically supplied primary factor and then there is another primary factor which you can think of as some fixed capital or for example the land that ports need to receive the shipments or like the trucks that the shipping industry has and so on so we have these two primary factors that contribute to a final good what is a bottleneck in this simple framework well it is essentially a decline in TFP or an increase in relative demand for the inelastically supplied primary factor so that's what's going to cause a bottleneck and really this heterogeneity in the supply elasticity is what matters so what does a bottleneck do in practice well it does two things first is going to lower the potential output of the economy and second is going to create an inflation output tradeoff so here I'm introducing a key distinction that I will maintain throughout the paper that is inflation can come from two things one is output is deviating from potential two is what I call the inflation output tradeoff that is inflation that would happen even if output is at potential so how do bottlenecks affect these two aspects well first why do bottlenecks lower potential output essentially because they shift expenditures towards inelastically supplied factors so they are kind of shifting expenditures towards a tight part of the economy so even if aggregate productivity is constant the fact that we are shifting demand towards a tight sector means that the efficient aggregate output is going to decline second effect of bottlenecks they create an inflation output tradeoff at least for consumer prices so the GPD flater why is that the case well here the intuition is a little bit more complicated but essentially we said these bottlenecks kind of like tend to shift demand and expenditure towards these tight sectors in the economy so in an efficient economy what would happen is that the price of these sectors shoots up and so yes demand shifts a bit but like there is some self-correcting mechanism where by the price increases and so demand kind of shifts back a bit away from these bottleneck sectors however if prices are sticky this price mechanism is not going to work or it's going to work only partially and so essentially in the sticky price economy demand for the bottleneck sectors think the shipment sector in the early phases of the pandemic remains too high because the prices don't increase quickly enough and so this beats up the demand and the price of the primary factors that are used in the bottleneck sectors like the shadow value of trucks or port land increases a lot okay and it is true that at the same time other primary factors like labor are in efficiently low demand but the price of these primary factors is not going to respond as much because these primary factors are more elastically supplied so their price does not respond as much to quantity so essentially on average factor prices go up due to these inefficiently high demand for the constrained factors and that's what eventually is going to drive the inflation output trade off because this average factor prices go up then good prices also go up and so on so here is just like the simplest possible example that I showed that I talked about we have this economy with just like two primary factors one has high elasticity like is like elastically supplied the other is in elastically supplied and the supply elasticity is captured by the inverse of fresh 1 over 5 in the equation and essentially if we are shifting demand towards the inelastically supplied good then this covariance is going to be negative which means that here I show aggregate output declines and that's literally we are shifting expenditure to the tight part of the economy so we get lower potential output let's consider the same demand shock like we're increasing demand for the inelastically supplied sector and look at the effect on consumer price inflation here we see that essentially consumer price inflation is going to depend on the deviation of output from potential that's what I call y bar here but it also depends on the covariance between the inverse fresh and the demand shock so in particular if demand goes up for inelastic sectors so this time is with a high fresh this means that we'll have positive inflation even if output is kept at potential so that's a general issue it holds in this example it holds also in a fully generally input output network and that's essentially the CPI or the GDP deflator are going to be subject to this endogenous cost push shocks coming from changes in relative demand or TFP across sectors interestingly I show that kind of the intuition that cross-sectional staff can cancel out on average still holds just not for the CPI you need to choose your average very carefully and I show how you can do this and essentially I show that there is always a price index whose weights do not depend on the shocks that is not subject to this endogenous cost push shocks and how do you get to this price index essentially you down weight the inelastically supplied goods so the volume of sectors should be down weighted inflation there doesn't mean that aggregate demand is changing it might just come from an increase in the relative demand for those sectors okay so this was just an example let me show you briefly how you can generalize this argument to add your economy construct and inflation the composition so first like it's a key definition is going to be the definition of natural equilibrium so because that's what determines what's the potential output and what's the output gap so essentially my definition of natural equilibrium is I consider a flex price economy subject to all the same shock as the actual economy with sticky prices and I define the natural output and natural employment of each primary factor and natural relative prices as the output employment and relative prices that would prevail in this flex price economy subject to the same shock as the actual one okay great so with this definition we can look at the main variables that I'm going to relate in the model so I will the indigenous variables will be sector level prices and inflation factor level primary factor level employment gaps and the aggregate output gap the exogenous variables will be essentially all the shocks that they throw into the model and here is the first important result so I'm going to throw basically all sorts of shocks into this model and it turns out that they enter the equations only through just one variable that is what I call the price wedges chi so the shocks that they include are sectoral TFP and preference shocks and shocks to desire markups and then I will have shocks to government spending and shocks to fiscal policy through lots of transfers on top of this there will be the monetary shock that determines aggregate demand basically all the shocks except for the money supply are going to enter the model through these price wedge objects which capture the difference between the initial prices the previous prices that the economy enters with and the desired relative prices that's the natural prices okay so with this notation I connect the indigenous variables the prices in the employment gap through the Phillips curves so here we have many sectors so we don't just have one Phillips curve we'll have one Phillips curve per each sector connecting inflation with the aggregate output gap okay so I'm not going to go through exactly what goes into the slope and the cost per shock for each sector but I wanted to recall that we can extend the usual Phillips curve to just derive it sector by sector and sectors will have different slopes with respect to the output gap and this is the key thing that these slopes will never be collinear with the effect of cross sectional shocks on inflation which is the blue term okay so this Kappa slope will be always non collinear with the blue term and that's how we're going to be able to separate the two in the data once I have the sector of Phillips curves I can just average them out across sectors and combine them into our favorite inflation index could be CPI could be GDP if later I am going to focus on two main indexes in the talk one is consumer prices or the GDP if later in more general terms and the other is going to be what I call the divine coincidence index what is this index it's the index that is that eliminates the endogenous cost per shocks alright so I've given that I don't have a lot of time left let me just go very quickly over this notation but essentially what I wanted to take away is that this model this complicated model with many sectors and primary factors can be boiled down to some equations that are actually very similar to the baseline UK engine model so we're going to have a factor supply equation which holds which looks basically like the consumption leisure trade off and we're going to have a pricing equation which is kind of a generalized version of the usual pricing equation and and then the very the actual novelty in the model is this new equation which is the factor demand equation that tell us how the relative demand for primary factors depend on the aggregate output gap and on relative prices and this is going to be the key part where the key equation that models the demand distortions that happen during a bottleneck so let me skip quickly to what the cost push shocks look like in the in the equation for the GDP deflator so this is like conditional on replicating and keeping output at potential what happens to the GDP deflator based on the sectoral shocks and they show that essentially inflation in the GDP deflator is positive even if output is aggregate output is a potential when productivity declines or demand increases for sectors that are more downstream or have more flexible prices that's the first term here which essentially tells us that we can define appropriate notion of pass through of cost shocks into prices and sectors that have high pass through will tend to create more inflation when they get a negative shock okay and these sectors are going to be easy to characterize their downstream and the price one. Second term tells us that the same kind of holds for primary factors so it doesn't just hold for sectors but if we increase the demand for some type of labor or capital that is used that has a more flexible price or is used by downstream and flexurized sectors we're going to get the same inflationary effect but the first term is the interesting one it tells us that we're going to have inflation in the GDP deflator whenever demand increases of the decline that's this term for inelastically supplied primary factors so they have high inverse, fresh, high okay so that's literally what I mean by a bottleneck we are increasing demand for a type part of the economy this is going to generate inflation even if output is high potential okay so it turns out that we can create as I mentioned a divine coincidence inflation index that doesn't suffer from endogenous cost push shocks and what's the weighting scheme for this inflation index well it does three things compared to the CPI first it weights sectors according to sales shares instead of consumption shares that's what I call sidebar here it means the sales share as opposed to the consumption share that they call better value my notation second these divine coincidence index is going to discount flex price sectors so delta is the probability of price adjustment at the sector level in my notation so sectors with high delta are going to be discounted and finally the divine coincidence index also discounts inelastically supplied goods okay so if you see inflation in say the shipment sector and there is a lot more inflation there relative to other sectors you think maybe there is an increase in demand for the sectors and this is a tight sector because it's constrained by the amount of land and tracks that it has okay so once we have this divine coincidence index we can use it to basically go into the data and disentangle which component of consumer price inflation comes from aggregate demand versus these cross-sectional TFP or demand shocks and how do we do it well basically the divine coincidence index to back out the output gap and that give us an aggregate component or kind of a demand driven component and then the residual of whatever inflation is not explained by the divine coincidence index is going to be coming from a cost per shock okay so this is what they do and I take the model to the data so let's skip this one and I calibrate the model to the account for the actual linkages between primary factors and sectors so expenditure shares of different sectors and more or less elastic primary factors and then I combine this data with actual sectoral inflation data to see which sectors experience more inflation and whether what does this tell us about the drivers of inflation so here is I mean the decomposition that you get depends a bit on the assumptions but here are a few robustness checks that I did and I actually found that the message is very consistent across different assumptions so here in this graph the yellow line is the kind of demand driven aggregate component or monetary policy driven component if we want the red line is the core PC and the blue line is the CPI so here in this graph I'm assuming that capital is fixed at this sector level and we see that in 2020 basically we get no demand driven inflation if anything a bit of deflation and then we get some demand driven inflation later if we allow for some capital mobility across sectors the answer for 2020 basically doesn't change but we see that the model attributes a lot more inflation to demand and that's kind of intuitive because the whole idea of a bottleneck is that you have some inelastic factor that is like stuck in a sector and sectors cannot acquire like the shipment sector cannot acquire land or tracks from other sectors right so if instead we relax this and we say well there can be some mobility then the model just thinks that bottlenecks are less important and so it attributes more inflation to aggregate demand another important thing is that we don't have great measures of wages because we have a lot of composition effects during the early phase of the pandemic so essentially the poorest people were the first to lose their jobs so the model is like the data that we have is underestimating the wage drop during the early phase of the pandemic and I think that this would likely lead us to underestimate the demand driven deflation in the early part of the pandemic so to address this issue I'll just basically remove all the weight that the index puts on wages and this is what I find so basically if we do our best to correct for this missing wage deflation what we see is that the model now does say that in 2020 the deflation part was partially demand driven and the PC was a good approximation for that but then the first spike in inflation in 2020 was still entirely supply driven but then the model will attribute a lot of inflation to aggregate demand even later great so to conclude I provide a new theoretical framework to think about how cross-sectional shocks to TFP and relative demand can affect inflation I make a distinction between inflation driven by output deviating from potential and whatever residual driven by marginal shock and they found that and I call this residual and inflation output rate of I find that this inflation output rate of increases inflation when TFP declines or demand increases for sectors that are downstream have flexible prices or are tight in the sense that they use an elastic factors I also show how to disentangle these two components in the data so how to use relative price data to understand which component of aggregate inflation is coming from demand in which one was coming from this cross-sectional shocks and I find that for the COVID episode the early phases were mainly driven by this cross-sectional supply shocks so bottlenecks and then in later phases aggregate demand was the most important factors happy to take any question and look forward to the discussion thank you thank you Elisa the discussant is Michel Garcia from CRE 10 minutes thank you for inviting me to discuss this great paper which is actually a part of also very interesting and very exciting research agenda alright so I guess I don't need to convince people once again that inflation is back in the spotlight and the question is what drives it to be more precise who is to blame I think is the more precise question and some people blame the aggregate factors or expansionary policy some people blame sectoral stocks global value chains bottlenecks in other words you know sources that are very granular and perhaps beyond the control of aggregate policy so what this paper does it tackles the question using a novel framework was I think all the necessary ingredients to think about this you have sticky prices to think about inflation you have many sectors to think about you know granular origins of shocks you have input output linkages to think about value chains and crucially and the last two things are the real innovation of this specific paper are the multiple primary factors so many different kinds of labor or capital and heterogeneous supply of elasticity to allow for certain factors which are you know in a very inelastic supply and any attempt you know to buy more from them could create bottlenecks alright so that's the paper and what it does it develops a very powerful and elegant decomposition in fact for any aggregate inflation index you take your favorite weights for an inflation index and you can show that aggregate inflation whichever way is defined can be decomposed into some aggregate component which is ultimately linked to the output gap and some cross-sectional component okay so that's the decomposition which at least it derives from any index and what she finds is that if you apply this decomposition to the COVID crisis in the US you see that the deflation and inflation in the early days of 2020 were almost entirely driven by cross-sectional factors whereas later in 21 the increase in CPI was driven by aggregate factors so that's the result of the decomposition in the paper so let me briefly walk you through the theoretical contributions of the paper starting from a very simple example so I think this equation has been in every single discussion of this conference so let me do it again so standard three equation New Keynesian model one sector the New Keynesian Phillips curve okay we can rearrange terms and say that aggregate inflation pi t minus the sum of endogenous and exogenous markups eta t is just the output gap so it's just rearranging terms okay and the right-hand side in this equation is entirely controlled by the central bank it's the output gap okay so this is the decomposition of the sort at least I considers just applied to the three equation model so inflation minus aggregate markup is controlled by the central bank so what Elisa does is she considers the very same decomposition just in the model with many sectors and shows that for the vector of sector specific inflations so bold pi t minus the vector of sectoral markups is again just the output gaps are scalar minus a vector of sector specific things and the crucial thing is to unpack those sector specific things so first you have again a vector of sectoral inflation expectations and desired markups that's eta and then you have a vector of sectoral price gaps and those two vectors enter the decomposition with some matrix V in front of them and absolutely crucially this matrix has zero rows which means any aggregate components in those sectoral vectors are completely nullified so any aggregate component in those sector specific things is not relevant for this decomposition beyond what is already contained in the output gap which is why they isolate specifically the sector specific components okay which ultimately means relative sectoral shocks matter even if the output gap is fully stabilized and I think this is a fantastic decomposition because it clarifies some occasionally sloppy thinking people are saying that sure sectoral things matter something happens in a given sector this something affects an aggregate object this aggregate object is not fully stabilized which is why aggregate inflation changes even if that aggregate object is entirely stabilized sector specific things matter even if you fully stabilize the aggregate output gap and that's the first message of the paper then the obvious issue is that to perform this decomposition in the data you need to measure the output gap which is not readily observed but then Alisa shows that it is possible to have a unique set of weights with which you aggregate the composition to completely eliminate the cross sectional components and she calls it the divine coincidence index so equation one on the previous page if you apply this unique sector specific divine coincidence weights you completely nullify any cross sectional influence which means you can track the output gap by tracking the divine coincidence inflation index that's the first problem solved then a second problem is that you still need to keep track of the etters those are the sectoral inflation expectations and the sectoral markup shocks and the way Alisa resolves this problem is by redefining the flexible price equilibrium so she makes all of these endogenous and exogenous markup shocks part of the flexible price equilibrium which by definition eliminates the etters and ultimately you will obtain the aggregation kind of result saying that any aggregate inflation index is a combination of aggregate forces summarized by the divine coincidence index which is completely immune to cross sectional shocks and only depends on the aggregate gap and whatever remains and then you can measure the left hand side as the CPI index or the PCE or deflator you can measure the divine coincidence index you can have the capac as some combinations of parameters from the model and the difference between the two is whatever is driven by cross sectional factors this is how you perform the decomposition and when you apply this decomposition to the US inflation experience you see that the early stage of the pandemic the blue line the aggregate CPI was kind of poorly tracked by the component of the output gaps the monetary policy component which pleases the conclusion that it was mainly driven by cross sectional factors whereas the later parts of the Covid crisis were mainly driven by aggregate components so that's the paper in a nutshell and I would like to make two comments which perhaps not so much want to change the paper but help kind of make some of the decompositions more robust and help us to think about how these sort of decompositions can be used later and perhaps apply policy practice so the first point is that as the paper actually acknowledges there are a number of channels that can contaminate this very very clean decomposition and kind of make things confusing whether or not it's driven by aggregate output gap, monetary policy or cross sectional factors so here's a list of some to me perhaps the most dangerous one is that those CHI components which matter for the decomposition include lacked sectoral prices as the only endogenous state variable actually in the model and if monetary policy has any persistent effects those lack sectoral prices could pick up persistent effects of monetary policy and that's something that I think is most dangerous in this and most obvious and something which I think we need to answer is whether or not this concern is quantitatively important and I think this could be done in two potential ways which are doable I believe the first one is a structural approach you can simulate kind of an exogenous central bank intervention in the model and trace out the response of the cross sectional component in the decompositive contractions see that in the model whether or not the lack sectoral prices would respond to monetary innovation under a realistic table rule I think it's the simplest thing but then this kind of exercise will be structural so it's very model dependent and I think there's a reduced form way of doing this namely measure the cross sectional component in the decomposition and then insert that into a local projection with some empirical measure of the monitor shock on the right hand side then you can see whether the cross sectional component of aggregate inflation identified in the data actually responds to an identified exogenous intervention to the aggregate component if this channel is not very strong that means this concern of kind of contamination and decomposition is not very long and in particular the second approach could be relatively agnostic in addressing that the second comment I would like to make concerns you know this specific role played by bottlenecks and primary factors in the decomposition and in the message of the paper more specifically I mean the decomposition which ELISA derives would actually hold in a simple multi-sectional model even without primary factors or bottlenecks I mean you don't need to go very far to see that if you go to ELISA's 2023 econometric paper which does not have primary factors or bottlenecks you have this equation for sectoral Phillips curve and if you just rearrange the terms here you would again go back to the decomposition which ELISA derives here so given that we can actually not observe either CHI's or ATIS in the decomposition the only way to distinguish between ELISA's new model and ELISA's old model is to say something about the copper parameters or to say something about the divine coincidence index so again the second difference is that when you construct the divine coincidence with bottlenecks you need to further discount sectors with inelastic factors so given that we cannot observe cross-sectional stocks actually driving inflation that's the only way to distinguish whether bottlenecks actually matter or not and a thing here as my final comment the approach one perhaps should employ is what would I get badly wrong if I ignored the extra layer of complexity for example imagine that I used the wrong divine coincidence index instead of the one derived in this paper I ignored the bottlenecks would this make my decomposition wrong in any major way if it does make my decomposition wrong in a major way this means the bottlenecks and primary factors should be part of these models or the second question which is related is you know imagine I want to fit a Phillips curve for the divine coincidence index again in a simpler model without the primary factors would it have much larger residuals or not so again would I be majorly wrong by ignoring the primary factors which leads me to my final question I want to add more layers of complexity to these models with also heterogeneity indulgenous price rigidity perhaps this divine coincidence index which you need to keep track of to approximate the output gap will also keep changing so the question is where do we stop and how majorly do we get it wrong if we ignore the extra layers of complexity and it's perhaps a bigger question for this whole literature but for now I would like to wrap up the paper and I look forward to the next iterations of this agenda Stefan Tepra from Bonk de France many thanks Elisa for the great talk I was just wondering the following it seems if I got to try that if I want to close the output gap I'm going to do it through a different combination of different goods in the sticky price economy that in the flexible price economy so that possibly I'm going to use different combinations of primary factors in the two cases so question number one is it the case that if I close the output gap I don't close the labor gap and if so would it be possible to do the same decomposition as you do focusing on the output gap instead with the labor gap and in which case would they give different responses to the question you're after yeah so great question so if by labor gap you mean both labor and all the other utilization gaps then that's identical to the output gap if you just look at literally labor and exclude capital utilization then the two gaps would be different and you would get a different I mean you would get a different Phillips curve but yeah I haven't thought how the decomposition would change but so actually one result which I did not present is that you can pick your favorite linear combination of factor specific gaps and construct the divine coincidence index for that so you could do a divine coincidence composition for the labor component as well I'm not sure how quantitatively that would be different um yeah I hope this answers the question yes and then yeah you would have also timed how to respond to the discussion great so can you still see my slides or do I have to share again please share again okay I think I should be able to can you see them now see it great so I think Michelle so thank you so much Michelle for super thoughtful and super kind discussion I think Michelle made two main comments or like three main comments so I'll try to address each of them in turn first like potential effects of monetary I mean we know that in this multi-sector models monetary policy has persistent effects on relative prices and so this would go this would be like a component of the price wedges in my notation coming from monetary policy so not exogenous to monetary policies so how large is that I actually did something similar to what Michelle was suggesting which is like this structural matter that Michelle was suggesting so basically I compute impulse responses for the model and based on those I back out what were the shocks that were driving the actual inflation data so what's at every point in time what's the monetary shocks and what's the cross-sectional shocks and the cross-sectional component of inflation and based on that I did do my decomposition kind of like based on a dynamic setting which accounts also for the effect of monetary policy on cross-sectional inflation expectations and this is what I find okay sorry I didn't realize that the figure was wrong size anyway I find that whatever component you get is basically super small so maybe you can see this figure let me share the figure directly I think I'm the goofiest presenter in the whole crowd okay anyway so in the like in the yellow line that's the baseline in the black line that's just literally the effect on inflation of the current monetary shock period by period and the two are very similar so that really suggests that the dynamic component is not really confounding the decomposition between aggregate and cross-sectional I just want to reiterate that I like to call the two components that I identify an aggregate component and the cross-sectional component because actually neither of the two is entirely driven or entirely by monetary policy both components have some elements that are endogenous and some elements that are exogenous to monetary policy this graph that I'm showing you kind of makes the case that the endogenous component of the cross-sectional piece is very small how about the aggregate piece it could be driven by monetary policy it could be driven by aggregate grid inflation those I cannot tell apart and to the second point that Michelle was making so that was about how important is the bottleneck aspect compared to what was already there in my job market paper that's a great question I don't have a figure to show the two of the compositions that I get are different if I use the model in my job market paper with only one primary factor versus the one with multiple factors and bottlenecks they are not massively different why? well there has been some correlation between capital intensity and inflation so that explains why accounting for the lower elasticity of supply of capital gives us a different composition in the early phase of COVID why are the two compositions not hugely different because essentially this divine coincidence index always end up assigning a highway to the wage component because it's very upstream and very sticky and so ultimately a lot of the composition is going to be driven by how much wages have moved and to the question like where should we stop I do think I mean I wrote this new paper because I really wanted to get to a model of bottlenecks and I realized that having these multiple primary factors is essential and you could think of multiple primary factors also as kind of decreasing returns that's isomorphic like if you have a sector with decreasing returns you can always think of there being a sector specific fixed factor so conceptually the two are similar so I think if you want to model a tight sector of bottlenecks you really need these multiple primary factors so I guess the ingredients I put in the model kind of reflect what I thought was the salient feature of the word that I wanted to capture but I don't see like actually I have some results that is fairly general which is that ultimately what matters in creating these cross-sectional effects are two things which is the supply elasticity and the degree of network adjusted wage rigidity so all the other kinds of heterogeneity that you can throw in the model are actually going to be second order so that's kind of a sharper answer like you can add more heterogeneity but at least in the setup that I am considering it's not going to have a distorted effect great thank you very much just give a hand thanks a lot so this concludes the first day of the conference so thanks a lot for attending tomorrow we will start at 10.30 and those who are invited to the dinner we can meet downstairs at 6.45 to walk over to the restaurant together or you can go directly to Steigenberg we will start at 7.