Aug 04, 2022

Episode 212: Prof. Ralph Koijen: Demand System Asset Pricing & Inelastic Markets

If you’re ready for a serious education on market elasticity, demand system pricing, and stock market flows, you’ve come to the right place (disclaimer: don’t expect light entertainment). Today’s guest is Ralph Koijen, AQR Capital Management Professor of Finance and Fama Faculty Fellow at the University of Chicago, Booth School of Business. Tune in for a fascinating conversation about some of the most fundamental characteristics of our economy. To say we learned a lot from this conversation is an understatement, and we’re sure you’ll walk away with just as many lightbulb moments and impactful lessons as we did.

 

Key Points From This Episode:

  • Ralph provides an in-depth explanation of demand system pricing. [0:02:48]
  • An example of how valuations can be affected while the connection between fundamentals and valuations remain relatively unaffected. [0:08:18]
  • How Ralph’s model for demand system asset pricing differs from other models. [0:41:26]
  • The two components that investor demand is made up of. [0:14:54]
  • Exploring the concept of latent demand and how to estimate it. [0:17:57]
  • How the price impact from institutions and elasticity of markets has changed over time. [0:20:34]
  • Understanding the surprising impact of households on stock market volatility in 2008. [0:20:34]
  • How latent demand can be used to predict differences in expected returns. [0:25:46]
  • Examples of factors that drive latent demand. [0:30:42]
  • The most impactful group of investors (and why this is the case). [0:33:17]
  • An overview of what would likely happen if the most influential investors switched to market cap indexing. [0:35:22]
  • How huge firms influence the setting of prices. [0:36:25]
  • Ralph shares his thoughts on the idea that index funds are distorting market prices as they continue to grow in magnitude. [0:35:22]
  • What demand system pricing tells us about the effect of socially responsible investing on prices. [0:43:01]
  • How US asset prices would be affected if foreign demand for US assets decreases. [0:35:22]
  • Inelastic versus elastic markets. [0:47:23]
  • Why prices are so much more volatile than fundamentals. [0:51:11]
  • Comparing micro-elasticity and macro-elasticity. [0:52:18]
  • Ways to estimate micro-elasticity and macro-elasticity, and the limitations of these approaches. [0:54:00]
  • Ralph’s estimate of what the macro-elasticity is. [01:01:00]
  • Risk factors that impact elasticity. [01:02:07]
  • An example which shows how flows work. [01:03:32]
  • Factors that impact how long the price impact of flows lasts. [01:05:24]
  • Dividend irrelevance in inelastic markets. [01:10:30]
  • The role of the increasing market share of cap weighted indices on market elasticity. [01:12:28]
  • How investors should behave when markets are inelastic. [01:15:11]
  • Ralph’s definition of success. [01:18:47]
Read The Episode Transcript:

Ralph, what is demand system asset pricing?

Okay, the idea behind demand system asset pricing is that what we typically do is that we’re trying to understand prices and characteristics, or firm fundamentals together. We oftentimes form factors based on characteristics, short stocks into portfolios and try to understand why certain securities have high or low expected returns.

Now, one piece of data that we typically don’t use, at least in testing models is look at portfolio holdings. The whole idea behind demand system asset pricing is to try to understand jointly the behavior of asset prices, portfolio holdings, and firm characteristics, or macro fundamentals. The reason why we think that’s relevant is that there’s a lot of questions that we are thinking about nowadays that involve changes in quantities. Let me just give you a couple of examples just to set the stage.

If you think about the whole transition from active to passive management, and then that involves large flows of capital, and wondering what does that do to prices. Similarly, if you think about all the growth in like, ESG investing again, a lot of change in flows and that impact, questions like, how much does that affect crisis? Also, in the context of QE, similar questions about Central Banks buying large amounts of assets, or quantitative tightening, not like reinvesting those assets. Then all of those questions, really what you want to have as a model, or a framework, it allows you to think about a shift in quantities, shift in holdings by one group of investors. Then how does that affect asset prices?

Then the standard model, there’s, of course, answers to those questions, but we haven’t really tested those dimensions of those models. The whole goal of demand system asset pricing is to just bring together, or develop models that simultaneously explain prices, characteristics and portfolio holdings.

What would the hypothesis beyond the shift from active to passive?

Typically, taking a little bit of a step back. I think, the overarching theme is that what standard theories implies that models are what we refer to as being very elastic. Meaning that if you get a shift in demand from one group of investors, then other investors are very quick to step in. As a result, prices don’t move a whole lot. What the data seems to suggest is that demand is a lot less elastic than what is implied by our standard models.

Now to make that a little bit more concrete, perhaps, and I will get excerpts of active-passive question. Where it is may be so very familiar to all of us? If you solved portfolio problems, think about mean variance, or you’re trying to fit a timing model, then without imposing additional constraints, you get very extreme portfolios. Because very small changes in expected returns lead to very big shifts in your portfolio weights. That feature is also part of asset pricing models. As a result, if you get a good small shifts in, or shocks to demands, like other investors would disappear, almost no impact on prices. You need a very large shift to really have a meaningful impact on valuations.

If markets are instead inelastic, then smaller movements in flows, or demand stocks have a larger impact on prices. Then you can apply that to specific questions. One of them being the transition from active to passive management. In that context, one of the things that we have done, which is joint work with Moto Yogo from Princeton, and Robert Richmond from NYU Stern, we looked at, first of all, if you look at a measure of how active the industry is all of institutional capital, one way of thinking about that is as a simple metric is to look at the aggregate, like active share.

That one declined a lot over the last 40 years. You can ask the question, did the decline in active share come about, because institutions changed our investment strategies? Or did capital move from more active institutions to more passive institutions? We find that at least if you look at the last 10 to 15 years, most of the decline in the aggregate active share is because capital moved from more active funds to more passive funds. The contribution of institutions changing their own active share is pretty modest.

Now you can ask the question like, okay, suppose we would undo that, so suppose you would use these kinds of frameworks once you’ve estimated them, to ask a question like, okay, what happens if you move capital back to those institutions where it came from, let’s say, a decade ago? What you find is that there’s a non-trivial impact on valuations. Some stocks were held by active institutions. They experienced outflows. Those stock prices came down and vice versa.

Now, then a related question that often comes up is, “Well, what did it do to things like price informativeness, market efficiency and things like that?” There, we find a very small effect. There’s a non-trivial effect on valuations, but it’s a fairly small effect on price informativeness. If you look at how well valuations forecast future profits of companies, that didn’t change a whole lot. You can ask them, why that is the case? That results critically depends on who lost capital and where did it go to?

If the capital moves from investors who are very good at forecasting future fundamentals, then markets will become more and more inefficient, if you wish. The link between valuations and future fundamentals would weaken. Instead, what we find is that the correlation between flows from one institution to another and how informed these individual institutions are about future fundamentals is very weak. It’s not the case that money systematically moved from more informed to less informed institutions to the other type. It’s more of a mixed result in the sense that impact on valuations, but in terms of the connection between evaluations and future fundamentals, there’s less of a connection there.

That’s making my brain hurt. How can valuations be affected, but the connection between fundamentals and valuations be relatively unaffected?

Right. The question is like, suppose you have two groups of investors, where one group of investors has very good information about future fundamentals. There’s another group of investors who are more, let’s say, sentiment driven. They are the ones disconnecting prices from fundamentals, like typical noise traders. Now, if the money moves from the noise traders to the informed investor, or vice versa, then prices become more informed. If instead, money moves from more “randomly,” so it’s not a strong pattern in terms of systematically money moving from one group to the other, but just capital gets reallocated more from active institutions to more passive institutions, then that could be an impact on valuations, that is not necessarily correlated with the ability for prices to forecast future fundamentals.

Wow. Okay. That makes more sense. I want to come back to just the concept of demand system asset pricing. A lot of our listeners are familiar with surprising models, like the CAPM, or the Fama-French five-factor model. How would you differentiate the model that you developed in your paper, demand system approach to asset pricing to those models?

Great question. Yes, yes. At a basic level, all asset pricing models start from modeling demand of investors, setting demand equal to supply and outcome asset prices. In the basic CAPM. Let’s say investors have mean variance preferences, you saw for optimal portfolio, supply equals demand and outcome prices. In that sense, demand system asset pricing is not a new theory, per se. It is just testing lag models along a new dimension. You could have always gone to the data and said like, well, what does the CAPM tell us and tell us the optimal portfolio that investors should hold? What do they actually hold?

What demand system asset pricing does is to say, okay, let’s bring in this additional information from holdings and see where it lines up with what our theories tell us. Let’s say that we go with the five-factor model, then what that would tell us is that investors build their portfolio based on those correct procedures are part of that model. Now, what do we find is that if you started looking at the holdings data, and we’ve mostly looked at certain filing, so that’s at the institutional level. It would be Vanguard as a whole and not Vanguard small cap growth fund.

Then there’s two very surprising facts. The first is that the median institution only holds 70 stocks. There’s thousands of stocks you can trade, but the median institution holds just 70 of them. Then you can say, “Well, that’s fine. It’s still a very diversified portfolio, potentially.” The second thing that’s surprising is that these portfolios are not particularly diversified. If you’re aggressive portfolio holdings on those characteristics, you get a very low R squared. Investors do a lot of things that are not well explained by those characteristics.

Now, then the other thing, which is a challenge, if you start thinking about those theories, is that okay, investors do like very different things. Then you would imagine if demand is very elastic, if correct risk change. Suppose profitability of a firm changes, or asset growth changes, then you would imagine they start changing their portfolios around very aggressively. That effect is also very muted. Think of it as like, institutions deviate from the market by a lot. Then those deviations are very stable over time.

Even though prices move around, characteristics move around. That is not something that you would expect if you’re constantly optimizing based on those characteristics. To go back to your question of how does it connect to traditional models, the core idea is really not so much about the models, as much as we’re going to test the same models along a different intervention. Then the key outcome of that is that demand appears to be much more inelastic. Portfolios are much less responsive to the information that we thought they would be responsive to, like the prices or characteristics. Then that, of course, leads you to think, well, what are then the right theories that would explain that? What are the further implications of those theories?

What new information do we get from this approach to asset pricing?

I would say, one piece of information and new answers. You can ask new questions. Let me start there, then I’ll get to your question about what information you get. The new answers is that you can ask those kinds of quantity questions that we started with, like, suppose I would ask you in the five-factor model, there’s been a shift from active to passive, or a growth in ESG funds, or more foreign investors, how did that affect the model? It’s hard to work out those questions. Here, you can ask those questions. That’s, I think one upside is that there’s a lot of questions involving quantities, and that’s why it’s useful to develop this. Now, what new information do you get is that think of it as a more disaggregated view of markets. That I think is the real upside of this approach is that normally, we see prices move around and we don’t really know why. Then, we tell lots of stories as to why price went up or down.

Now, what do we do over here, if we cannot go back to the level of demand? Again, this is implied by any asset pricing model. Now, that means that any move in prices, you can trace it back to individual investors. If GameStop goes up by a factor of 10, then you can trace back in that case, how important are retail investors? How important are various institutional investors? Now, what that allows you to do is that you can start to think about, first of all, why do prices move? Which investors trade? How much do they affect prices?

Secondly, you can start to think about a lot of the questions that we normally think about in terms of, let’s say, return predictability, at a much more disaggregated level. To give you a concrete example, one of the things that we normally do is we take characteristics, and we forecast returns. That’s the vast, vast majority of empirical asset pricing. What you can do over here is you can start to forecast demand, and then add up demand across all the investors, then set it equal to supply and outcome prices. Instead of directly forecasting prices, if one group of investors does respond to a particular signal, let’s say, momentum, but other investors do not, and those investors hold different stocks, then it may be the case that momentum works for one subset of stocks, but not for others.

It opens up a whole new area of thinking about return predictability from the bottom up. Again, this is consistent with all of your standard models. It’s just based on leveraging the holdings data, you can start to do this bottom-up approach.

What role, or do they still play a role, characteristics like market equity, book equity, profitability, like the Fama-French type characteristics?

They play a role, obviously. They explain a certain fraction of returns. What we find though, that if you start from holdings, and maybe the easiest way to think about it is that – your investor’s demand is made up out of two components. One is that if my view of future cash flows and things like that doesn’t change, but as the prices move around, then how quickly do I trade out of a particular stock? That part I’m going to refer to is like a price elastic component. Then there’s some additional information that I may receive about the firm’s future fundamentals, riskiness and stuff like that, and that leads me to trade.

Now, that second part, think of that as a shock to my demand. In that part, if you see how important characteristics are, and you added up the prices, it’s really just around 10% of the variation, 10% to 20% of variation in returns that you can explain with that.

Wow.

The residual part, the part of that demand component that you cannot explain with characteristics is driving most of the variation. That’s about 80%. Now, the good news is that tells us a little bit about what we don’t know about returns. This is a new way of disaggregating that information, because normally, if you have gross returns in the cross-section on changes in characteristics, you get a very low R squared, that’s 10%, 20%. That residuals then can like, who knows what’s going on?

In this case, you could ask, “Okay, how important are different investors to explain that residual variation?” Then if you want to think about predictability and things like that, then it’s really about predicting that residual component, more so than the components related to characteristics. Now that said, as you go back to Cameron’s question about what new information do you get out of this, if you could understand a little bit better also why certain factors, or why certain characteristics, like how the pricing of certain characteristics changes. For instance, if you think about value not performing particularly well over the last decade or so, and maybe minus the last couple of months, then you could ask which investors change their demand for particular characteristics.

You could imagine that there’s a particular appetite for growth of the firms that that group of investors changes over time, and that’s ultimately why value underperforms. Now you can trace it back to different types of investors. That may help you to understand also going forward what the performance may be of different factors. Characteristics are not as obviously, like we want to get rid of characteristics altogether, but there’s a lot of movement in prices and a lot of variation in demand that is not well captured by characteristics.

Do you know the answer to that question about value? About which investors are responsible for value underperforming?

No, no. That one, we haven’t done. It’s a very, again, the moment you have – the simplest way maybe to think about it is that once you can decompose the movement in any stock into the demand of each and every investor, then you can always re-aggregate it by momentum, by value. You can go through all of your favorite anomalies, or factors, and understand how important are, let’s say, retail investors for momentum. How important are foreign investors for carry trade, stuff like that, or particular countries for carry trades and things like that. You really get a new perspective on what drives markets. One, give you an interpretation. Two, I don’t know for other things like understanding riskiness, like expected returns, it gives you another look at that.

How do you explain what latent demand is?

Yeah. That’s really that residual component that I was mentioning. Things like, I choose my portfolio. Part of my portfolio choices, I see where the prices are higher, or low given everything else. That’s my demand sensitive component, then I get my demand shift, or my demand shock. That part is in part explained by correct, or six doesn’t explain that much variation. Then there’s this big residual component. That residual component, that is what we refer to as latent demand. If we add it up across investors, then that explains about 80% of variation in returns.

One of the things that you could have thought about and what could have been the case is that you open up the holdings data. It could have been the case that sure, investors hold 70 stocks. Sure, if you run regressions of holdings on those characteristics, you get a little R squared. If you can take those error turns, and you add them up across investors, it doesn’t amount to a whole lot and it just cancels out. What we find is that it doesn’t cancel out and this is really the main driver of return variation.

Now, in part, it’s interesting, because it tells us that there’s a lot of things we don’t yet understand and a lot of return variation that we cannot connect to observable factors. That’s really what we’re now interested in, you have to understand.

How do you estimate latent demand?

Oh, so the way it works, I’m going to give you the simplest versions first, and then we’re going to tell you the obstacle, why that simple version doesn’t really work. The simple intuition is going to be the following. Essentially, you run a regression of portfolio weights on prices, characteristics, and then the error term that’s late in the net, okay. Now, there’s a problem with running a regression, because if lots of investors like a particular stock, that is not well captured by characteristics, let’s say GameStop. A lot of retail investors like GameStop, well, their latent demand is going to impact prices. If latent demand is very high, it’s going to push up prices. Well, now you have a correlation between prices and latent demand, so you can’t really run that regression.

The challenge in estimating these demand models is to get the component that is sensitive to price, like right. If prices move around, how much do I respond to that? Once you have that part, then latent demand is really just a residual from that model. Now, in terms of economically, what you want to think about in terms of latent demand, it’s really everything that we cannot observe with correct risk. Because obviously, there’s a lot out there. Let’s say Netflix, I don’t know, changes the number of subscribers and things like that. There’s lots of information that we don’t have in Compustat, obviously shifts investor’s demand. All of that ends up in latent demand. Sentiment trading, like GameStop again, is a good example, perhaps of that. Yeah, so that’s how we estimate it and how to think about what could be driving that.

Man. So many questions. Empirically, how has the price impact from institutions changed over time?

Our estimates imply that the price impact has gone down, so markets have become more elastic over time. That’s going to position with other estimates that people have provided. You see, the elasticity has two features. One is the elasticity is going up, so price impact is going down. The second thing that you see quite strongly in the data is that the price impact goes up in recessions. In that time, demand becomes more inelastic and price impact goes up.

Why do you think markets would have become more elastic over time?

It could be related to lots of things. One could be simply trading technology may have improved, or maybe easier to collect information, like investment become more agile to respond to information, aggregated information and may respond to the bid more aggressively. Now, that’s not to say that elasticity is just still very, very low. Just orders of magnitude. The thing that’s striking maybe, just orders of magnitude, normally what we – in standard models, price impact. If you trade 1% of a stock, then the price impact will be 1 over 5,000.

Essentially, that means that if you buy 10% of a stock, prices would move by essentially nothing. Okay, now, few of us would probably think that if you buy 10% of a stock, prices wouldn’t move at all, but that’s really what the standard model implies. What the empirical literature tells you is that the price impact single stock is order of magnitude, one. You buy 1% of a stock, you move prices by 1%. There is a surprising agreement actually among studies, using different data, different countries, different methods. The wedge between theory and empirical is a factor of 5,000. It’s massive.

Wow. Is it households or institutions that explain more of the variance in stock market volatility?

Okay, so we looked at this in 2008. In particular, in the context of there was some debate about our large asset managers, systemic or not. I’m going to make a distinction and what I’m saying between large and smaller institutions. Suppose you do the following exercise in 2008. You think the top 30 institutions, they manage around 6 trillion dollars. Take all other institutions, they also manage around 6 trillion. Take households, they also manage 6 trillion. Now, I have three groups of exactly the same size.

I’m going to ask using our calculations of how much did each group contribute to the cross-sectional variation return? Not the whole market downturn, but just in the cross-section. Then you find that large institutions contribute very little. They just approximately run tension and variation in returns is due to the large institutions. Then there’s a larger fraction that’s coming from smaller institutions, and a particularly large fraction during that period was coming from households. It’s the upside down of what people have thought about in terms of the large institutions are the ones that are causing a lot of the price movements.

The reason we don’t get that, and it’s a very direct fact in the data is that large institutions, if you look at their portfolios, they’re fairly close to holding the market portfolio. If anything, they overweigh the largest and the most liquid stocks. That means that if they experience inner outflows, they can have scaled on all the assets in proportion to their market cap. It doesn’t have much of a cross-sectional impact. Small institutions, they act like small institutions, but as a group, they’re doing the same thing. As a result, they are going to have a larger impact on prices, and that’s even larger for households.

The ranking we get is that large institutions have the smallest impact on prices, followed by smaller institutions, and then the households are the most important one in moving prices during that period across stocks.

Do you know if that generalizes outside of that period? Do households tend to have a lot of impact on volatility, or the variance of volatility?

Yeah. We haven’t done this for each and every period. The fact that large institutions will have a small impact on prices, that’s going to generalize just by the nature of the structure of their portfolios. Because for them to have a large impact on prices, what you would need to see is that the very large investors would have a large overweight on certain sectors, or certain stocks. That’s simply not what you see.

One of the things if you just measure how to do a simple calculation of active share, and link it to the size of the institution, then as the active share declines very sharply with size of the institution, and as a result, the price impact that they will have is going to be smaller.

Any hunch on what the impact would be of gamification apps, like Robinhood?

No. No. The big unknown that we have in our data, and that’s just a data restriction is we have pretty good data on institutions from 13F filing, holdings and stuff like that. The way we measure retail investors is just as a compliment. Apple is just a 100 shares, 70 are being held by institutions, 30 are being held by retail investors. Now, what fraction is coming from Robinhood traders is much harder to establish.

I see.

We always have to look into retail investors as a group. We do have new work, though, using high-net worth individuals, where we cover a little over 2 trillion dollars in assets. We actually have their portfolios as well. That will allow us to answer some of these questions. Now, how much of those are – those are probably not your Robinhood group. At least, we can remove that group and get a more refined measure of the complemental debt, I guess, in terms of retail traders. Directly knowing how important Robinhood was for GameStop and things like that is with the data we have, at least, I think, it’s hard to tell.

Can latent demand be used to predict differences in the expected returns?

Yes, so you can. What we did is, and this is joint work with Moto Yogo. What we did is we measure latent demand for each and every institution. That’s the part in your portfolio holdings for a given stock that we cannot explain with prices and characteristics. It seems that, let’s say, think of it like this, according to the model, I should hold 10% in Apple, but I actually hold 14%. In that case, my latent demand is positive. One salient feature in the data, is that if an investor likes a particular stock now, it tends to still like that stock next year, so latent demand is high, but it tends to mean revert.

What you can then do is for each and every investor, you could forecast how quickly late in demand mean reverts. Again, added up across all the investors, and that gives you how much demand is going to change for a particular stock in the next quarter, or in the next year. That’s something that we did. If you form a strategy based on that, then that generates alpha relative to standard factor models, like a three-factor model, five-factor model. Intuitively, one big advantage that you have here, and that I think is something that is very hard to get from other methods is that normally, if you try to build a strategy with characteristics, then what you do is you link characteristics to future returns using some historical period of let’s say, 10 years, 20 years, and so on.

Now, what is the main advantage of using holdings data, is that when the world changes, holdings change right away. If there’s breaks in the data, you would notice it right away. Think COVID. In the fourth quarter of 2019, Carnival Cruises and Zoom were not particularly unique in terms of latent demand. Well, that changed a lot in the first quarter of 2020. Imagine you run your family regressions using the last 10 years of data, you’re not going to detect anything. If that’s going to be very much your own judgment call, without a whole lot of measurement to see, okay, how am I going to predict returns going forward?

Well, now what can you do with the demand-based approach? You can say, “Well, the reason why Zoom went up so much, and let’s say Carnival dropped so much is because of these particular groups of investors.” What do we know about this group of investors? Well, for some of them, demand is very proficient. For some of them, demand is much more transitory and it mean reverts more quickly. It means that on a stock-by-stock basis, by re-aggregating those investors, you get an estimate of expected return that is stock specific, based on who actually owns that security and how those investors typically trade?

Even if you get very radical changes, think about inflation, one of the big obstacles people are talking about is that you can’t really use historical data to forecast the impact of inflation on strategies, because well, we haven’t seen a whole lot of inflation. What are you going to do? Now, you convince again, like with demand-based approach, you can start thinking about, well, who’s driving prices in real yields, nominal yields? Who drives breakeven inflation? What does it tell you about expected returns in bond markets going forward?

Do we have that data ex-ante? Like you said, you can build a strategy that has an off with this, but is it tradable?

We lag it a lot, actually. Normally, so certainly, our findings come out 45 days after the end of the quarter, and we lag it by six months. The way I think about our results is very much as a proof of concept. Could you potentially do something like this? In ongoing work, we’re trying to build on this in a much more systematic way and use all the new tools that people have developed in big data machine l4earning to see, can you use those methods to forecast future demand? Because if you think about it, if you thought about big data with a couple of 100 characteristics and lots of stocks, well, here, you have really big data, because now for every stock, you have lots of investors and every investor has lots of different positions.

Now, you can really start to think about the Amazon model, where, well, you bought a hat and a scarf, you live in Chicago, maybe you also want to buy gloves. In our case, well, you traded out Google and Amazon, maybe not that big on tech anymore, you may also sell some of your position in Microsoft. All the tools that people have developed in machine learning, computer science, those become directly relevant. Now in asset pricing, because we’re truly back to the level of demand, not so much for your hat and your scarf, but now for financial securities.

To your point, can you use it in practice? Obviously, another angle is with real money in terms of the standard metrics of is the data available in real-time? The answer is yes. How much more can one get out of this approach? I think that’s one of the things we’re exploring now. What we did in the first paper, where we introduced a framework. One of the questions that we got a lot is like, okay, you can estimate all these things, but how do we know that there’s any information in here? I guess, to Cameron’s point, maybe told us measurement error.

One of the ways that we went about this is, okay, if the system works as we think it works, then you should be able to forecast demand and should translate in return predictability. That part, at least as a very first pass works. Now we’re trying to scale that up using more advanced methods.

As a trading strategy, is that arbitrage? Are you running a risk premium? How would you explain that?

That’s a very difficult question. It goes back to what drives latent demand? Because most of the variation is coming from that latent demand. It could be the case that part of it is driven by risk factors that people – if you think about my COVID example from before, then it seems very possible to feel one reason why my demands shift is for Carnival Cruises and for Zoom, none of my characteristics have really changed at that point. Of course, investors have information about all these cash flows are very much exposed. Those are not, people may disagree on that. That’s all fine. Then it would be, I don’t know, perhaps, it’s risk compensation.

If you think about latent demand in the context of GameStop, there’s risk in there too, but it’s probably more sentiment driven and things like that. One of the ideas we’re working on now is to try to disentangle from what drives latent demand, what drives latent demand at different periods for different groups of investors. One of the things that, for instance, you can do with this and that relates to this active-passive comparison we had in the beginning, is you can take latent demand at the investor level, and you can ask like, how well does investor A’s latent demand forecast future fundamentals? Or does it forecast future returns? Or, maybe it mean reverts and it’s just like sentiment.

We’re doing this now sort of an investor-by-investor level, stock-by-stock, then you can re-aggregate that. Then you can say like, well, one of the reasons why maybe value underperformed in the first couple of years, let’s say the first couple years of the last decade was because of fundamentals and later on, it was driven by sentiment, things like that.

That’s crazy. Well, it’s, I guess, it’s the 90% of price variation that you’re digging into there. It just seems like there’s so much information in here that we don’t have right now.

Right. I know. The part that excites us is that that because it’s at the investor level, you can really get into a – those investors who did very good at forecasting future fundamentals, those investors seem to be early on in, let’s say, trading momentum. Their demand forecasts future price soak other investor’s demand. It could be the case that maybe some hedge funds are very good at forecasting flows that are coming from pension funds in the future. That’s why they trade in. While others drop out, because maybe they think that demand is going to go down, or they got excited about some stock, and then that’s mean reverting.

Just because you can do this investor by investor, you get a whole wealth of new information in terms of trying to understand why prices move, gained and turn it into predictability. How much of that actually is some risk? How much of it is, I don’t know. At the end of the day, we very much think that that is related to risk, in a sense that it’s not pure arbitrage, it seems. None of the numbers that we have suggest anything close to that. It does give you lots of new ways of looking at markets, hopefully.

Yeah. Sounds like it. Do you know which types of investors are most influential in setting prices?

Yeah, so that’s right. One of the things that we’ve looked at is, suppose that investors would get in our outflows, how much would different types of investors move prices? What we find there is that the most impactful group per dollar of assets that they manage are hedge funds and small active investment advisors. The intuition being is that that’s a combination of two things. One is, if you look at their active share, how much they deviate from the market, that is a measure of how much they disagree with consensus, if you want. That thing where the market is consensus, their demand is very different from consensus.

If they experience outflows and someone else has to hold those stocks that they weren’t really willing to walk before. Now, the second thing that happens is that if you look at the elasticity of different investors, then hedge funds are very price elastic. If they see something that looks out of line with their demand, they’re very agile to move in. Other investors are less agile. Now, if you pull money away from hedge funds, then other investors have to hold it, but they’re less responsive to price. Prices have to move a lot more for other investors to step in. If you combine the two things together, their demand is very different from what other investors want to hold. Plus, to convince the other investors that they have to hold these other securities, prices have to move more, their price impact per dollar that they manage is larger.

Then, further to Ben’s question, which type of investors are most influential when linking the firm characteristics to valuation ratios?

Yeah. It depends on which characteristic you look at. Again, it’s really the same types of investors, simply because they move away the most from simply holding the market. It’s really the more active and the more price elastic ones that have the largest impact, because it’s just you pushing securities with certain characteristics, or because of a latent demand. You push your securities onto the other investors, prices will have to move more to convince them to do so. Change the connection between prices and fundamentals.

What would happen if those most influential investors did switch to market cap indexing?

Yeah. It mostly would lead to, and it’s going to do two things. One is that, obviously, there’s going to be repricing of securities. The second thing is it would make the market more inelastic. Because, essentially, like the elasticity of the whole market is a valuated average of the elasticity of each and every investor. If money moves away from the very elastic investors, then whoever’s left in the market is going to be less elastic. Any flow, or any demand stroke is going to happen, is going to have a larger impact on prices. That’s also where the whole debate about active versus passive is a hard one to settle in the sense that if you want to know whether the growth in passive has made the market more elastic or not.

It really depends on where the money came from. If the money came from very sleepy buying old households who rebalance their 401k once every seven years, maybe it made the market actually more elastic, if they hold certain strategies, versus if it gave them very agile hedge funds, it’s a different story.

How influential is a huge firm like Vanguard in setting prices?

Well, if they just hold the market, let’s say for organizations, say that their direction very low. Let’s say that they’re close to holding the market. They’re not that influential in setting the cross-section of prices, but they do lower the elasticity of the market as a whole. It goes back to this argument where the overall elasticity of the market is the valuated average across all the investors.

Now, if you’re a pure index fund, the elasticity that you provide to the market is zero. Because if GameStop goes up by a factor of 10, you just hold on, a pure market weight as an investor. As a result, if you move money from the active to passive funds, or to various – it’s also makes things potentially more inelastic. In terms of how much their own flows are going to move back the cross-section of prices, the answer is going to be very little. That goes back to the conversation we had, for instance, 2008, they just scaled down all stocks in proportion to the market cap. That’s going to have a pretty even effect on prices.

Is there a relationship between elasticity and market efficiency?

I don’t think we know. That’s the honest answer. The way I think about the whole conversation about market efficiency and excess volatility and why prices move so much, it’s a combination of two things. One is flows in and out of the market and demand of different investors, then how those flows get amplified in the market. Inelastic markets, only tells you something about the second part. It only tells you that a given stock gets amplified much more than what it would be in even a very elastic financial market.

It doesn’t tell you whether the flow, or the demand of different investors, whether that one was informed and tell you something about future fundamentals, or whether there was sentiment-driven and potentially moving prices away from fundamentals. Our next paper is going to be all about what drives the flows? What is latent demand? What are the shocks? Because I think your question is a very important one, and the one that we also want to answer. Unfortunately, we only have half of the answer, we think, at least – that’s what our current agenda suggests is that market are more inelastic than what we typically thought. That means lots of shocks get amplified much more than before. What it doesn’t tell you yet is that the sharks that are hitting markets are more or less informed. That’s really where we are going next.

Very interesting. Okay, we talked about Vanguard and their influence on prices. Based on that, so maybe I’ll just reiterate what I think you said. Assume Vanguard is just cap weighted for the sake of argument. They’re not affecting prices, necessarily, but they’re making the market less elastic. Is that right?

Yeah. Someone who holds purely to market, if you would go from a situation where let’s say you have a bunch of fairly agile investors, and you move to a situation where that money goes to a calculated investor, then you would lower the elasticity of the market.

But not necessarily affect prices? Or would there still be a price effect?

Well, it depends on what the old investor did. If the old investor was overweighting stock A and underweighting stock B, then of course, there’s going to be price effects on that. Now, if you think about the new money moving into Vanguard and becoming the largest share, and otherwise, would have gone to other stocks, again, that would have an impact on prices. In terms of if you take again, like the example of 2008, or I think other times of stress, where you see large flows out of certain types of institutions, how much do you affect a cross-section of prices? The answer is not so much, because you just scale down everything roughly in proportion. You get a very equal demand stroke across different securities.

Now, where things would get very different is that suppose that Vanguard would introduce a technology fund and becomes very, very, very large. Again, of course, it’s a whole different scenario, because now suddenly, you can get shocks across different sectors.

Okay. Based on that, we often hear this idea, or get questions about the idea that index funds are distorting market prices as they continue to grow in magnitude. Is there any validity to that or reason it would be true?

Yeah. It’s not very obvious to me. In terms of an impact on elasticity, I think there is some evidence that they may have. In terms of directly distorting prices and making markets less efficient or something like that, from what I know, I think there’s at least from our calculations, that’s not something that we see. If you look at, going back to what we talked about before, if you look at where the money came from that went to passive investors, and see whether it came from investors who are systematically more or less informed about future fundamentals, that correlation is essentially zero.

It’s not that systematically, it went similarly from informed and uninformed investors to more passive institutions. Given the absence of that correlation, at this point, I don’t think there’s much evidence that I’m aware of at least. Maybe there’s other evidence I don’t know. Based on that calculation, it’s not that obvious to me.

Okay. I’ve got one more question on just the growth of indexing concept. If the growth of cap weighted index funds makes markets less elastic, what is the implication of that for investors?

Yeah. The implication could be that if you get large shocks to markets, if you get larger flows to markets that have a could have an outsized impact on prices. The same size shock is going to have a larger impact on prices. The interesting thing, let me just give you one other example, which is interesting in this context of inelastic markets. There’s a recent paper by Jonathan Parker, Antoinette Schoar and co-authors, where they look at the growth of targeted funds.

Now, targeted funds, they add a very, very small amount of elasticity to the market, because let’s say, there’s 60/40 stocks and bonds. If the market rallies, they sell a little bit of equity. If prices go up, they lean a little bit against that, but it’s very, very minimal. Whatever it will suggest is that the growth of targeted funds made markets more elastic. What it means that the market was even more inelastic before, if those very inelastic investors make the market more elastic. That’s why it’s very tricky to start, I don’t know, not tricky, but when as we’re careful making statements about, let’s say, targeted funds make markets more inelastic, because well, if what households were doing before, is to lock up their money, be buy and hold and do absolutely nothing, but actually, maybe they made markets more elastic.

I think, the proper accounting is to say, okay, where does the money come from? What was their elasticity? Where does it go to? Of course, if everything happens in that space of very inelastic investor, the bottom line is that the whole market is quite inelastic.

Amazing. Do we know which individual firm is most influential in setting prices?

Oh, we could compute it. We didn’t do it. Per dollar of AOM, who would have the largest impact? We didn’t do that. We could, but we haven’t done that. Yeah, it goes back to the point like, once you have everything at the stock investor level, lots of calculations one can do, this would answer questions like that.

Cool. What does demand system pricing tell us about the effect of socially responsible investing on prices?

There’s a couple of ways to approach that. I think, first, let’s think about what additional information it can provide. There’s a lot of debate about, for instance, different ESG scores that are out there, that are not very highly correlated. We don’t really know which scores investors pay attention to. Now, that’s a question you can ask with managed system asset pricing, because what you could do is to say like, well, there’s different scores, there’s scores from the MSCI, just analytics, and so on. You can see which scores correlate with demand of different investors.

Now you could say, okay, this fraction of total market cap is attending those scores and distractions attending to other scores. That part, you can easily do. You can also see how much of that appetite for particular scores has changed over time. Then you could use the model to sort out how is it growth into attention to green stocks, how much it affected valuations. That’s something that conceptually can be done with once you have multiple scores and figure that out. What the model would suggest, or what the framework would suggest is that the impact on valuations would not be zero.

If you had an extremely elastic market, then what would happen is if someone really likes green stocks, you will give a lot of the green stocks. If you don’t care about green stocks at all, you would give your green stocks are very happily to that green investor, and there’s a very modest impact on prices. An important flip side is that you would see very large flows in financial markets. That’s typically not something that you see. If flows are modest and demand shifts a lot, then it has to go with prices. In terms of do we know how much of the movements in prices is because of the additional demand for ESG and things like that? That’s not something that we know at this point. Conceptually, it can easily do that.

What effect would we expect on US asset prices if a change in the US dollars reserve status affected foreign demand for US assets?

Okay, great. Another paper we’ve done with Moto Yogo, we look at a model of demand, but of the whole world. We think about it like, think as a country now as an investor, and you choose between short-term bonds, long-term bonds and equities across all countries all over the world. Now, one fact that’s very striking is that suppose that you look at UK investors. You ask, well, what is your demand for different countries? Given their characteristics, given their evaluations, how much do you invest in German equities? How much do you invest in Italian government bonds and things like that?

Estimate the model for the whole world minus the US. Now, predict what do investors want to hold of you as assets? Okay, so you get some prediction, and you look at actual holdings, and as a massive gap. Foreign investors have this very strong outsized demand for US assets. What it means is that suppose that for whatever reason, that special demand for US assets wouldn’t be there anymore, because maybe investors get concerned about certain aspects, or risk and things like that.

Then, you can compute the impact on prices. One of the things you will see is that the demand for US government bonds would go down a lot. In terms of order of magnitude, the impact on prices would be one and a half to 2% in terms of yield. Treasury yields will jump up by one and a half to 2%. There’s actually a couple of recent papers that find a similar order of magnitude, coming from the demand of foreign investment. Now, that one is relevant for a bunch of questions that we’re thinking about now. Imagine, if you’re interested in the impact of demand for government bonds and broader implications, think about a fragmentation tool that the ECB is now considering. It’s really asking a question, how much of Italian government debt would ECB have to buy to keep yields below a certain level?

Or, questions about fiscal capacity. How much government debt can the US issue before yields really start to rise? Those questions you all want to consider in a global context, because if you don’t want to buy US bonds, then that capital is going to go somewhere else. Or, investors get worried about Europe, then it’s going to have a spillover effect on US demand. We’ve estimated at least during the last 15 years, that there’s very strong demand for US assets, has a large impact on valuations. You also can use that framework to start thinking about to what extent that privilege that US experienced is still going to be there in the future.

It’s crazy stuff. Okay, I want to move to your more recent working paper on inelastic markets. We’ve touched on inelastic markets. There may be some overlap in these next questions. I think for our listeners, a lot of this has been new information. It was for us, too. I think the overlap is actually okay in this case. All right, so you already touched on this question, but I’m going to go again. Can you re-explain what it means for the market to be inelastic?

Yes. Let me also clarify perhaps a little bit what the distinction is between what we’ve just talked about and the work with Xavier Gabaix from Harvard on the Inelastic Markets Hypothesis. The first work we talked about with Moto Yogo and Rob Richmond is on the cross-section of US equities, largely. What we work on in the paper with Xavier is on the aggregate stock market. It’s really motivated by – it’s very large swings in prices. We have very little understanding. There’s lots of theories why prices move up and down. There’s one feature that rational and behavioral models have in common when it comes to the ag and stock market, is that markets are very elastic.

Now, order of magnitude is striking. The elasticity that you get out of standard theory is order of magnitude of 10, or 20, okay. That means that if you buy 20% of the stock market, prices would move by just 1%. The Norwegian sovereign wealth fund would sell the holdings that they have. It would be just a blip in the market. You wouldn’t notice it. Before lunch, it would be gone. It seems quite striking. We explored the idea that, well, maybe markets are actually quite inelastic meaning, that small effects on small flows into the market, or demand shifts can have a larger effect, okay.

Now, why did we get here? For the aggregate stock market. There’s two ways of getting into this. One is that you ask, well, who owns the market, and who can make this market so elastic? Suppose in March of 2020, the stock markets fall down 30%. More of those agile investors have rushed in to buy the market down, if you think that market is mispriced. What we do is we can first go investor group by by investor group.

You look at mutual funds, there’s three groups of mutual funds. You have pure equity funds, you have some targeted funds, 60/40, 70/30 funds. They have bond funds in this very small supply group of balanced funds who can be a bit more agile. Even there, not a whole lot of movement. Think of it as like, it’s either fully equities, or it’s a balanced fund. Those can provide very little elasticity. If you’re an all-equity fund and a market goes up or down, you can buy individual securities, but the market as a whole, you can’t really do anything about it.

Next one, ETFs. Even more experienced. Pretty much all equities, or just spots. Now go to pension funds. Well, what they do is they have their target allocation of let’s say, 60/40. If markets fall and then gradually fade back to target. Then you keep going and keep going. Then at the end of the day, there’s not a whole lot of capital left, that is also agile. We then ran a survey amongst academics and industry experts and asked like, okay, suppose that there’s a 1-billion-dollar inflow into the market, how much does it move the price?

Before the release of our paper, the overwhelming response was zero, we are gradually moving up to some response. We also asked them like, okay, what is the mechanism? Who’s providing this elasticity? Then the answer was it’s either hedge funds, or broker dealers. The good thing is those, I think, we can rule out in the sense that if you look at what happens to hedge funds during downturns, I think 2008, well, hedge funds experienced outflows themselves, or risk constraints are binding. They’re not the ones that come in very aggressively.

If you look at broker dealers, they’re just really, really small. Broker dealers just own 1% of the market, less than that. If you get a pretty large inflow into the market, they wouldn’t be able to absorb that. The mental picture that emerges from that is that there’s lots of investors who hold either fully equity, or some fixed share allocation. There’s very few investors that can actually very aggressively time the market. With that in mind, if you add all that up, then you get to a situation where markets are potentially quite inelastic.

Why is understanding market elasticity important?

Because we want to understand why markets are so volatile. We just see on days without much news that stock prices can move by 1%, 2%. We see that during COVID, stock prices fall 30%. Of course, there’s fundamental reasons. It’s not that there’s, sometimes there’s news, but the amount of volatility in markets is much higher than the volatility in fundamentals.

One of the things that people have thought about for a long time, going back to Schiller’s seminal work is seeing, why are prices so much more volatile than fundamentals? To us, there’s two parts to that question. One is like, okay, what drives those flows into markets? Second, how do those flows get amplified? What we’re suggesting is that much smaller flows into markets can have a much larger effect.

Next question is, what drives those flows into markets? Then you can get into questions about advice, anything that’s rational or not, or things like that. Depending on maybe which group it comes from, during what episodes it happens and things like that. Ultimately, I think it’s really just to understand why markets move, why are markets so volatile, and how to interpret market movements.

I just want to come back to make sure something is clear for me and for people listening. Before when we were talking about asset demand systems and we were talking about inelasticity in that respect, we were talking more about cross-sectional inelasticity. Now we’re talking about the fluctuations in the whole entire market.

Yeah, that’s right. That’s a great observation. We refer to the former as the micro-elasticity is like Apple versus Google. Then at the stock market level is the macro-elasticity. Now also, in the theories, there’s a big distinction between that. Because why is the elasticity so high in the first place at the individual stock level, like the 5,000 I mentioned before. If you think about the standard CAPM, all that matters is your stock’s data and a little bit of idiosyncratic risk, which diversifies away.

Apple and Google are very close substitutes. All you have to tell me is the beta and some ideas to credit risk and off you go. At the level of aggregate stock markets, the substituted would be, let’s say, bonds, or maybe some corporate bonds. Those are less similar than Apple and Google. The idea is that the more similar two securities are, the higher the cross-elasticity between the two. The extreme example would be, take a 10-year government bond and a nine-year government bond with 11 months to maturity. Well, those are essentially the same maturity.

If I would tell you that it would be really big price fluctuations, you would say like, well, that is too good a deal, and it will be massive arbitrage there. Now, the arbitrage between stocks and bonds is a lot less obvious. Because you can see big swings in bond yields, you can see big swings in stock markets that can last for long periods of time. The closer the substitutes are, the higher the elasticity. That’s why in all the CAPM and so on, so the elasticity of individual securities is 5,000. The elasticity of the market as a whole is 20. Empirically, both of those are a lot lower. You’re absolutely right to point out the difference between the two types of elasticity, which is very important.

Okay. I’m glad I asked that. On market elasticity, how do you estimate it?

Okay. That one is a bit more involved. In general, the way how you estimate elasticity is the following. I’m going to start with the micro-elasticity and then we’re going to go through the macro, because the micro-elasticity is easier to explain. The way it works is, in general, and this is just the only way to estimate these things, is that there’s a demand shock to one group of investors. That one’s going to move prices. Let’s take the traditional example is index inclusion.

As a thought experiment with one stock gets randomly added to the index, okay. There’s a group of index investors who now have to buy that security, okay. You get a demand shock from the index investors. Now, someone else has to give up their shares. Now we can measure like, okay, there’s 5% demand for that particular security from index investors. How much do prices need to go up for other investors to give up their shares? That’s the way we normally estimate the elasticity of the investors who are giving up their shares. That’s the general recipe.

The only thing that differs across different studies is what shock you use to what group of investors? You can think about the index inclusion being one of them. There’s a paper that looks at changes in Morningstar ratings. Then you see fixtures and flows, and that changes prices. There’s people who have looked at dividend payments of firms and you can use those as ways of moving prices around. What we’ve done in the paper with Xavier is that we have a separate methodology that we developed in a separate paper, where essentially, the idea is that you’re isolating demand shocks, using credit demand shocks to different groups of investors.

You see, you get a shock to the demand coming from retail investors, like the GameStop example we had before, or you get a demand shock to mutual fund investment, or things like that. Then you can use that to see one group of investors suddenly wants to hold more of the stock market. How much do prices need to move for all the other investors to give up some of their shares, or to accommodate the demand shock of these other investors? That’s in the background what the methodology is, in terms of estimating those elasticities.

Now, there’s another approach, which is more – at a more bottom up, if you wish, which is more of a calibration and a direct estimation, but I think it goes a long way to understand what’s going on, is that you can add up group by group, as I mentioned before, what fraction of holdings is in either all equity funds, or in balanced funds. The reason is that for an all-equity fund, the elasticity is zero. If you have, let’s say a 60/40 fund, then the elasticity is one minus the fraction they invest in equities, okay.

If you have, let’s say, a fund invests 80% in stocks, 20% in bonds, the elasticity is 0.2, that would give you a price impact, or a multiplier of one over 0.2, which is fine. What it means is that you buy 1% of the market, prices go up by 5%. You can also start going through group by group and measure what fraction of capital is elastic, and you get a sense of a very large fraction is not very responsive to prices.

What are there limitations to these approaches?

Well, the limitations are that you’re trying to isolate a shift in demand to one group of investors alone. The main threat to not measuring elasticities correctly is that you don’t isolate the Sharpe to just that one group of investors. It’s in fact correlated with the demand still of the other investors, okay. Suppose I’m trading and stock markets, I’m buying the market today. The stock market goes up by 2%. Well, it’s probably not because of me buying the market, but there’s lots of other investors who also, whose demand also changed on that day.

The main challenge is that if I don’t properly isolate a shift in demand for just that one group of investors, but instead, it’s actually, there’s a lot of – given demand, which is correlated with the demand of these other investors, okay, then you’re mismeasuring, because there’s actually, the change in prices wasn’t cost, but just a smaller group of demand. It’s actually a much broader group of demand. That’s where this whole literature is about trying to isolate those events, and trace back shifts in demand of different investors.

People have used index inclusion dividend payments, like changing Morningstar ratings, or regulations in Chile of pension funds that have changed just like IPO restrictions in China. There’s lots and lots and lots of things. People are using stimulus checks now that came to households and so on. The good news is to the big picture across 10-plus papers that we cited with Xavier, is that there is some range of elasticities, but there’s a surprising amount of agreement in terms of the elasticity at the stock level are right around one. Elasticities at the level of factors in the stock market appear to be below one.

That said, this is a whole new literature and I’m sure we’re going to learn stuff over time. To be cautious, the reason why we called our paper The Inelastic Markets Hypothesis was to say, “Okay, this seems plausible to us. Here’s the first estimate that we put on the table that seems much lower than what you would have expected. We think it’s important to measure this, but we’re very open to other methodologies, new approaches to measure this.” The papers that have come out of after us, they find the similar analysis, but I’m sure as everything in economics, this is not the last word, I’m sure.

We had Gene Fama on our podcast a while ago, and we asked him about your work. He laughed and said, “Oh, his office is right down the hall from mine.” Then he said, “Well, it’s still just a hypothesis.”

Yeah, that’s right.

Basically what you just said.

No, no. That is right. Let’s put it like this. I think at a single stock level, people have been trying to estimate that since the mid-80s. There’s a lot more evidence on that. The gap between an elasticity of 5,000, that theories imply, and the elasticity of and one that people estimate is the first thing. Now the second thing is that as we talked about before, the macro-elasticity is below the micro-elasticity. Because stocks and bonds are less close substitutes than Apple and Google. Given that I have a lot more experience, estimating micro-elasticities, and that’s an upper bound on the macro-elasticity. Well, when the micro-elasticity is right around one, then you would expect the macro-elasticity to be at a one or lower.

By that logic, there’s a lot of papers at the same time that different countries, different methodologies, that also have an elasticity that’s fairly low, there needs to be something very systematic that all these studies miss, which it’s a post-analogy.

Super interesting.

That’s where this comes from. I guess, with efficient markets, people are still going on and on about that.

Yeah, that’s true.

Nothing genius. Yeah.

That is also true. That’s still a hypothesis, too.

Yeah. We’re very open to alternative methods, measurement and so on. I think, it’s a very – it was an important question. We’ve talked in the beginning about what the upside is, if we can measure this accurately for lots of applications. We think that at the very least, we should have theories and models that explain holdings data, alongside prices and characteristics and macro-variables. I think, that’s where we’re at at this point.

You mentioned a couple of times that the macro-elasticity is below the micro-elasticity. Do you have any quantification of what the macro-elasticity is?

Yeah. Our preferred estimate is right around 0.2. That gives you a multiplier of five. You buy 1% of the market to move prices by 5%. Now, there’s uncertainty around those estimates. We get somewhere between three and eight, in terms of the range of estimates that we get. At the individual stock level, it’s around one. You buy Apple, and prices go up by 1%. That’s the order of magnitude that you get. Now, again, if you think about the idea that targeted funds, let’s just make it concrete. Using a targeted fund make the market more elastic. That means that the elasticity of targeted funds are below what it was at the market level before that.

Now, the typical targeted fund has let’s say 60% in equities, 70% in equities. That already gives you an elasticity of one minus that. 0.3, 0.4. Invert that, and you’re right around there. It’s almost an undergraduate example where you say, okay, you have two funds. One is a bond fund, one is a balanced fund. You move $1 from your bond fund to the balanced fund, how much do prices move? You suddenly get these very large multipliers. Then you realize at the market, I know seems a lot like that.

Does inelasticity extend a common risk factor, such as size and value?

Great question. Yes. The answer is yes. It really is, the more aggregated you go, the lower the elasticity. Individual stock level, very idiosyncratic. You had the highest elasticity, which are still fairly low. If you didn’t go to larger stock still in the cross-section, elasticities fall. The elasticity of Apple and Tesla is much lower than of some very small biotech firm. The reason is that they are important, like they get a large weight in many cap-weighted indices. You have to hold it. You can’t deviate all that much from those large stocks.

The demand for those securities is more inelastic. Now, you go from small cap to large cap. Now, if you go to level of factor, size and value, there’s one aspect by Jay Lee, who’s trying to estimate it for size and value factors. They get also elasticities, or multipliers, I should say, also, between two and five. You’re right. This logic applies to essentially all markets, not just equities markets, also bond markets. But the elasticity should vary and conceptually, what you want to think about is what close subsidy you have available. Think about treasuries and maybe save corporate bonds, or credit close substitutes. That’s what gives you some more elasticity there.

Okay. Now, we’re talking about flows at the aggregate market level. How can there be flows into stocks, for example, when for every buyer, there’s a seller?

Yeah. This took us a while. Yeah, it’s surprisingly tricky, to be honest. Let me give you an example. Then I’ll tell you how to properly measure this. Suppose like, you go back to my earlier undergrad example, you had a bond fund, you have a balanced fund. Balance fund holds 80% stocks and clinics and bonds. I suppose, I moved $1 from my bond fund to the balanced fund, then what you will get if you were to do – you work out the calculations, and you can see that prices go up. If you would think about what happened in terms of equity flows, there’s no equity flows, whatsoever.

Clearly, we would agree that money went into the stock market, because they move money from my bond fund to this balanced fund. How do you properly compute flows? You cannot just add up flows in just a stock market, because then for every buyer, there’s a seller. The way you do it correctly, is that you have to equity weight the flows into every institution. What it means is that the bond fund get an outflow of $1, and the balance fund gets an inflow of $1, okay. Now, the total flow across the two, obviously, is zero. If I equity weight them, then I put 100% weight on the balance fund, because it holds all of the equities, I put zero weight on the bond fund. The equity weighted flow is one.

What you can show is that this is just – it’s accounting. It just follows very directly from just market clearing that the way you should aggregate individual investors and the way you should aggregate flows is using equity weights and not asset weights. Because if you do things using asset weights, then you get into that for every buyer, there’s a seller. The ideal measure that you want is you take every investor, you look at all their investments and you measure the flow, the new capital they allocate to all of the markets, okay. Now you aggregate those across investors using the share of the equity market that each of them holds, and that’s the flow into the stock market.

How long does the price impact of flows last?

Yeah, so it’s a very important question, but one that’s even harder to answer. What we’ve focused on is on quarterly price impact. We go out up to four quarters. It’s really hard to measure this in one quarter. If you go up to four quarters, then the interval widens so much that there’s not all that much you can say after one year. It all depends on how persistent, or how transitory flows are. If you get a permanent shift in flow, so let’s say the Norwegian sovereign wealth fund decides to increase allocation to equities from 60% to 70%, and it’s going to stay there for the foreseeable future, then at least what our model implies is that prices are going to be permanently higher.

That’s important insight, I think, because it also revisit all of the work we have on to the testing, let’s say, market efficiency. Because normally, the events study graphs that we look at, the ideal graph is prices are really flat, then they jump up, and then they stay there. In our model, that doesn’t tell you anything about market efficiency. If there’s a completely uninformed flow, that’s going to stay there forever, it’s going to shift demand forever, prices are going to be permanently higher and that’s it.

If let’s say our example, from before, Vanguard introduces a new fund in healthcare, it attracts a certain amount of assets, it’s going to permanently shift those prices. Expected returns are going to be permanently lower. Now, the impact on prices is, of course, much larger than on expected returns. You can get a big impact in prices, but a very small impact on expected return, so it’s very hard to detect empirically. That’s why empirically, it’s very hard to make statements about what is the impact of flows on unexpected returns, on price over what Verizon, just because of the noise that you get at longer horizons is just very large, and that makes it very hard to get precise estimates.

Okay, hold on. You broke my brain for a second there. How can prices change and expected returns don’t?

No, no. They both change.

Okay.

Let’s say you get a permanent inflow into, let’s say, US assets. Let’s say, my Norwegian sovereign wealth fund example from before, and suppose that they buy 2% of the stock market. Prices go up by 10%. Then what’s going to happen is that the expected returns are probably going to be lower in each and every period in the future, okay. The impact on expected returns is much smaller. If prices go up by 10%, then expected returns go down by 10% times the dividend yield. Let’s say the dividend yield is 4%.

Understood.

The expected returns go down by just 40 basis points. If you go down just 40 basis points, it takes a very, very long period of time for us to be able to detect that impact. Even though we really want to know what is the impact on the long run impacts, some papers would try to measure this up to one year out for individual securities. In our theory, it critically depends on whether the flows are permanent, or whether they are transitory. The empirical literature makes some distinction, but not always. Yeah, so that is where we are with that. Quarterly, we can measure reasonably well, up to a year, we see no mean reversion but then confidence intervals get so wide that there’s not a whole lot you can say.

Do we know what causes flow?

Not yet. We’re hopefully getting there. What we’ve done is the following. Given that we have a new way of measuring flows, we try to implement that empirically and say, what is the flow into the stock market? Now the problem is that it’s actually surprisingly hard, because the data that you need for this is that for each and every investor, you need to know what is their new money that they allocate to all the asset classes. The data that we have typically is either just equities, or just for a small set of institutions, like pension funds. We have both. There are equity and fixed income and so on allocations. That’s really what we need for this.

We need to make some assumptions there and we try to measure flows into the stock market. We show that those are correlated with prices. Then we correlate it with two things. One is macroeconomic fundamentals with GDP growth and things like that. That doesn’t do a whole lot. What does work very well is investors survey-based measures of expected returns. When you see that investor’s subjective expectations of returns go up, that correlates quite strongly with our measure of flows. What is perhaps most interesting to us is we’re now trying to do in different settings to measure flows closer to theory. In other countries and in Europe, we have much better data on this, and we actually can measure flows more accurately, and then we can answer some of those questions. The biggest question to us at this point that we’re working on is trying to understand what drives flows, what are those shocks to latent demand, which is all one in the same.

Interesting. That’s what you’re talking about earlier about relating it back to market efficiency. Is that right?

Yeah, exactly. The whole idea is going to be to try to understand what drives shocks to financial markets with lots of people thought about. This approach gives you a new way to thinking about that. Then all the answers can be a lot more nuanced in terms of, is it all sentiment, or is it all an irrational policy? It could be that some investment in some periods, even in some segments of the market are more informed than in others.

Unreal. Is dividend policy still irrelevant to the valuation of shares if markets are inelastic?

Okay. Now it depends. That one could also change. Think of it as follows. If you do repurchases, versus you pay out the dividend, then now it becomes irrelevant. What the investor is going to do with that dividend? If the investor is going to keep the dividend, and not reinvest it in markets, then dividend policy will not be neutral anymore. Again, it’s very hard to get good data on that to know exactly how much of the cash being paid out is being reinvested. In principle, it could be the case that dividend irrelevance has been broken by inelastic markets.

Would that be a good thing, or a bad thing for dividend-focused investors?

That part is not so obvious to me. I think, there’s more that bigger picture – The way I would think about it more is that it actually has an impact on the firm’s policies. One of the things that now changes is that if your payout policy can have an impact on valuations then obviously, it may give firms an incentive to choose one versus the other. Then the same way, if you think about going back a little bit to the first part, if you think about investors having a special demand for certain characteristics, let’s say, that investors really like firms that do a lot of R&D, or they really like firms that do a lot of investments, or that expand globally, well, then they get firms an incentive to undertake those corporate policies.

It’s not just an implication for investors. Actually, firms now facing inelastic markets, where investors have an appetite for certain characteristics may actually change corporate policies to cater to that demand, and that may influence valuations. That’s, I think, a more direct consequence. Depending on the outcome, it will also have an impact on expected returns of firms. Hence, for investors who would like to hold dividend paying stocks, but that really depends on how large the effects are and whether it is relevant or not.

What role, if any, does the increasing market share of cap weighted indices have on market elasticity?

At the level of the aggregate stock market, then, if you just hold on to the market weights of everything, you provide no elasticity, whatsoever, if you’re just a buy and hold investor. If you think about the level of the aggregate stock market, and you think about asset classes, the more common policy is to hold light fixtures. Then you provide some elasticity, but it’s very low. That was like, you invest in extending stocks, elasticity is 0.4. To the extent that more and more investors move to buy, and it’s really buy and hold, just hold the market portfolio across many asset classes, or fixed shares, it makes it a lot easier for us to compute elasticity, because we know exactly what else is for those investors.

To what extent it makes the markets more or less elastic goes back to what did they do before? If the target investors have attractive 401k sleeping investors, then it’s going to make markets more elastic. If it came from investors, otherwise, who were much more like timing the markets. That, I think, is also at a high level, I think the deeper question is why investors don’t time the market more aggressively, right? Why do we see these fixed share allocations and why don’t we see investors time the market more aggressively?

I think lots of people have expressed this idea that okay, it’s very hard to time markets. There’s lots of uncertainty that comes with it. Faced with that uncertainty, the best thing you can do is you just hold a share portfolio. Well, there’s a flip side to that, and the flip side is that markets become very inelastic.

I’m not sure I understand the implication there.

Suppose that before, investors were trying to find new markets, and they thought when the stock market is falling a lot, then we aggressively step in and things like that, then you would provide more elasticity to the market. Now, if in reality, it’s very hard to time the market, because there’s so much uncertainty in whether prices fall because of fundamentals, or because of expected returns, then it may be optimal for lots of institutional reasons to just hit aggregate. At the level of asset classes, our strategic allocation as a pension fund is just 60/40, stocks, bonds. We revisit it every five years, and that’s about it.

Well, that policy of fixing those shares, that leads to very inelastic markets, because you can respond very much at the asset class level to movements in prices. I think lots of people agree that timing markets is very difficult. There’s a lot more cross-sectional strategies and things like that. The implication of that is that if people don’t time the market very aggressively, then we know directly what the implication is. That’s what our paper points out, is that these fixed share strategies, that’s what a lot of investors have gravitated towards, just because it’s hard to time the market, consequences that the markets put in the last.

Super interesting. Okay, last question for you. Should individual retail investors, many of whom are listening to this podcast, should they behave any differently if markets are indeed inelastic?

I think not so much. Retail and institutional investors alike. Because a fixed share allocation and not timing the market and maybe the optimal thing to do, faced with a lot of uncertainty. If you don’t know what expected returns are at any given point, the optimal thing, maybe to just hold a fixed share allocation and gradually trade back to where it started. We’re just drawing out those implications, but it doesn’t mean that you can do much better.

I think, the simple and this survey, we’ve done a lot, also with investors. If you think about, I think, last year, in 2021, when markets rally a lot, and you drifted away from your target allocation and you now think that expected returns are lower. You were 60/40, now you’re 70/30. Corporate bond returns over the year are pretty flat, like stock markets rally, so you’re out of line with your target. What do you do? So suppose that the expected return on stocks went from 5% to 2.50%, then what most people answer in that survey is like, “Oh, well. We would gradually trade back from 70/30, maybe to 65/35, 60/40.”

Very few will say like, “Oh, well. The equity risk premium when it half. Before it was 60/40, I should be 30/70 now, because the equity risk was much, much lower.” That’s what the standard model would imply. This notion that you gradually trade back to target, because it’s just probably hard to know what expected returns are at any given point, there may well be the optimal thing to do. It means that markets are inelastic, but that’s not the whole lot you can do about it. The only thing that, again, I think, to the earlier discussion that we started with is the upside of all this, but I’m not sure that’s for retail investors. The upside of all this is that it can make markets more understandable, interpretable, because we know who’s trading, who’s moving prices.

If you could forecast that demand, then that can help you to forecast future returns. There’s the upside there. That only works in inelastic market, because otherwise, the price impact of any individual investment would be very, very small. That’s some implication is there. In terms of, is there something you can do very differently, just because you now know that markets are inelastic, the answer is not so much. I’m afraid so.

Do you think we’ll see, like there are all these factor investing products based on the Fama-French factors and other factor models, will we see a, I don’t know, a latent demand factor fund at some point?

It’s worth exploring. Let’s put it like that. I think it’s worth exploring to think where you can use with – I mean, put a leg as to restrictions that you need, under which you can bypass holdings and directly go to forecasting, returns are pretty strong. In a sense that investors need to be very much alike, so that the demand of different investors are not very relevant in forecasting future returns. Now, how large those gains are in practice, accounting and certainly, you have an estimation and things like that, that I don’t know.

Got it.

One of the things we’re exploring now more for the greater good of academic interest, it’s just understand where if you use some of those tools, like we talked about before, to forecast latent demand, and then forecast future returns, how much can you gain beyond directly forecasting returns? Obviously, there’s a lot of interest in the industry on studying flows. Essentially, what this provides is maybe a framework to say like, well, I don’t know, we see them all the time in the media, like okay, this flows from this segment of the market to that segment.

We’re essentially providing a framework to put all that together and how it links to prices. Given that there’s a lot of attention on trying to use flows to forecast returns, perhaps there’s something there that one can explore.

Super interesting. We do have one final question. I didn’t send this one to you, so if you want to skip it, no worries. But we typically ask all of our guests. I just forgot to include it. How do you define success in your life?

Okay, that’s a good question. Success, is if I understand that are how financial markets work and how to interact with the macro economy. The bigger goal is putting there’s a lot of things in finance that we don’t understand, like two basic questions, why the markets go up and down? Why do certain factors earn high returns? Of course, lots of people have thought about that. We’re just trying to approach it from a different angle by really starting at the industrial level and bringing a whole lot of new data to make markets more understandable.

Then our next steps are going to be to connect that to the macro economy and global economy. That’s really where our research agenda is heading. Success would be, if we have this podcast again in 10 years that I can tell you a little bit more about how these things work, that will be successful for me.

Awesome, great answer. Ralph, this has been fantastic. I learned a ton going through your research and even more talking to you today. Really appreciate it.

Thank you so much for having me.

About The Author
Cameron Passmore
Cameron Passmore

Cameron Passmore has been a leading advocate for evidence-based, systemic investing for over 20 years in the Ottawa area. Today, Cameron and his team serve a broad range of affluent clients across Canada.

Benjamin Felix
Benjamin Felix

Benjamin is co-host of the Rational Reminder Podcast and the host of a popular YouTube series.

Meet With Us

Meet With Us Flyout Form
What are you looking for?