Financial Economics for Real People

0:00The conference has been great so far, but I have to admit that as soon as I saw the schedule, I knew it would be good because I think this is the only financial conference this year with two speakers in a row named Jean. And that's a plus at any conference. I think they should all have more speakers named Jean, so I knew we were set here. When I was preparing the talk, I talked to Lew a little bit about the time factor and managing the time talking. And Lew had said, well, you know, a big thing is your pacing. Now, you know, I didn't want to write that much, so you have to excuse me.

0:47I'll be talking like this. What I want to talk to you about today is I'll call it an Austrian sojourn through the land of modeling. It's something that's extremely prevalent in the financial industry and I joined an equity trading firm and found that they were making extremely extensive use of mathematical models in their Trading. Now, as an Austrian, this was something puzzling to me and I wound up, you know, I've spent a lot of time thinking about this the last few years, puzzling over, you know, what was going on, how did this fit with what I knew of Austrian theory. Hopefully for the people who are listening, this will give you a That's a way to view what's going on with modeling in the financial industry.

1:54You may be someone who's ambitious enough to attempt to create your own models to use in your trading, or you may invest with someone who trades off of models, who uses models and they're investing. I think in either case, hopefully the view I give you of what's going on with modeling will be of some cognitive assistance in looking at that issue. So I'll start out by telling you about my journey into the land of models. I joined a very successful equity trading firm about three years ago. Now, I'll explain very successful.

2:39When I joined, they were in their 15th straight profitable year. The three years since have all been profitable. The year before I joined, the return on partner's equity was only 20%. The year I joined, they had an 80% return on partner's equity. It wasn't because I joined. The next year, they had a 106% return. And for 2001, I think the return is 40%. More astoundingly to me, the firm hardly ever had a losing day trading. The firm posts running P&L to the internal web on a number of the areas it trades, so everyone inside the firm every 10 minutes can see our P&L as we progress through the day.

3:38and once in a while there's a down day but pretty much the firm does not lose any day and they're doing all this they're achieving these returns and and not having losing days not swinging for the fences or trying to make a killing in any particular stock they look to make three cents a share that's That's their goal. If they make three cents a day on the shares they trade, they're extremely happy. This performance caught my eye. This was not a random walk down Wall Street. As a student of economics, this looked to me like something worth studying and trying to figure out what was going on.

4:32So first I'm going to tell you a little bit about what they do so you can see how they're using models in their trading. I'll digress for a moment. A sidelined piece of advice, don't try this at home. I bring this up because I've seen a lot of ads for day trading operations and you read this this booklet and get this piece of software and you'll be a day trader just like the pros. Well, after being at this firm, I consider this something like someone who says we'll give you this helmet and shoulder pads and this pamphlet on how to play football and you too will be able to play in the NFL.

5:19You know, these people have hundreds of trader years of experience on their trading floor. They have millions of dollars of investment in technology, high-speed data feeds, teams of programmers, a math PhD generating models for them. You're not going to be able to do that at home. There are exceptions. You could be the Michael Jordan of day trading and play in the same field as them. But for the average investor, you know, don't think you can enter into this arena against people like this on short-term trading. As I said, they use quite a bit of mathematical modeling. A lot of the trading is automated and in fact they count on this for their edge in a lot of areas. The computer can spot This is not, hopefully, a price discrepancy and trade on it before a human trader would have time to react and pick up the phone and execute the trade.

6:28There is a downside to this way of doing things. One day a trader at the firm accidentally bought 102,000 shares of a stock as a result of a program bug. He meant to buy 2,000 to hedge a position and the program put out the order 51 times before someone killed it The stock happened to be Enron and the day happened to be the day before the stock collapsed from 4 to 50 cents We also on occasion we found ourselves accepting our own offer for a stock The program puts out a bid at a hundred and a few seconds later the program says, a hundred looks like a good price, I'll buy. So we have an occasion of traded with ourselves.

7:17So there is a downside to this kind of thing. I'll look at one aspect of the business. They do what's called risk arbitrage. The math PhD in the traders have spent a couple of years There's refining a model of merger deals. The trader who's going to trade the deal, input some parameters to a program, basically babysits the program through the day and allows it to trade, you know, half a dozen traders in that area are trading several million shares a day, several thousand trades. They may buy and sell the same stock 50 times in a day, coming in and out. Our exceptions are times when they intervene and we'll look at that because that's important.

8:04What they're looking for is mispricings. Let me, right under death, we'll put this. Let's say that Sun Microsystems is going to buy Apple Computer. And we've got Sun trading at 22 and Apple is at 16 the day the merger is announced. Now we'll imagine it's going to be a one-for-one stock deal, so the day the merger goes through, one share of Apple is going to exchange for one share of Sun. Now, what we know at the time of the announcement is, if the deal goes all the way to completion, the day the deal completes, they're going to arrive at the same price.

9:01We know that because the instant before the deal goes through, if they were trading at different prices, there's essentially money lying on the table for someone just to pick up. If right before the deal happens, we're this 20s arbitrary, nice round number. If sun's at 20 and apple's at 19, right before the deal goes through, you buy as much apple as you can and get $20 an instant later. So we know that these prices are going to converge to this point should the deal go through. So we've got a convergence over time. In the meantime, though, there's risk that the deal will not complete. You may have a regulatory agency stepping in. You may have something like Dynergy Enron where son looks at Apple's books and says, whoa, you've been lying to us, you know, this isn't what we said we were getting into. You may have something like HP Compact where a major shareholder starts to create a fuss that looks like it's going to go through I guess but you know major shareholder

10:04could stop the deal. So what they're doing is they're trying to factor in the risk and find a band between which these stocks should trade. Anytime, let's say here they think there should be a 10% difference in the stock price. Well, they find Sun trading at 22 and Apple at 18. They're going to sell Sun, buy Apple and and they wait for the stocks to move back within what their model says the band should be, then they reverse what they just did and they'll do that all day long. If those stocks keep doing this, they could do it 50 times a day and, you know, the volume, they like volume, they like stuff happening, so the prices keep moving.

10:56Their returns were down a little bit last year because volume was down. So I saw them doing this and you know it's very puzzling to me because I'm convinced that the Austrian critique of mathematical economics is sound and I'll read you something from Mises. He says logic and mathematics deal with an ideal system of thought. The relations and implications of their system are are coexistent and interdependent. We may as well say they are synchronous or that they're out of time. Within such a system, the notions of anteriority and consequence are metaphorical only. They do not refer to the system but to our action in grasping it. The system itself implies neither the category of time nor that of causality.

11:45There is functional correspondence between the elements but there is neither cause nor effect. What distinguishes economics from the logical system The theorem is precisely that it implies the categories both of time and of causality. So just to explain what's going on here in this statement, Pythagorean Theorem, A squared plus B squared equals C squared. There's nothing here about cause or time. B and A don't cause C to be a certain length, and C doesn't cause A and B, and it's not like B emerges and then subsequently A and C emerge, this is a timeless form.

12:37Cause and effect and before and after are absent in mathematical functions. With action it's different, we have to understand that for the human actor there is a past that's the soil of their action, there's a present where they can sow the seeds of acting and a future in which they hope to reap the fruit of that action. And a human actor must believe that they can cause and effect or they won't act. So this is Mises' basic critique, quick version of why mathematics is not the essential way to understand economics. I'll give you a quick example from the neoclassical world of the reverse view.

13:25Steve Landsberg in his micro-textbook Price Theory says it is important to distinguish causes from effects. For an individual demander or supplier, the price is taken as a given and determines the quantity demanded or supplied. For the market as a whole, the demand and supply curves determine both price and quantity simultaneously. Now if you think about what he's saying in the common little supply demand curve graph you'll see in a micro textbook, no human is able to do anything about a price. The curves set the price. So if apples are 59 cents Well, the curves determine the 59 cents. No human had to do anything to bring the price to 59 cents.

14:21It leaves completely unexplained how a price can change. Clearly, prices only change because of human action. Someone decides, I can lower the price to 58 cents and my cost of production will go down, so on. I can make a bigger profit. Or if I bid this, I can get more and I have a good use for them. So the fact that these give us a rough picture of market behavior is the effect of human action. It's certainly not the cause of human action. Mises says the mathematical economists disregard the market process and evasively amuse themselves with an auxiliary notion employed in its context and devoid of have any sense when used outside that context.

15:12Because we're dealing with human action, we're also not going to get the constants that the physical sciences will give us. Whatever you choose or decide today, tomorrow you have the opportunity to choose differently. So, you know, someone who says we've determined that the price of wheat will always trade at one-fifth of the price of, you know, barley. All that takes is market participants making a new decision, and that relationship will be gone. So Mises says there are no constants in human action. Well, so this leaves us with a puzzle. All these investment banks and trading companies and so forth invest all this time and money into modeling. Now, are they just nuts? Well, you know, in at least the example of the firm we've been talking about, they clearly aren't nuts. They're doing something, they're getting So how can we square this use of mathematical modeling with the Austrian critique of mathematics in economics?

16:27So the conclusion I was finally led to by thinking about this was that mathematical models are useful for describing equilibrium-like phases of markets. I want to give you a little analogy. You have a batter in a baseball game and he swings at a pitch. Well, you can use a mathematical model to describe the path of the bat, given the initial force he applied to the bat and what's likely to happen to the ball and so forth. What the mathematical model is not going to get you is whether he's going to decide to check his swing. so you have phases in markets where the market will behave a good deal like a model now with the art with the merger deals that our traders are trading I said that they pretty much babysit the program through the day but what happens is they they watch carefully what's going on with the the stocks in the deal and And at some point, they see something going on that isn't going according to the model.

17:42At that point, they stop the program. And immediately they try to figure out what it is that went on. And it may be there's an announcement has come out, a major player has entered into the market for those stocks, could any of a number of things. What's happened is there's a new interpretation of the data in the market, and at that point they need to rethink the situation. They're again engaged in entrepreneurial activity, and they may be able to change their parameters, re-input them, and start it running again, and it may do just fine.

18:24One thing we can say about any sort of model in the financial markets is that it makes no sense for anyone who contends they have a model that always will work that they've arrived at the universal model. This can't possibly make sense. Fisher Black is a famous person in the finance markets and he was backing the capital asset pricing model putting forward his version of it and he He actually wanted to achieve general equilibrium and thought that by everyone following his model that the economy could achieve what Mises called the evenly rotating economy. Basically everything could settle down and conform to his beautiful mathematical model of what the economy should be like.

19:16But once you look at the factor of human action, you see this doesn't make any sense. No one's profiting anymore. Why? Why are they paying attention to the market? In fact, the market will get wildly out of equilibrium with everyone following a model because there's nothing, there's no reason to pay any attention to it. You also have the paradox that if everyone's following the same model, when that model generates a buy signal, well, everyone's Who's going to buy? From who? Who's selling to them? There's no one on the other side, so the idea of a universal model or a model that's always correct is nonsense.

20:03What you have is Terry Narden, who's a political philosopher, gave me, I heard him at a conference and he said a phrase that sort of crystallized my thinking on this for me. He said mathematical models in the social sciences are disguised descriptions of a practice. What that means is you have certain institutional structures, certain common interpretations, certain traditions, practices that people are employing. And while those practices and interpretations are in place, you'll get a regularity in that area of society.

20:53And that regularity, you may be able to generate a disguised description of it in a mathematical model. But that will hold only as long as those interpretations, those practices and customs hold as well. Karen Vaughan makes a very similar point. She says, in the case of action without significant reinterpretation, we try to explain a set of choices and their consequences within an established culture and an established market, within a given set of institutions. Because there are established institutional parameters, parameters, we can make informed theoretical predictions about the outcome of any action.

21:40In the second case, where new interpretations are being formulated, where there's innovation, we are asking questions about the process of market creation and institutional change brought about by the discovery of new knowledge or the perception of previously unimagined opportunities. With changing institutional parameters, we can predict very little even in principle Since we cannot know in advance what is going to be learned or perceived. So, if you're looking at employing models as an aid to investing, what can you take out of this insight? One thing is that modeling doesn't eliminate entrepreneurship.

22:31The people who are creating the models at my company are constantly innovating with their modeling. They're constantly making entrepreneurial judgments, and they're constantly ready to stop the model and make a new entrepreneurial judgment. At any point, the market data seems to call for that. Mises says, like every acting man, the entrepreneur is always a speculator. He deals with the uncertain conditions of the future. His success or failure depends on the correctness of his anticipation of uncertain events. If he fails in his understanding of things to come, he is doomed. The only source from which an entrepreneur's profits stem is his ability to anticipate better than other people the future demand of consumers. If everybody is If it is correct in anticipating the future state of the market of a certain commodity, its price and the prices of the complementary factors of production concerned would already be adjusted to that future state.

23:37Neither profit nor loss can emerge for those embarking upon this line of business. So that leads us to another consideration of modeling. As I said before, a model that everyone's using or everyone knows about is extremely unlikely to be successful. The interpretation is already out there. The price will already have adjusted. Now, there are a couple of common exceptions brought up, the dogs of the dial, the January effect. I'm no student of those. They're puzzling to an economist because if they really work and they always produce profit, why hasn't everyone piled in and eliminated that profit? As I said, I'm no student of those particular theories of trading.

24:25Another lesson is you have to constantly watch, if you're capturing a pattern of social practice in your model, you constantly have to watch for that pattern dissipating because it will. As I said, these traders and the modelers at my company are, are, while they are still making money with model A, they're working extremely hard coming up with model B. They never stop doing that. They never conclude that we've got the model and this is it. The traders explicitly say, no, you know, the profits from that will go away after a time. Market conditions change or other people come in and pick up on the pattern, because whatever pattern you saw you're not the only one who can see that and they'll gradually take part of those profits for themselves until the profits are competed away. As I said I would say don't trust anyone including yourself who thinks they've got a universal model for trading. A universal model implies that we're going to wind down to general equilibrium, to the evenly rotating economy and that state

25:47of affairs is logically impossible. Mises says there's nothing driving it anymore, there's no human action left. And as I said, you also find that if such a model were true, then you get the ridiculous state where everyone in the market is buying and selling at the the same time. The market is driven by differences in valuation and a market where everyone values everything the same is an impossibility. Okay, that's it. Hopefully something you can take out of that. Thank you.