Tag: modeling

Macro is *Hard*, Edition #293483343643

I began teaching a course on introductory macro this past Saturday at a college here in Pune. I often tell my students that my job in a macro course is to leave them more confused at the end than they were at the start. That always evokes laugther by way of response, but as anyone who has learnt (and especially taught!) macro will tell you, I’m quite serious.

Macroeconomics is hard, it is confusing and as the person responsible for teaching it, you’re always on your toes, because you’re never sure if you’ve understood it yourself!

And I really do mean that, it is not a rhetorical statement. My own PhD is in macroeconomics (business cycles, more specifically), but I’ll happily admit to still not being sure about what exactly causes business cycles, what (if anything) to do about them and when to stop doing whatever it is that we’ve chosen to do about them. And I suspect that most macroeconomists will tell you the same thing.

This humility stems from a very good reason: macro is hard.

It is hard for lots of reasons, and not to get too meta, but quite a few debates within the field are also about which of these reasons are most relevant, and whether the relevance changes over time – and if so, due to which reasons!

But if I were to try and write a simple post for people who have no formal trianing in macro about why macro is so hard, here would be my reasons:

  1. Macro is really about trying to figure out everything that goes on in an economy, and if you try to think about all the things that go on in an economy, you very quickly realize that figuring them out is even more challenging.
  2. Time and uncertainty!
    • Macroeconomic decisions take time. It takes time to decide to start a new factory. It takes time to figure out the financing. Land acquisitions, regulatory approvals, construction delays will all add weeks to the planned schedule, if not months, and sometimes years.
    • These expensive decisions are made at the start, but there is no guarantee that macroeconomic conditions will be the same at the finish of the project as they were at the start. You want a relatable example? How sure are you that macroeconomic conditions will be the same when you graduate from college – as they were when you enrolled in it?
  3. The way macroeconomic variables interact with each other isn’t known for sure. We think we know how inflation and unemployment are related to each other, but we can’t really say for sure. We think we know how exchange rates impact the domestic economy, but we can’t really say for sure.We’re still figuring out how monetary policy and fiscal policy should interact in theoretical models, let alone in reality. The impact of monetary policy in America today on India’s economy tomorrow? Don’t get me started. I can go on, and folks with greater expertise than me will prbably not stop for years.
  4. Life has a way of throwing up surprises that macroeconomic models never thought about. You could (and probably should) blame macroeconomists for not getting enough finance into their models prior to 2008, but who, pray, could have foreseen 2020 and 2021? How do you come up with models and policies on the fly in such a scenario? And then, just for fun, throw in a jammed Suez canal. Life, I tell you.
    We call these things exogenous shocks in macroeconomics, but the name hardly matters. Reality will always be more complex and more unexpected than any model you can come up with, and that’s just a fact.
  5. Counterfactuals are impossible to test. How do we know that Ben Bernanke did the “right” thing in 2008? We don’t! What if he had done x instead of y? There’s no way to test this, since we can’t turn the clock back to 2008, and ask Mr. Bernanke to, well, do x instead of y. This is both a problem and when it comes to critiquing models, a great convenience.
  6. Attitudes towards risk, and the propensity to copy what others are doing change according to your outlook towards the macroeconomic environment. You can call this animal spirits, but what you’re really saying is that you don’t quite know how to think about it, even less model it cohesively.
  7. Building a model – any model – requires simplification. When you build a model, it will by definition be an approximation. Unfortunately (and I wish this weren’t so), this very real limitation isn’t always front and centre within the field while developing models.
  8. What are you optimizing for when you build a model? Is it fidelity to reality or is it a beautiful model that may or may not have anything to do with reality? Again, I wish this weren’t so, but the answer isn’t always clear cut.
  9. Any field that uses the pool player analogy is a field that is, by definition, unsure about how the world works.
  10. No matter how much data we have access to, there will always be data points that we cannot capture, and we don’t quite know how these data points, and their unavailability, will impact our understanding of the economy.
  11. Social structures, psychological make-up, cultural parameters will all have an impact upon the decision making capabilities of individuals, but quite how this works (and that too across space and time) isn’t well known. For example, how would your grandfather have reacted to the prospect of not being employed upon graduation? What about your dad? What about you? What does this say about the nature of India’s changing economy, and what does it say about cultural norms and expectations? Is your answer likely to be different depending upon how much your family earns, where in the country you are located and your family size? Can we model this? (Hint: no.)

So sure, I’ll teach them about the variables, the models and the case studies.

But I’ll let you in on a dirty little secret, so long as you promise not to tell anybody: I’m just not sure if I really and truly understand what I’m teaching in macro.

The TALISMAN Heuristic

Thinking about real world problems is impossibly hard. Any story you tell yourself about the world is necessarily an abstraction.

What does the world abstraction mean? The Merriam-Webster dictionary tells us that “from its roots, abstraction should mean basically “something pulled or drawn away”.” So when you tell yourself a story about the world, you are pulling something, or drawing something away from reality.

What are you drawing away from reality? The parts of reality that seem important, interlinked and informative to you. So for example, when you wonder why the prices of onions are so damn high, you try to “pull” out of our reality those parts that you think will help you explain why the prices are so damn high.

Now, you are the captain of this ship – the one that is about to undertake this intellectual voyage of discovery. You are free to taken on board any parts of reality that seem relevant to you. Unfortunately you cannot take on all of reality, because then, hey, you aren’t pulling or drawing something away from reality. You’re trying to take on all of reality! And that is difficult impossible to do.

So you might choose to take on weather patterns, inflation, and the part of the country that you live in. These might help you arrive at a way to think about this specific real world problem: what is causing the prices of onions to be so high? Could be, you think to yourself, because of unseasonal rains, could be because of high inflation and it could be because you live in a tony part of a town/city that tends to have high prices of vegetables.

Congratulations, you’ve built a mental model! Don’t worry (yet) about whether the model is correct or incorrect. Don’t worry about whether you can gather the data required to test out your model. Don’t worry about whether the model will work next year, or in another part of the country. You’ve fashioned for yourself a story, and the story goes like this: x is seen in the world because of y. Specifically, high onion prices are seen in your world because of the weather, because of inflation and because you live in (say) Pune.

Savor this moment of victory, for we’re about to add in some complications.

The first complication is that you haven’t taken into account everything that influences the prices of onions. Maybe there’s a transporters strike? Maybe there’s been a pest attack on onion crops? Maybe a restaurant in your area has purchased an unusually large quantity of onions just a little before you went out to buy onions? Maybe the vendor was in a bad mood, and is charging you high prices for no good reason?

Some of these questions make sense, others do not. My point is that once you start to think about the problem in greater detail, you might realize that there are many other things apart from your three factors, that at least have the potential to raise onion prices.

But pah, you say to yourself. By this logic there will be no end to this exercise. You have, you tell yourself, chosen the factors that are likely to explain most of the increase in prices. Sure, you say to yourself, there are other causal factors out there. But these three? These, you aver to yourself, do most of the heavy lifting. And so you have chosen to “pull out of”, or “abstract away from”, reality these three alone.

A good modeler always bears two things in mind, therefore: her skill is about identifying((and then verifying – this exercise us economists call econometrics, and we get very excited about it)) the factors that are most important. But a good modeler also always worries about whether she has missed an important factor. A good modeler therefore always walks that painfully thin line between certitude and hubris. And this is hard.

But now we’re faced with a new problem. Of the three factors that we have chosen, which is the most important? Is it all about weather, and not at all about inflation and location? Or is it almost entirely about inflation, and not so much about the other two? Or… you get the drift.

Which, finally, brings us to the point of this essay: The Truth Always Lies Somewhere in the Middle. Corner solutions aren’t impossible – it is certainly possible that it is only the weather that is causing the prices of onions in your neighbourhood to be so damn high. But I would say that it is unlikely. Location almost certainly is an influencing factor. And so also is inflation.

In fact, it’s worse, because for reasons discussed above, the truest shape to surround The Truth is some impossibly complicated polygon. We’ve chosen to abstract away from this polygon only three factors, and so we have the luxury of thinking about where The Truth might lie in this triangle.

But even in this simplified model, we should fight the urge to corner The Truth into a single vertex. It’s almost always more complicated than that.

  1. Is Thai cuisine good or bad?
    If you were to ask me, good! But are there Thai dishes that I don’t like? I should be clear: this is not a dish I have eaten, but I (unfortunately) have a mental bias against even trying this dish. Given what little I know of Thai cuisine, the loss is almost certainly mine – but hey, it is what it is.
    So is Thai cuisine good or bad? If you were to ask me now, after that last paragraph, almost entirely good.
    You see how what I choose to abstract away from reality helps me learn more about where The Truth might lie?
  2. Is Sachin Tendulkar a great batsman?
    In my opinion, almost definitely so. Now, I’m a Sachin acolyte. But even I, a rabid Sachin fan if ever there was one, know about his fourth innings average. I know that McGrath got the better of him in ’99 and (sigh) ’03. And so on and so forth. So on the Great-Not Great spectrum, I would place him very very close to the Great end of the spectrum.

Reasonable people can and should disagree about where on the spectrum The Truth lies. But a discussion becomes impossible, and therefore counterproductive, if you insist on clinging to just one tiny little dot in reality called Great (or Not Great).

This applies to economic models, political leaders, vaccination policies, American Presidents – and Thai cuisine and Sachin’s greatness, and everything else besides.

The Truth is mostly unknowable (and that is bad enough). But for us to have the hubris to think that we can pin it down to just one part of a binary is an extremely dangerous thing, and I think we would all do well to try and not fall into that error.

What explains the title of this post?

Well, I have Aadisht to thank for it. He has noticed, as perhaps you have, my tendency to use this phrase quite often in my posts here: The Truth Lies Somewhere in the Middle. Now, the abbreviation of this phrase doesn’t exactly roll off the tongue easily. TTLSITM isn’t likely to win me any marketing awards, alas.

But consider this magnificent wordplay:

Truth Always Lies Inexorably Somewhere in the Middle of Assertion and Negation

That’s a talisman I’m very happy to claim ownership of!

There just remains the small matter of deciding upon a suitable compensation for Aadisht’s time and expertise. But if you think about it, the idea was mine, and it was just the acronym formation that he contributed.

So between refusing to even acknowledge his contribution and giving him all the credit…