Monday, October 22, 2018

Let's not assume that

I accidentally set off several threads when I tweeted that maybe empirical evidence should guide what we think about the economy rather than pronouncements about what money is. One of the shorter sub-threads (and pretty much the only one I understood what people were talking about — j/k) included Nick Rowe and his old post on (the lack of) evidence favoring of fiscal or monetary policy. He's great because regardless of what you think about the ideas behind the toy models he builds or parables he tells, they're remarkably clear illustrations of the ideas. I'd recommend even the most hardened heterodox MMTer read his blog (here's a good one on stock-flow consistency).

The summary of Rowe's post is that if you have a fiscal or monetary authority (government, central bank) that targets some some variable it can affect — possibly imperfectly — under the assumption of rational expectations, then there'd be little evidence that the instrument used to target that variable had any effect. The fluctuations in the instrument or target variable are going to be the authority's uncorrelated forecast errors. It's "Milton Friedman's thermostat" (also well-explained by Nick Rowe in another post using an analogy with a driver on hilly terrain). The conclusion is that you should expect little evidence that fiscal and/or monetary policy works even if it does.

I’m pretty sure that it was JK Galbraith (with an outside chance that it was Bhagwati) who noted that there is one and only one successful tactic to use, should you happen to get into an argument with Milton Friedman about economics. That is, you listen out for the words “Let us assume” or “Let’s suppose” and immediately jump in and say “No, let’s not assume that”.
If assuming that a central bank with rational expectations stabilizes the economy will produce no evidence that a central bank with rational expectations stabilizes the economy, then what we have is effectively unfalsifiable (in the useful sense of Popper).

Let's not assume that, then.

What use is it to make these assumptions? They essentially prevent learning things about the economy. In fact, the most useful thing to do in this case — even if those assumptions are true — is to assume the opposite: that central banks (or fiscal policy) has no effect on the macroeconomy. Incidentally, this would produce exactly the same observation of a lack of correlation between the authority's inputs and the target variable output. In the worst case, at least you'd learn that you were wrong if the assumptions were actually true. And if you discovered robust empirical regularities about e.g. fiscal policy mitigating unemployment, then you'd learn that those assumptions of rational expectations and policy effectiveness are wrong in some way. It's a win-win.

You as the theorist should endeavor to maximize the ways in which you can be wrong through observations because that's how we learn [1]. If your preferred framework makes it impossible for data to shed light on it, then the best evidence you can provide is to assume the opposite and show how it fails to capture the data. These frameworks run the gamut from specific mathematical assumptions to more philosophical ones, but they have a single purpose: protecting beliefs from data. If this isn't your aim, the best course of action still would be to lean over backward against this bias [2] and seek out how you might be proven wrong.



[1] High energy physics (the particles and string theory stuff that people often think of as "physics" in a similar way to the way people think of macroeconomics as "economics") has been thought to be in a kind of existential crisis because it is too good at explaining observations — there's no place high energy physics is wrong, so we can't learn anything new.

[2] It might just perceived as bias by others, but that's the breaks. If we think you're biased to adopt a framework that lets you keep your beliefs by escaping comparison with the data then it's unfortunately on you to disavow us of this belief.


  1. The central problem though is still that there simply isn't enough data to begin to resolve most questions in the field.

    1. I agree that there is an uninformative data problem (as Noah Smith writes about). That's somewhat orthogonal to what I am talking about, but has a parallel issue -- in a sense it is the models that are assumed that make the data uninformative. The data is uninformative because the models are too complex to be rejected by it. It's a different kind of immunization against the empirical data, but the effect is similar: models aren't rejected by the data.

    2. Anti make a good point here. However, I want to add something else. Noah Smith et al often argue that there is not enough data. However, next minute they proclaim their own genius in solving problems in the absence of that data. That’s not consistent with evidence-based anything.

      You say that the answer should be to use simple models. However, simple models of the economy always fail. Marxism, Libertarianism and Monetarism are all simple models. All have failed.

      In the real-world, I’d say that the correct answer is that we should be cautious. However, economists are anything but cautious. That is a major problem.

      Also, if we need more data, or more accurate data, or more timely data, we should ask what data we need and how long it will take us to collect that data. If we assume we would need data from around 20 Great Recession-type events in the US, and if we assume that we have two of these each century, we would need 1,000 years to collect that data. What should we do in the meantime?

      Economists like to say there is not enough data when someone else makes a proposal. However, miraculously, there is always enough data when they make a similar proposal. That’s not consistent.

  2. I enjoyed this post as I enjoy almost all of your work, but you appear to endorse an argument made by Nick Rowe that I think might benefit from a bit more probing (even if, in the end, everyone agrees it's correct). I mean, on one hand it seems obviously true – the driver on hilly terrain was particularly useful for illustrating how it might happen that a statistician or econometrician might not have useful correlations to latch onto when it came to trying to establish a correct relationship between pedal, hill, and speed, in particular, which are the causes of which effects. But less obvious, to me anyway, is the logical step that seems to say, "an econometrician does not have useful correlations to latch onto, therefore there is NO way to establish a causal relationship." Perhaps it's just the flavor of "uncorrelated, therefore hopeless" that is rubbing me wrong. Perhaps it is having been exposed of late to the work that Judea Pearl and others have done on causal inference over the past 20-30 years that has me convinced that, now when I hear "correlated" or "uncorrelated" and the issue is "causation," I think I have a hammer (or think I know a group of theorists who might) and so am always hoping to find a nail.

    Sorry if this comes off as so much gibberish. Again, Rowe's argument seems very good as far as its claims go, but perhaps along the line of Rowe's own "ask the driver" option, suppose one started, not with the statistics, but with a model informed by experience of how and in what order different variables influence the system and its outputs – Pearl seems particularly taken with the use of directed graphs for this purpose – do you think one might be able to get a different result and escape the problem of "Friedman's thermostat?"

    I have to admit, though, the longer I think of the driver on the hill (or the hypothetical technocrat with a hand on fiscal or monetary levers, as the headwinds or tailwinds facing the economy are shifting this way and that), the more of a long shot it seems to escape Rowe's conclusion. You larger philosophical points, on good versus bad theorizing, empirical reality, etc., which you once again make well and from yet another interesting angle, are also well taken. Thanks, again.

    1. Thanks, and yes I agree that I am essentially ceding (for rhetorical purpose) the thermostat argument to make a more general point.

      I had previously talked about the underlying issue with the thermostat idea in an older post:

      The TL;DR is that in order to make this argument, you have to assume you have a really good model of how the economy works in the first place that had to be deduced before e.g. the central bank could be effective. Otherwise, how did you know the gas pedal (monetary or fiscal policy) does anything at all?

    2. Anonymous: "suppose one started, not with the statistics, but with a model informed by experience of how and in what order different variables influence the system and its outputs"

      Economists seem to imagine that the only way of understanding the economy is using uninformative high-level statistics. However, that's not true. If we split the economy up into components, we can study each component before we try to understand the whole.

      In medicine, we don't try to understand the whole body as a single undifferentiated system, as the starting point in our study. Instead, we try to understand the basic components and their causes and effects e.g. the heart is a pump - it pumps blood - the blood transmits oxygen.

      We don't start with statistics and then try to figure out the existence of the heart. That would be dumb.

    3. "If we split the economy up into components, we can study each component before we try to understand the whole."

      Not necessarily, and there are many lines of reasoning (from general arguments about emergence to the specific SMD theorem) that tell us this is actually unlikely for macro.

      Sure, it sounds reasonable but there is literally no reason for this to be true. In the natural sciences, we can generally make this assumption because the mechanisms of natural sciences are (luckily!) such that effects e.g. decrease with distance. This allows us to isolate subsystems in most of the natural sciences, but we have no idea how to actually isolate subsystems in economics. You can't take a market for one good out of an economy and study it separately. I wrote more extensively about this here:

    4. One of the main symptoms of the 2008/2009 crisis was a massive failure in the banking system. It is perfectly possible to study the banking system in isolation to diagnose what occurred. Indeed, it is not clear how you would diagnose a failure in the banking system without studying the banking system. What does your technique have to say about how to avoid a similar banking crisis in the future?

      In the Great Depression (and James Stewart movies), many banks went bust with the loss of people’s life savings. This happened because people worried that their savings were not safe, so they panicked, tried to withdraw their money, and made a bad problem much worse. Policy-makers thought about this component of the economy and introduced deposit insurance. As a result, we did not see a repeat in the 2008/2009 banking crisis. Deposit insurance had a massive positive impact on the 2008/2009 crisis. Without it, the crisis would have been much worse. However, the thinking about deposit insurance didn’t involve any mathematical models. It was about risk management. Risk management is a more important tool for real-world policy design and delivery than mathematical forecasting.

      Further, many policy interventions are aimed at individual industries or products or geographic areas. I seem to recall an intervention specific to the car industry in 2008/2009. I spent about five years working on policy interventions in the UK agriculture sector. I can assure you that it is possible to study one industry in isolation to others. One of the biggest macro problems today is the inability of governments to tax multi-national digital businesses effectively. As far as I can see, your techniques have no relevance to any of these topics.

      Further, when multi-national businesses move profit to avoid tax, they distort high-level statistics such as GDP. This raises the question of what benefits we think we gain from analysing GDP at a high level without controlling for such distortions.

      You have confused your narrow economic forecasting technique for economics. Even if your technique were the most successful in the history of unconditional economic forecasting, it would still only ever be one tool in a much larger toolkit.

      When we study problems in any man-made system, at any degree of granularity, we need to start with the problems we are trying to solve and then ask what analysis techniques we need to solve those problems. You have never done that. Instead, you have taken your analysis technique and run with it to see what it can do. That’s admirable. However, it means that you often reject other, equally valid, ways of thinking about the economy which are more relevant to practical problem solving.

  3. I agree with a lot of this, but I will focus on where I don’t agree.

    Best practice in policy is called “evidence-based policy” precisely for the reasons you state. However, policy-making is not science. For example, policy-making involves determining which facts are important. If you think that GDP is the only thing that matters, whereas I am happy to sacrifice some GDP for the sake of a more equal society, we will not agree on the best policy. However, there is no objectively correct answer on whether my bias is better or worse than yours. That’s why we have elections.

    In terms of mathematical models, the three most important points are as follows.

    First, as policy involves changing the behaviours of factions in the economy (business, banks, rich households, poor households etc), useful models must include the concepts of these factions explicitly, so that we can test different behaviours and see how they might impact the economy.

    Second, as different policies involve different factions, we may need multiple models which focus on different sets of factions for different policy areas.

    Third, the factions we model must represent real-world factions. Again, that is precisely because, if we want to test our assumptions about faction behaviour empirically, we must be able to measure them, so they must exist in the real-world.

    Your models fail the first point as you ignore factions. That’s why you couldn’t meet my challenge about advising President Obama in 2008-2009 except in very vague terms.

    As far as I can see, models based on the representative agent fail the third point. There is no such thing as a representative agent in the real-world, so we can’t test assumptions about the agent’s behaviour empirically.

    So, what is the answer? Well, I’d suggest that the first steps towards the answer are probably in Wynne Godley’s work. Godley was a policy-maker before he was an academic, so he would have understood the things I have said here. Unfortunately, his modern-day advocates have misunderstood his models. The real USP of his models is that they are based on realistic factions whose behaviours can be tested empirically. His advocates, on the other hand, go on and on about “stock flow consistency” which is just an internal integrity check.

    On a separate topic, some policy-making is about ensuring that processes & systems are robust under stress. The key point here is not to forecast the timing of an adverse event. Rather it is to minimise the damage from the adverse event whenever it does occur, or, ideally, to prevent the adverse event from occurring. We have talked before about avalanches and earthquakes. They are a good analogy for this sort of policy. For example, we should talk about risk factors rather than causes e.g. a heavy snow fall is a risk factor for an avalanche, but the cause can be a random event like a single skier moving across a ridge. That is an important distinction. We don’t assume that avalanches are not associated with heavy snowfalls just because we have empirical evidence that one or more heavy snowfalls were not associated with an avalanche.

    This type of policy thinking gives rise to what I call “the paradox of risk”. If we recognise the risk of an adverse event but take actions to mitigate that risk so that no adverse effect occurs, the result is indistinguishable empirically from a situation where the risk did not exist. If I’ve understood correctly, that’s part of the explanation for Milton Friedman’s thermostat.

    In the 1990s, the IT industry warned that many computer systems would fail at the start of 2000 with catastrophic results. There was enough warning of this risk to allow anyone with a sensitive system to update it. The result … nothing happened in 2000. Many people then claimed that the problem was a hoax from the IT industry to drum up business. However, that was not true, and we need to understand why that matters in managing risk in any policy area.

    1. Per your there most important points:

      1. ... useful models must include the concepts of these factions explicitly ...

      It is bad modeling methodology to include things in the model that do not improve agreement with empirical data, or add large amounts of complexity relative to the quantity of available data. It's basically saying the model has to look like what your preconceptions of it are, but more generally is called over-fitting.

      2. ... we may need multiple models which focus on different sets of factions for different policy areas ...

      This may well be (there are different models in physics that apply to protons or planets). But how do we choose between these models? Ex post is just wrong ("I know after the fact that the 2008 crisis was a shadow bank run via Diamond-Dybvig"), and in order to choose models ex ante requires some specification of model scope.

      3. ... the factions we model must represent real-world factions ...

      This is a fine goal, but sets an artificial straitjacket: there exist classes of models ("effective theories") that could forecast well and tell you almost everything except for the set of observables that you specify. Would you throw them away because of that? Example in physics: we know little about quarks at the QCD scale, but have tons of accurate theories for various hadrons at that scale. The actual true and correct theory (we know it! it's QCD!) is in fact completely intractable, so we have to use effective theories.


      One general lesson that I think is great to take from physics and that physics would do well to popularize outside of the field is the idea of not imposing unnecessary structure on the world. In physics, it's usually put as "Nature doesn't have to follow your rules" (mathematical elegance, showing effects in certain observables). The economics version would be "economies don't have to follow your rules" -- we shouldn't expect our intuition or our human feelings about how things work to produce accurate descriptions of economies.

    2. No this is wrong. You haven’t taken enough time to understand the problem that we’re trying to solve. As a result, you are forcing your solution onto the wrong problem.

      Let’s go back to my challenge about advising President Obama in 2008/2009. I want to change the challenge slightly. For the rest of this comment, I am now a very abrupt gatekeeper to the President. You need to persuade me you have something useful to say before I’ll give you access.

      Last time, you said you would propose a work programme. That’s not a policy – it’s a vague idea. What size of work programme? Over what time-period? What if we doubled the size and halved the period? Or some other combination? Why is a work programme better than a tax cut to businesses? Or a tax cut to households? Or both? What size of tax cut would be equivalent to the work programme? What is the probability of success? And so on.

      What I need from you is quantified analysis of the impacts of these options (and any other options the President comes up with). Otherwise, in my challenge, I’m showing you the door. How will you meet this challenge?

      You say that you are a scientist. You say that numbers are important. You say that we need mathematical models. However, when I ask you a question on policy, the science, the numbers and the mathematics all disappear. You give me only “a work programme”. That’s not good enough if you want to see the President.

      What’s going on here is that you have not thought about how you would answer MY question to MY satisfaction using the types of standards we have discussed e.g. we want to make assumptions that we can test empirically.

      Real-world evidence-based policy-making is more like medicine than physics, so you also need to justify to me why using a physics-based approach is better than the medicine-based approach used elsewhere.

      Further, let’s stick with the medicine analogy. If we were talking about the human body, you are saying, effectively, that we can understand it only by assuming that it is just a collection of atoms. No structure is allowed in your models for some reason. In the real world, we use multiple perspectives of the body e.g. organs, skeleton, surface view; and multiple scales e.g. whole body, brain, area of brain. We also envisage multiple scenarios e.g. healthy adult body, adult body with cancer etc. We can use all that information and more to help diagnose and solve problems. When we find a new disease or find a flaw in a treatment, we need to make assumptions about risk factors, causes, effects, side-effects etc. We then test them empirically. Over time, this process leads to better diagnoses and treatments.

      Unless you can explain otherwise, your atom-based approach can do none of that either in either medicine or policy-making. It is just not correct to say that structure is not important in human understanding of how systems work, particularly when the systems are man-made.

      Also, why are physicists allowed to have more levels of structure in a single atom than you will allow in an entire economy?

      Also, I saw elsewhere that you say you are a Marxist. Where are your mathematical models justifying Marxism? Where is the evidence? Where are the assumptions we can test empirically? Marx used a basic structure of capitalists versus workers. You can’t be a Marxist without this structure, but you say that structure is not allowed. Marxism failed miserably in 20th Century Europe. Should we ignore that empirical evidence? Why?

      Abrupt gatekeeper storms out of the room. Are you up for the challenge or not?

    3. What size of work programme? Over what time-period? What if we doubled the size and halved the period? Or some other combination? Why is a work programme better than a tax cut to businesses? Or a tax cut to households? Or both? What size of tax cut would be equivalent to the work programme? What is the probability of success? And so on.

      What I need from you is quantified analysis of the impacts of these options (and any other options the President comes up with). Otherwise, in my challenge, I’m showing you the door. How will you meet this challenge?

      Look, I've already done this when I analyzed the effect of immigration on output:

      Instead of determining the size of a work program that would offset a recession, I showed how much of a recession would be caused by removing a segment of the labor force. If that's not good enough for you then I can't imagine how plugging in different values for your 2008 scenario is going to convince you. But if we take the 2008 recession as 10% of NGDP, then we'd need to increase the labor force by about 1.7% to 2.5%. Taking the smaller number, that's 2.7 million people at an estimated direct cost of 135 billion dollars per year (using a 50k median wage) or 1.4 trillion over 10 years (as policy packages like this are usually priced in the US).

      Also, I saw elsewhere that you say you are a Marxist. Where are your mathematical models justifying Marxism? Where is the evidence?

      That is something of a joke, but is based on the reality that the "quantity theory of labor" is an accurate model of nominal output as well as e.g. this:

      You also mention a couple of things ... "your atom-based approach" ... "you say that structure is not allowed" ... that I do not understand. These are not accurate representations of my approach. I assume individual humans are too complicated to be understood with simple models, so I don't try. I also don't *assume* structure — that is not the same as saying there is no structure.

      I can't imagine the hubris it takes to assume you know how humans work or assume you know the structure of the macroeconomy. I mean, that's what mainstream economists do all the time, so you're in good company! I just don't get it. Why? How can you predict an individual's behavior? How can you divine the structure of the macroeconomy by just thinking about it in your armchair?

    4. Another way to see that calling my approach "atom-based" are the various times I've said the econ doesn't have a second law of thermodynamics. If it was "atom-based", then the second law of thermodynamics would be a guaranteed theorem.


    5. Jason: “How can you divine the structure of the macroeconomy by just thinking about it in your armchair?”

      I might ask you the same question.

      One of the many differences between physics and economics is that, in economics, there are many conflicting perspectives.

      There are only a very limited number of conversations you and I could have about sub-atomic physics. First, I could listen to you talk about it, and ask questions for my own clarification. Second, we could have a non-technical conversation where, for example, I could ask what value society gets from continued research into theoretical aspects where we have little or no likelihood of empirical validation. That’s about it though. Essentially, you would talk, and I would listen.

      In economics, there are many more perspectives, so many people can contribute something and challenge other perspectives.

      I spent 30 years working in business and government. As a result, I know how businesses work, and I know that economists’ conceptions of business are both shallow and flawed. As I have said before, one of the biggest problems in economics is that economists are not the recognised experts in many aspects of the economy. I also know how policy-making works. I worked at the boundary between policy design & delivery. I am not an expert in policy design, but, over many years, I have spoken to many people who are. I know how they think. I know what support they would expect from economists. I know how to challenge policy-based thinking. I have also had a lot of experience with sorting out dysfunction in businesses and government departments, so I know how to diagnose dysfunction and how to get to the heart of the causes of the dysfunction. The academic economics profession is dysfunctional. I see economics from those perspectives.

      Also, I am prepared to invest time accessing insights from non-conventional perspectives e.g. obscure economics blogs by physicists. Also, I invest effort in non-mainstream perspectives from professional economists, with an open mind on what I will find. Also, and most importantly, I applaud the student societies which advocate pluralism. IMHO, they are the only hope for the future of the profession. However, even they spend too much time talking and thinking about themselves, rather than challenging themselves on how they can best provide a useful service to society.


    6. Three random examples.

      First, this year’s Nobel prize in economics was won by someone for his contribution to innovation. I was interested to understand what insights this might have for practitioners who spend their lives innovating and bringing about change in many applied fields. However, as far as I can see, the prize-winning contribution boils down to vacuous generalities such as “innovation comes from ideas”. Society already knows that, so these ideas appear to me like the Emperor’s New Clothes. Now, my perspective may be wrong. However, unless economists learn to express themselves more clearly, we need to challenge them to discover whether there is any substance behind the Emperor’s Clothes.

      Second, at the microeconomic level, I’d say that the single most important economic number for any business is the break-even price for its products. Thinking about break-even price is the easiest way to understand why prices are sticky in a recession. We don’t need macro-economic thinking to understand this. However, the concept of break-even price does not seem to exist in economic textbooks or economic discussions. Again, we need to challenge economists on why their perspective is valuable when, in this case, it conflicts with the real-world view of practitioners.

      Third, and most pertinent to you, macroeconomics will only useful if it helps policy-makers who follow best-practice in policy design and delivery. That means that economists need to start by understanding how policy is developed in the real world, and what data & tools are required to promote the use of evidence in that process. Only then can we think about what sort of models are best-suited to helping that process. That’s why the best historical macroeconomists have been the ones who helped solve real problems and focused their theory on ideas to support real-world problem-solving e.g. JM Keynes, Wynne Godley.

      The biggest problems in economics all stem from the fact that academic economists (and physicists) seem to have no concept that alternative perspectives are useful (or even exist!). That is why peer review is a joke in economics. All that happens is that people with the same narrow perspective, and sense of superiority, all agree to agree that they all agree with each other. You even do your own “peer” review, so all you are doing is agreeing with yourself.

      I value your contribution to economics as you provide a perspective which challenges the conventional view of mathematical models in economics. As far as I can see, you run rings round the professional economists in this area. However, I also note that hardly any of them comment on your posts unless you name them explicitly, and none of them take you up on the challenges you set.

      One useful technique in assessing dysfunction is to look at personal networks. I sometimes look at your Twitter activity, not for the content as I can get that here, but for evidence of who engages with you from the economics profession. My very ad-hoc analysis is that the many of the people who engage with you are Post-Keynesians – even though you seem to disparage everything they do. I have seen little engagement from the likes of Noah Smith. I also note that he never challenges himself to learn or write about approaches such as yours or other left-field perspectives. Note that, in my analysis of dysfunction, engagement with alternative perspectives = openness to new ideas = a good thing.

      Where I challenge your perspective is that you seem to assume that the only valid perspective is that of a mathematical modeller of relatively uninformative high-level statistics. Your post here is about challenging other people’s assumptions. However, we all make assumptions. The smartest people are often the ones who are willing challenge their own assumptions rather than just other people’s.

    7. Jason: “You also mention a couple of things ... "your atom-based approach" ... "you say that structure is not allowed" ... that I do not understand. These are not accurate representations of my approach”

      I think there is a language problem here. When I started reading this blog, you compared your thinking to the behaviour of gas atoms in a jar. My shorthand understanding of your technique goes back to that point. If you say that is wrong, that’s fine. That’s why I call it atomist.

      Let me take another tack.

      In diagnosing problems in human systems, it is often valuable to ask people to draw a picture of how they see the context of a situation or problem. When you do this, you see that people often conceptualise a situation very differently. Most people assume that other people conceptualise the world in the same way they do. However, that leads to endless trouble. This is obvious when you think about examples but hardly anyone does. For example, a physicist and a mechanic might conceptualise a car very differently. Neither view is wrong. Both are relevant to understanding a car and diagnosing problems.

      One of the oddest things about economics is that you rarely see a picture of the economy. That’s one of the reasons it is so difficult to get into economics. There is no simple picture of either the scope or the contents of the economy. As a result, there are insufficient shared concepts for sensible debate. People make up their own concepts complete with their own vocabulary. Chaos follows as effective communication is impossible. That’s part of the problem with the schools of thought.

      In general, as far as I can see, there are two very different popular conceptions of the macroeconomy – which correspond to political bias. The Monetarist conception says that the economy is made up of markets, and markets are made up of participants. The Keynesian conception says that the economy is made up of sectors or, more generally, factions. In this respect, the Keynesian view is similar to the Marxist conception that the economy is made up of capitalist and worker factions. These different conceptions lead to different models, different vocabularies and different biases.

      Jason: “I can't imagine the hubris it takes to assume you know how humans work or assume you know the structure of the macroeconomy. I mean, that's what mainstream economists do all the time, so you're in good company! I just don't get it. Why? How can you predict an individual's behavior?”

      Keynesians think in sectors (or factions). They don’t think about individual behaviour. The sectors are the structure of the economy. Policy-makers also think in terms of sectors and sub-sectors.

      My assumption is that you have assumed that other people have a Monetarist conception of the economy i.e. economy / markets / market participants. That fits with your comment here to me. However, that is not correct.

      Even at the microeconomic level, businesses rarely think about the behaviour of individual customers. They think in terms of trends in demand, seasonality factors and ad-hoc factors. The oil majors will think about demand in the Seattle area. Individual customers are pretty much irrelevant. The idea that individual people are relevant at the macro level is preposterous.

    8. Jason: “How can you divine the structure of the macroeconomy by just thinking about it in your armchair?”

      I might ask you the same question.

      I was referencing comparison to empirical data. If one just declares the economy to be one thing based entirely on ideas that originated inside one's own head and have not been compared to empirical data (that I may be able to access from my armchair these days, but would not be available if no one left theirs), it has no, as economists put it, external validity.

      One of the many differences between physics and economics is that, in economics, there are many conflicting perspectives.

      This is incorrect. There are many, many conflicting perspectives in physics. I actually wrote about this making some analogies with econ on this blog not too long ago:

      When I started reading this blog, you compared your thinking to the behaviour of gas atoms in a jar. My shorthand understanding of your technique goes back to that point.

      I've occasionally made analogies using that behavior (it's the simplest way to achieve a system where agents explore the entire available state space which is the actual requirement). But I think you are taking the "atomist" view the wrong way. When Boltzmann and company were developing thermodynamics, they did not know atoms existed or how they worked.

      Ever since my 7th blog post back in April 2013, I have been stressing this implication. I am trying to build a framework where I am not sure people exist nor am I sure how they work.

      I've actually criticized the "atomist" approach where people build little models of 'human atoms' (i.e. agent based modeling) on multiple occasions. E.g.:

      ... to be continued.

  4. By the way, it is the absence of factions in the conceptual thinking of Monetarists (and other right-wingers) that allows them to see policy makers (who should be included in models as separate factions) as Gods or Devils. For example, as the Central Bank is external to Monetarist models of the economy, Monetarists can imagine whatever powers for the Central Bank they desire. As a result, they conclude that, if their preferred policy fails, it was because the Central Bank wasn’t trying hard enough. That’s why Monetarism is a religion. There is no possibility of testing their assumptions.

    Interestingly, a few years ago, there was a major debate between Nick Rowe and Scott Sumner on one side, and many of their commenters on the other side. Nick and Scott called the commenters “people of the concrete steppes”.

    In your terms, the commenters were trying to get Nick and Scott to define assumptions that could be tested empirically. Instead, they made up a funny name for the commenters. I like Nick too. He is transparently a good educator. However, his limit, as with most other economists, is that he will not recognise his own untestable assumptions.

    1. I was one of those people of the concrete steppes!

  5. Thanks Jason. I really appreciate your saying that.

    Let's assume the car driver has irrational expectations. Specifically, assume the driver thinks that hills are always half as steep as they really are. So he sees an uphill stretch, presses the gas pedal, but not enough, so the car slows down. (Or he sees a downhill stretch, eases up on the gas, but not enough, so the car speeds up.) We would see a correlation between gas pedal and speed, but with the wrong sign. The gas pedal would look like a brake pedal; pushing down on the gas slows the car down.

    How to test the effect of money on inflation? One example: in 1992(?) the Bank of Canada said it was going to use monetary policy to bring inflation down to 2%, and keep it there. And that is (roughly) what happened. Either the Bank of Canada got very lucky, or else monetary policy worked in (roughly) the way the Bank of Canada thought it worked.

    Or we look for historical examples where the supply of money is not controlled by a central bank trying to target something, but by ships bringing in loads of cowrie shells that they accidentally discovered by sheer chance (at least we think it was mostly sheer chance). But those aren't modern economies, of course, so we can't be sure the lessons from those historical examples still apply today.

    1. It is not good enough to think of an example and then generalise e.g. Bank of Canada in 1992.

      We need to consider different scenarios. Suppose we were studying a car. The base case is a sober driver in good weather with a car which works. However, we needs to consider other possibilities e.g. icy weather and a drunk driver, or a car with a flat battery.

      We need to itemise as many scenarios as we can imagine and then test our theories against EVERY scenario. Otherwise, we will be caught out every time something outside the base case occurs.

      That's a key problem in economics. The idealised view of the world means that economists often think only of the base case.

    2. Hi Nick,

      Actually, I posted about the Bank of New Zealand "getting lucky" in exactly this way just the other day, and have now updated it with the same model making the same argument for Canada:

      The summary: you can forecast New Zealand's and Canada's price level in 2018 using data from before the mid-1980s by simply assuming the inflation shock of the 1970s had a Gaussian (i.e. normal distribution) shape (effectively this is what the dynamic information equilibrium model does). Since you can do that, it's not obvious events in 1989 (New Zealand) or 1991 (Canada) should have any effect. Central bankers could get "lucky" and target 2% inflation just by looking at data before 1970 (in Canada) when inflation was "about 2%".


Comments are welcome. Please see the Moderation and comment policy.

Also, try to avoid the use of dollar signs as they interfere with my setup of mathjax. I left it set up that way because I think this is funny for an economics blog. You can use € or £ instead.

Note: Only a member of this blog may post a comment.