Thinking in Systems, Ch. 4: Why Systems Surprise Us

this video captures chapter four from this book which i'm going through chapter by chapter and this chapter is particularly relevant to economists she actually comes up with a critique of econometrics she talks a lot about economics including behavioral economics bounded rationality so this is going to be a fun chapter for me to talk about so basically systems surprise us we can't predict the future the economy does crazy things and econometricians never really capture that in their models so the question is why do systems surprise us and i think it's helpful to think of systems as

these complex things with many different threads that are interconnected and if this diagram looks like a cell in a body then that actually makes sense because cells are systems as our body systems as our economic systems and the key is any model you come up with is going to have only a few of the threads of the system working together in fact that's the point of a model is to sort of simplify the many interconnected forces so that you can isolate the effect of those those two or three or four different forces on each other

without taking into account the full complexity of the system so the short answer to why do systems surprise us is just the fact that the models we have in our heads and in our math of systems are inherently limited they they simplify systems down to one or two different factors and maybe we can make good predictions as long as the system is sort of only limited or only working on those two or three factors but when systems surprise us it's usually because some other factor that's at play in the system that's intertwined with the things

we're studying in ways we hadn't considered those factors come up and they they make an impact in a way that just wasn't built into our models now she points out something that is absolutely true which is that a lot of times our models can have truly valid assumptions and yet they completely miss major things that happen in the system and other times we can have untrue assumptions or invalid assumptions in our models and still make good predictions and that's just because of the complexity of systems okay in this chapter she cautions against making too much

of our analysis of systems dependent just on events because events oftentimes result from some thing in the system sort of popping up its head that we hadn't expected and really systems analysts need to think about the underlying system structure and how that relates to events that pop up and resulting oscillations okay she gives a critique of econometrics which i think is a valid critique and my guess is most econometricians would agree with her analysis here which is that econometrics basically takes flow variables it's sort of variables over time as they change and comes up with

a model that connects those flows and the problem with that is that the flows are only part of the system and the flows are oftentimes reacting to the stock within the system in these feedback loop fashions and oftentimes the reason the flows move together or moves move in opposite directions or whatever is is because there's correlated to some other factor inside the system they're responding to the dynamics deeper in the system so if you pick any random five flow variables that are built into the system you can build an econometric model that sort of relates

those those variables but that doesn't necessarily capture anything about the underlying structure of the of the economy or of whatever system you're looking at and because of that econometric models are really good at predicting the near future where everything's sort of going along just like it has been but they're not very good at predicting the long future when the system sort of changes and evolves such that different limiting factors different limiting stocks within the system become relevant and of course what do we care about we care about the change moments or at least we care

about the near future a little bit of course but a lot of times what we're really putting a lot of effort into predicting is the long-term moments of change and econometric models are not good at that now in talking through these examples one example she gives is just gdp of course that's something that economists track a lot and gdp is a flow variable it is sort of the extra productivity added in a given year but she makes the point that actually one thing that matters a lot is the stock like how much physical capital how

much human capital how much social capital has been built up in a system and what it does the next year depends not only on what it did last year but also on the built up system that's sort of supporting that underlying growth and gdp alone completely ignores that so so that's what she means when she says econometrics focuses on flows and oftentimes ignores the underlying stock in a way that is that can be fatal for analysis and she kind of makes the same point where she's saying okay why are people surprised when they try to

go in and fix uh developing countries by just you know adding some extra flow variables without considering the underlying stock one of her section titles is linear minds in a non-linear world and this is actually something that economists completely get right we know the world is not linear and as a matter of fact this is why economists actually sort of say you have to go back to the economic theory when you're doing econometrics you can't just go out there and do econometrics without considering economic theory because economic theory brings in the non-linearities so you can

kind of think about are we at a place where we're hitting diminishing marginal benefit or increasing marginal cost and let me draw those so the key here is that when you're doing your kind of metrics and analyzing flow variables you're going to get some you're going to capture some linear relationship between two variables that might look like this and the point is as the system grows and expands eventually if it's a benefit you're thinking about there's going to be diminishing marginal benefit eventually that relationship is going to change such that it's much flatter and there's

a huge difference between this and this and her example was when you fertilize a field uh you add you know 10 units of fertilizer and the crop yield increases by 50 well that that might be measured down here if you project forward you're going to be way off by the time you add you know 100 units of fertilizer and there's going to be some point at which by which you're adding too much fertilizer and the diminishing marginal benefit actually turns into a um into a bad thing adding extra fertilizer uh is is making the yield

go down so economists are very much in touch with this we know that things change and that there are these non-linearities and that's why we absolutely insist on this back and forth conversation between the modeling the theory and the econometrics she points out that limiting factors can change and of course she's said in previous chapters that the factor in the system that is most scarce or most most limiting that's what's going to determine the behavior of the system and i really like her example of this she says okay if you have a company and you

have sales force who's trying to get more people to buy the product and you have a production line that's producing the product either of those could be limiting factors so you might imagine the production line is producing really fast they're doing great but the sales force can't sell well if you hire a bunch of superstar sales people and those sales people try to sell and try to sell and they do a great job of it what could happen is you get a change in the limiting factor such that the sales force is selling so many

and you're getting a ton of orders in that enthusiasm is high and suddenly the assembly line cannot keep up at that point you've had a change in the limiting factor previously the limiting factor was your sales force now suddenly the limiting factor is the production team and that change happened because of a positive thing because of this uh outsized growth in the sales success and she she points out the fact that the success of a system like how well it does when it sort of gets on a roll often times that success can lead to

a change in which factor is limiting and relating this to economics she points out that the whole field of economics was developed in the uh you know 1920s 1930s i mean obviously there's a longer history of economics but a lot of the models we use really came about in an era that was a different era from the the one we currently face so capital and labor is really the focus of economics that's what you talk about in your econ 101 classes but it's possible that we've actually reached a point where there are different limiting factors

such as information such as social capital such as human capital so our economy may actually look really different in terms of its limiting factors but the textbooks have not really changed now i'm not really complaining about that i think it's helpful for economics to use really clear visualizable examples for the textbooks and then once people get down the basics then they can broaden their economics way of thinking to include squishier things like social capital and information capital and network effects which are really the big factors today but because those the factors today are all kind

of intangible and difficult to think about i think the e-context books are written the best way they can be written now one short point she brings up that i think is relevant to this question of sort of the information economy is that information can switch overnight information changes really fast that's both the stock and the flow of information can potentially be instantaneous whereas physical capital like machines and robots and even human capital are things that change only slowly over time and i think that's something that's worthwhile for economists today to consider that we have these

amazing tools for sudden changes in information but the rest of the economy the physical world economy moves more slowly the last section of the chapter that i want to talk about is her section on bounded rationality now i teach behavioral economics and bounded rationality is one of the key concepts in that course i will say i'm not 100 sure she's using it correctly most of what she talks about in this section is really about imperfect information which is part of classic economics she also talks about the tragedy of the commons which is captured in classic

economics now i think bounded rationality is relevant she just doesn't articulate anything to indicate this is something that's on her mind but so what is bounded rationality bounded rationality is basically the fact that each human is limited in the the amount of time and cognitive effort they have to give so they cannot make fully rational decisions so bounded rationality is essentially this meta model where people say okay because i cannot make every decision with enough effort to be fully rational i will optimally allocate my decision-making effort toward the things that are most important and of

course the result of that is imperfect information and therefore um she's relating this of course to systems and the key with systems is feedback loops which often depend on information so if you have bad information feeding the feedback loop the system can kind of go off the rails because of that bad information that's sort of embedded in that feedback loop she talks about tragedy of the commons which can lead to addictive products where you're creating a product and maybe it has a negative moral side effect such as creating addiction but it leads to super competitive

behavior in fact in some markets you will be competed out of the market if you don't do that anti-social thing and the antisocial thing can be create an addictive product uh harm the environment there's lots of different things like that now um some of what she's talking about here can be wrapped up in some of the game theory that we teach in classic economics multipolar traps capture some of this so i really think classic economics actually has everything you need to do what she does under the section of bounded rationality but i mean this this

chapter is not a comment per se on economics she just talks a lot about economics and i think she gets a little bit of it somewhat wrong now the last thing i wanted to say from this section is she points out the fact that sometimes you have feedback loops with really long time horizons and i mean climate change is one of these examples where the time horizon where this is going to become a huge irreversible problem is a longer time horizon and you may need people with the foresight to predict that to prevent it so

she's pointing out some of the really important problems in economics and she's pointing out why it's difficult for us to predict these things it's because our econometric models really can't do the job now the tools of the economic framework actually can do the job it's just that when you look at a complex system you can pull out a bunch of different parts of that system and look at them individually and see the tragedy of the comments at play or see imperfect information and the necessary uh things that are going on with that but when you

sort of hold that up to the system as a whole it's really hard to know which of these different factors in the in the vocabulary of economics is going to dominate and create feedback loops that matter the most 10 years from now and 20 years from now