Quick Comment On SFC Models And Mainstream Approaches

by: Brian Romanchuk

Cullen Roche has written an introductory article on Flow of Funds approaches to economics: "Understanding Flow of Funds Accounting." He asks why (post-Keynesian) Stock-Flow Consistent (SFC) models have not caught on with the mainstream. I think this is easily explained by the buried mainstream assumption that economic models must be based on the mechanic that prices bring supply and demand into balance. The post-Keynesian SFC models (as described in my primer) call into question this role of prices. In this article, I am just commenting on why this distinction exists, and not attempting to say which is better (although long-time readers will know of my bias towards SFC models).

Cullen Roche:

Although the term "stock flow consistent" has become popularized by Post-Keynesian Economists, it's really just macro integrated accounting. So, please don't confuse "Post-Keynesian" as having a monopoly on macro accounting or flow of funds accounting frameworks. In fact, if you've ever cracked the Federal Reserve's Z.1 report, probably the most important data set the US government compiles for the economy and the source of much of the St Louis Fed's FRED database, then you've dabbled in "stock flow consistent" modeling which is more commonly referred to as "flow of funds" accounting by the Fed. This approach to economic modeling was introduced by Morris Copeland in the 1940s when he brought it to the Federal Reserve.¹ Although it is widely used in the Fed and on Wall Street it hasn't made much impact on more mainstream academic economic modeling techniques for reasons I don't fully know. [emphasis mine - BR]

(Although he has a lot more to say, I am just responding to the highlighted text.)

Firstly, I agree with Cullen Roche that the name "Stock-Flow Consistent Models" is somewhat unfortunate; we have a label that is not in perfect accordance with what it refers to. I will use the term "SFC models" to refer to the models developed by post-Keynesian economists (such as in the text Monetary Economics by Godley and Lavoie). Technically, these are models in which stock and flow accounting is done correctly (unlike many classical models), but also contain post-Keynesian behavioural equations. One can imagine models in which the accounting for stocks and flows are done correctly, but do not have post-Keynesian behavioural equations. (As an example, take the nonlinear versions of many DSGE models.)

In the same way that we can have an "Austrian" economist who does not come from Austria, we have to accept that there are models which have consistent stock and flow accounting, but are not "SFC models." I apologise to the reader, but it's too late to change the terminology.

Why Not SFC Models?

The complete answer to the questions as to why "SFC Models" have not caught on with the academic mainstream probably involves academic politics. However, the most reasonable explanations are:

  1. model-consistent forward-looking expectations, which is related to the "Lucas Critique";
  2. more experience with fitting mainstream models to data;
  3. the use of prices to bring the model into "equilibrium."

There is a lot of controversy around "micro-foundations." It has often been argued that SFC models are unacceptable to the mainstream because they are not "micro-founded," in the sense that the model solution is the result of an optimising decision by households. Although there is a fundamentalist fringe who probably believe that, I would argue that why "micro-foundations" matter is because of my first point (model-consistent forward-looking expectations).

Rational Expectations

I touch on model-consistent expectations in my previous article - "Finding the Solution in a Simple SFC Model." These models can have "point-in-time" model-consistent expectations, but they are not forward-looking (they do not incorporate the expected path of economic variables) - "rational expectations." These expectations are dealt with at the aggregate level, which might not correspond to any individual entity within the aggregate.

Do forward-looking expectations matter? In certain cases, the answer is certainly yes. As a strong believer in the rate expectation theory, bonds need to be priced based on the expected path of the short rate. (In fact, Modern Monetary Theory is much more fundamentalist about rate expectations than many mainstream economists; they are forever finding stories which explain how mysterious risk premia will cause the bond market to crash.) More importantly for macro theory, if you have a currency peg, expectations about that peg matter a lot. However, once we step beyond cases like those, it is unclear how much we need to worry about long-term expectations (assertion based on empirical work done by post-Keynesian economists).

If the SFC models did add in forward expectations, I would guess that most of the "Lucas Critique" might be answered. (It has been a long time since I read up on the Lucas Critique, and so that is just my impression.)

(The obvious question is: can it be done? I would argue that working in forward expectations to a model you can already solve is going to be more tractable than the approach generally used in DSGE models: write down a complex nonlinear model, and then jump (somehow) to a log-linear approximation.)

Model Fitting

The second point - that there is more experience in fitting DSGE models to data - probably matters a lot, but it is not necessarily an inherent weakness of SFC models per se. For someone at a central bank, fitting a model to data matters, so this is an important point. The edge in model fitting for DSGE models (at least the linearised variants) reflects the greater mass of researchers working on the topic. For SFC models, this is a bit of a chicken-and-egg problem; without a similar mass of researchers, advancement in this area is going to be slower, reducing the pace of acceptance. [Update: added this paragraph.]

Post-Keynesian Behaviour A Step Too Far

However, the real bone of contention revolves around the role of prices within the model. A model in which prices do not bring supply and demand into balance is a step too far for the "mainstream." (By contrast, post-Keynesian argue that most prices within the economy are administered; only a handful of "flexprice" markets act like mainstream theory suggests - such as the financial markets.) For example, the model I solve in my linked article has fixed prices - I would expect that many economists would argue that it is not an economic model by definition. Furthermore, the linear production function is also excluded by definition - everybody knows that marginal productivity is decreasing.

These objections revolve around microeconomics, and so it might be thought of as a variant of "micro-foundations." In this case, we are defining "micro-foundations" as being that aggregate behaviour matches what is expected based on microeconomic reasoning. (It is unclear to me whether that qualifies as "micro-foundations.")

If we take out the post-Keynesian behavioural assumptions, we end up with a nonlinear DSGE model. In some cases, these can be solved by brute force calculation. However, the usual response to log-linearise, breaking the accounting identities. In other words, the mainstream values their behavioural equations over keeping the accounting consistent. Since that it is a judgement call, there is no reason to expect their preference to change.

[Update] The exact complaint about post-Keynesian assumptions appears more complex than I discuss here, as pointed out by Philip Clarke (@pilkingtonphil). Since it appears that the mainstream largely ignores the SFC modelling literature (are there any citations?), we cannot be entirely sure what the true underlying objections are. Since I wanted to keep the discussion limited to "reasonable" objections, that's not a direction I wanted to take the article. (For example, post-Keynesian article submissions might be blocked since they do not cite the authors favoured by the anonymous reviewers at mainstream journals. Although this is probably happening, it is not a "reasonable" problem from the perspective of anyone outside of academia.)