Firm performance, macroeconomic conditions, and “animal spirits” in a Post Keynesian model of aggregate fluctuations

Abstract:

We construct a multiagent system (MAS) model of cyclical growth in which aggregate fluctuations result from variations in activity at the firm level. The latter, in turn, result from changes in “animal spirits” or the state of long-run expectations (SOLE) and their effect on firms’ investment behavior. We focus on the impact of publicly available information about macroeconomic conditions—analogous to the press releases of national statistical agencies—on changes in the SOLE and hence the amplitude of aggregate fluctuations. Our results suggest that the amplitude of fluctuations is reduced by extremes of attention or inattention to aggregate economic performance, but that this relationship is subject to complicated (and possibly complex) phase transitions exhibiting extreme sensitivity to initial conditions.

Key words::

• aggregate fluctuations
• animal spirits
• cyclical growth
• multiagent system
• sentiment
• state of long-run expectations

JEL classifications::

• C63
• E12
• E32
• E37
• O41

Notes

¹This information is analogous to the regular press releases of statistical agencies such as the Bureau of Labor Statistics and Bureau of Economic Analysis.

²Kahneman et al. (1986) actually refer to “reference transactions”; “reference points” generalizes this terminology to allow for the fact that not all significant economic events take the form of transactions. See also Ball and Moffitt (2001) for an example of an application of this concept to economic decision making.

³Kregel (1976) identifies feedback from disappointed short expectations onto the state of long-run expectations as the essence of Keynes’s “shifting equilibrium” model of effective demand. See also Dutt (1991–92) and Setterfield (1999) for further discussion and one-sector models of this shifting equilibrium system.

⁴Equation (7) also codifies the nature of agent interaction in our model. MAS models often feature networks in which one agent must be linked to another in order for the two agents to interact. Our model, however, does not involve network connections. Instead, firms pay attention to a “blackboard” from which they derive common information about the performance of the aggregate economy. It is reference to the blackboard that (assuming $${\kappa }_j\ne 1$$ in Table 1 discussed below) is the basis of firms’ interaction. Put differently, instead of the “direct” interaction between individual agents that characterizes networks, our model exhibits “indirect” agent interaction, resulting from the sensitivity of individual firm behavior to aggregate economic outcomes that are a product of the actions of all firms.

⁵It is possible that in any period t and for any firm j, none of the conditions in the first column of Table 1 will be satisfied. In such cases, the SOLE will remain constant and firms so-affected will begin to converge to a steady state rate of capacity utilization consistent with the value of α established up to that point. Note that as long as κ ≠ 1, however, the traverse toward this steady state may never be completed. This is because any sufficiently large subsequent fluctuation in aggregate capacity utilization will cause firm j’s SOLE to vary once again.

⁶This is true as a matter of logic, but does not enter into the formulation of the aggregate (one-sector) structural model from which the MAS model in Equations (7)–(9) is derived—hence the amendment to the model discussed in what follows.

⁷The requirement that α be nonnegative follows from the canonical configuration of the Kaleckian growth model on which our MAS model is based, in which α≥0 is part of a parameter configuration that ensures the existence of positive and stable steady-state rates of growth and profit in all periods.

⁸Note, in fact, that we need to know the value of u_t to make the calculations described in Table 1 and so to update our simulations between periods.

⁹On the concept of the reaction period, see Asimakopulos (1991, ch.7).

¹⁰Note that, because there is only one firm, the degree of isolation (κ) is irrelevant in this exercise.

¹¹The precise value of this path-dependent steady state will depend on the terminal numerical value of α that enters into the calculation of u^*in Equation (12).

¹²The criteria in Table 1 involving Δu_jt − 2 play no role in determining Δα_jt in the first period of our simulations. This is because of the steady state prehistory with which we endow our model in its initial set up, as a result of which the criteria involving Δu_jt − 2 are always false in the first instance following the initial shocks to firms’ capacity utilization rates.

¹³A good example of this is what happened at the opening of the Millennium Bridge in London. As recalled by Shin (2005), shortly after the opening of the Millennium Bridge over the Thames in London in June 2000, the bridge began shaking so violently that it was subsequently closed for more than eighteen months. Engineers discovered that the event was triggered by a lateral disturbance at 1 hertz—which could result from a normal human footfall, or a gust of wind. And as Shin notes, once the initial disturbance occurred, because people react to their environment, they adjusted to the disturbance in a similar fashion at the same time (in an effort to steady themselves on a moving bridge). This synchronous behavior fed back to the bridge, causing it to move in a more exaggerated fashion, which caused further synchronous adjustments by the people on the bridge, and so on—resulting in the violent shaking recorded by BBC cameras covering the opening of the bridge. In short, the initial shock was amplified by adjustments within the system. These adjustments were likely to recur in the event of any subsequent small shock, resulting in repeated violent shaking of the bridge in the normal course of its use—a fact that resulted in its closure until shock absorbers were installed.

¹⁴It may appear strange to refer to “aggregate fluctuations” when there are only two firms. Note, however, that our two-firm model is analytically equivalent to a two-sector model (the like of which is abundant in the macrodynamics literature) and that most macrodynamic analysis, drawing on representative agent methodology, features only one firm.

¹⁵The reference to an absence of orderliness when κ = 0 is based on observed firm behavior outside the steady state. When both firms achieve the steady state then behavior does, of course, become perfectly orderly. But since steady states are, by definition, associated with an absence of change, this is not an orderliness in behavior that will produce fluctuations at any level of aggregation.

¹⁶$${\hat{\sigma }}_u$$ is calculated as the standard deviation of simulated capacity utilization over 100 periods (the length of each individual run). It is therefore a measure of the average amplitude of aggregate fluctuations over time.

¹⁷Refer back to the introduction for examples of such studies.

Additional information

Notes on contributors

Shyam Gouri Suresh

Shyam Gouri Suresh is in the Department of Economics, Davidson College, Davidson, NC.

Mark Setterfield

Mark Setterfield is in the Department of Economics, New School for Social Research, New York, NY.