Government as landscape designer: A behavioral view of industrial policy

Paper highlights

Industrial policy changes the landscape on which firms search. Incentives make some choices more attractive; regulations remove choices from consideration. This paper models those interventions while allowing firms and governments to be boundedly rational.

The model qualifies two common claims. Policy instability is not always harmful: changing incentives can teach firms about different parts of the landscape and dislodge them from local peaks. And a government with limited predictive ability can sometimes improve firm performance, particularly when it uses incentives and revises them periodically.

How the model works

Firms conduct local search on a performance landscape. A government imperfectly estimates which choices are promising and intervenes through incentives or regulations. The simulations vary government ability, policy intensity, frequency of change, and the interdependence that makes the firm’s environment more or less complex.

Main results

• Periodic policy changes can improve search through a training effect—exposing firms to useful information—and a dislodging effect—moving them away from a local peak.
• When government ability is limited, adjustable incentives generally work better than regulations because firms retain the option to reject bad guidance.
• Complexity raises the intervention scale needed to alter firm behavior, while also limiting the damage caused by inaccurate policy.
• In a low-complexity setting, even a small but misguided intervention can be harmful because it can redirect search effectively in the wrong direction.

Why it matters

• Industrial policy changes the landscape on which boundedly rational firms search; firms do not instantly move to a global optimum.
• Policy instability can generate learning by exposing firms to several regions rather than keeping one region persistently attractive.
• Incentives preserve firms’ ability to use local information, whereas regulations remove options; this difference matters when government knowledge is limited.

How to use this paper

Cite this for

• A behavioral model of government as a designer of the search landscape firms face.
• The training and dislodging mechanisms through which policy changes can affect firm search.
• A comparison of incentives and regulations when governments and firms are both boundedly rational.

Useful for teaching

• Industrial policy as a change to the landscape, not only a shift in prices or constraints.
• Why policy instability can sometimes aid search by moving firms away from local peaks.
• How bounded rationality changes standard claims about government intervention.

Careful claim

The model identifies conditions under which changing incentives can improve firm search; it is not a general claim that policy instability or low government ability is beneficial.

Abstract

The strategic management literature has built rich and behaviorally plausible models of firms, yet the industrial policy literature has overlooked nuances in firm behavior. This paper bridges these two literatures by incorporating increased micro-level realism to examine how industrial policy affects firms. More specifically, we develop a formal model to study how commonly held results on the effects of two prominent types of industrial policy—regulations and incentives—change when we account for the behavioral aspects considered in strategic management.

We specify conditions under which, contra results in the industrial policy literature: (i) policy instability can be beneficial (through what we term the “training” and “dislodging” effects) and (ii) firm performance can benefit from the industrial policy of a government with limited ability for identifying and enacting optimal policies. We also show how environmental complexity, an understudied factor in that literature, is a strong moderator of the effect of industrial policy. Managers can use these results to devise better means of coping with and leveraging the effects of industrial policy. Our findings also have implications for organizational search and R&D governance.

Other links

• SSRN preprint

Last updated 2026-06-21