Per my last post, the key takeaway from Dani’s industrial policy teachings is not to target sectors, but specific market or government failures.  Only if these failures are present is government intervention justified.  The big ones we should be looking out for are learning externalities and coordination failures.

But…it seems one can always justify learning externalities, whether in the form of learning by doing or learning about costs.  I didn’t get into these on the last posts, but let me dive in a bit on them now here.

Learning by doing refers to productivity increases over time getting because employees learn either from experience, each other, or via training.  The problem is that they leave and go to other firms.  Because of this firms reap the benefits of each other’s training, creating a positive externality.  This means the private benefit from training is less than the social benefit for training, so it is underinvested in relative to the social optimal.  If we assume that this happens to a greater degree in “modern” sectors than “traditional” sectors, then at an economy wide level, this creates underinvestment in the sectors that have increasing productivity over time because of learning by doing.  As a result, subsidizing training to align social and private benefits shifts economic activity towards more productive sectors and creates growth.

But anytime you have employee turnover (i.e. always) you’ll have this externality.  If labor laws were rigid enough to fully internalize learning by doing you’d have a government failure that would almost certainly constrain growth even more than this learning externality.  So pick and sector of choice and you’ve got your justification for industrial policy promoting that sector – at least via training subsidies.

Same thing goes for learning about costs when it comes to new sector activities.  Learning about costs essentially refers to the fact that entrepreneurs don’t know the cost structure of new activities in an economy until they try them, so whoever is doing this “learning” assumes all the risk if the activity turns out to be unprofitable.  If it does turn out to be profitable, free entry and competition would enable other firms would benefit from their cost discovery process, and the initial firm wouldn’t receive adequate compensation to justify taking on the risk in the first place.  Thus firms need to receive some rents, at least in the near term, to justify taking the risk to learn about costs.  This is completely analogous to R&D and patents. 

But again, any time there’s a new activity; you’re going to have this learning about costs externality.  So again… pick your new sector of choice and you have your justification for industrial policy, so long as it’s targeting this externality.

It’s important to remember Dani’s arguments that you should formulate policy to target the externality, not just head straight for tax breaks or credit subsidies.  But I’m unconvinced in reality there’s ever a case where government intervention can’t be justified, especially when certain sectors have political support for other reasons.