Streetwise Professor

Recently there have been quite a few articles declaring the death, or at least the senescence, of the shale boom. This from today’s Bloomberg, about the Permian specifically, is one of the more optimistic takes.

The gravamen of the argument of those digging shale’s grave is that the most promising prospects have been drilled already. This is no doubt true–and I’ll present some evidence of that shortly–but it’s hardly the entire story. When one looks more comprehensively at the shale boom, it becomes clear that it was driven by technological progress that has overwhelmed the traditional sources of declining productivity in natural resource extraction.

I recently completed a paper on the shale boom. It examines the sources of productivity growth in both oil and gas unconventional wells. In particular, it quantifies the impact of learning-by-doing on productivity growth on a well-by-well basis in all of the major production basins.

The empirical framework captures three potential sources of productivity growth. Firm specific learning, basin-wide learning, and exogenous technological change. As is conventional, I measure the former effect by the cumulative number of unconventional wells drilled in a basin by a given firm prior to drilling a particular well. The second effect is measured by the cumulative number of unconventional wells drilled by all firms in a basin prior to the drilling of a particular well. The last effect is captured by a time trend (again conventional in the learning-by-doing literature dating back decades).

I examine a variety of productivity measures. In what I consider the most novel and potentially interesting part I also look for evidence of cost-reducing innovation and learning effects.

Productivity measures include things like initial production, maximum production, production over the first 12 months, and decline rates. I find strong evidence of firm-specific learning effects in the first three variables, but not so much in decline rates. (Interestingly, learning does not appear to improve drilling speed, contrary to empirical findings in conventional wells.) I do not find strong evidence of industry-wide learning effects.

The last finding sheds light on the exploitation of most promising prospects first. The cumulative basin-wide experience variable is also impacted by this effect. More wells drilled means more industry experience, but it also means more of the good prospects have been drilled. Those two things offset, leading to coefficients that are small positives or actually negative.

The crucial thing to note is that productivity increased from 2011-2020 (in oil) despite the impact of going to progressively less promising sites. This demonstrates the importance of learning and exogenous technological change.

The cost results are the most fascinating to me. I regress the number of wells drilled in a given month in a given basin against the learning variables, input cost variables, a time trend, and price (instrumented for gas to take into account endogeneity–the oil price is reasonably exogenous). I find strong industry-wide cost reducing effects of learning. Specifically, holding price and input costs constant, the number of wells drilled in a given month increases strongly with cumulative industry experience in a basin. That is, cumulative experience shifts out the supply curve. This is evidence of declining cost, and in particular declining fixed cost.

Here again you would expect that the exploitation of the low hanging fruit first should lead to higher costs as cumulative experience grows. But if that effect is there, it is overwhelmed by learning-driven cost reductions.

Based on this research, I am more bullish about the prospects for unconventional production growth in the United States than the conventional wisdom is. The conventional wisdom focuses on a single margin: the stock of potential drilling locations. That totally overlooks the real shale story: massive technological improvement, largely driven by learning effects. Those learning effects work on a variety of margins, including getting more out of a given well, and reducing the cost of drilling a well.

In essence the conventional wisdom is like neoclassical growth theory, in which diminishing returns are the depressing fact of life. But as modern growth theory emphasizes, technological progress has overcome diminishing returns. That’s why we are so rich–far richer than neoclassical growth theory can explain.

My interest in learning-by-doing dates back decades, to my amazing experience of taking bob Lucas’ undergraduate economic growth course at Chicago: Bob decided to teach the course as a way to master the growth literature, and so those fortunate few of us in the class were witnesses to the genesis of his research on growth, which is more important than his (still important) macro/money research for which he won the Nobel.

I wrote a few papers in grad school on LBD, for example showing how learning effects drove productivity growth in US gun manufacturing (at Springfield and Harpers Ferry Armories) in the 19th century. Researching learning in shale gave me an opportunity to dust off and update that previous research interest.

I would also note that shale pessimism is focused on oil. Gas has continued to go great guns–despite the fact that the same diminishing returns effect should be operating there as well. US gas production has continued to grow, especially in Marcellus and Permian. The latter is largely associated gas, but the former is not. This is a productivity story, and it’s not like diminishing returns don’t operate in Marcellus.

Indeed, gas supply growth has been so robust that prices are hovering around $2/MMBTU–back to the level prior to the spike in 2021-2.

Of course one cannot count on the rate of technological improvement continuing at the rate observed in 2010-2020 (for oil) and 2006-2020 (for gas). But one should certainly not discount it, and one should definitely not ignore it altogether and focus only on a source of diminishing returns.