Accounting for these dynamics matters: ignoring adjustment costs understates substitution potential and overstates needed carbon taxes. With the improved model, Germany could cut fossil use with a 50% lower carbon tax and ⅓ the output loss. 🌍

When output rises, electricity use jumps 3x more than fossil fuel use. Standard production models predict equal responses so they miss an important real-world friction. Thore shows this can be explained by adjustment costs in fossil fuel use.

Observable traits (like location, industry, or products produced) can’t explain this heterogeneity. Something else shapes how “green” a plant’s energy mix is. Thore digs deeper into how plants adjust fossil vs. electricity use when demand shifts.

Fossil fuels make up 70% of energy use in German manufacturing. But across plants, energy mixes differ a lot: there’s 5x more variation across plants (even within the same 4-digit industry) than within plants over time.

🎁Bonus takeaway: Even without a slowdown in innovation by dominant firms, concentration can quietly slow growth by hampering the reallocation of resources.

When concentration rises, productivity first accelerates—as resources shift toward more productive firms—and then slows as reallocation opportunities diminish, mirroring the U.S. productivity boom of the late 1990s and subsequent slowdown.

📊The model replicates key patterns in the data even though firms’ productivity processes remain unchanged.

In economies with distortions, the theory predicts a stronger slowdown when sales are more concentrated than costs.
📉 A 10 pp increase in the gap between sales and cost HHI predicts about a 13 pp decline in five-year productivity growth.

In efficient economies, the model predicts that higher sales concentration slows productivity growth.
📉Empirically, he finds that a 10 pp rise in the Herfindahl index of sales concentration across industries predicts about a 3 pp decline in five-year productivity growth.

Juan develops a multi-sector model with granular firms that are hit by random productivity shocks.
💡The key idea: When production is concentrated in a few large firms, idiosyncratic shocks generate smaller reallocation gains — a mechanism Juan terms the granular drag on growth.

🧠 Feedback also explains a sizable share of manager quality, measured by value-added to worker productivity.
💡New insight for management: feedback is more than information transfer. The tone and information shape worker productivity and retention.

🛠️ Constructive feedback doesn’t affect future code quantity and lowers quality, as developers revisions to reviewed code crowd out careful new code development.

❌ Toxic feedback reduces future code quantity , quality, and retention
⚪ Non-toxic criticism has no such detrimental effect.
✅ Positive feedback increases productivity, retention, and even spillovers to coworkers.

Using LLMs, she classifies 200M+ feedback messages by tone (🤬toxic/nontoxic, 👍positive/negative) and content (🛠️constructive/non-constructive), leveraging random reviewer assignment to identify causal effects on developer productivity and retention.

A counterfactual 10% tariff on imported inputs would offset most of these gains and compress the firm size distribution.
The analysis links technological changes in input market efficiency to policy instruments like tariffs that affect firms in similar ways.

In general equilibrium, a 40% reduction in supplier search costs raises:
📈 Real GDP ~13%
🏢 The share of output produced by the largest firms ~12.7%
Aggregate activity shifts toward the most productive firms, while non-importing firms lose market share through general equilibrium effects.