## Notes from 16 February 2026 [[2026-02-15|← Previous note]] ┃ [[2026-02-17|Next note →]] I found a fascinating working paper today via [[Mark Thomas]]’s excellent new Substack, _[Enabling Acts](https://open.substack.com/pub/markthomas10x/p/january-19-2026?r=10o1&selection=d705ad6a-a46e-4a98-a954-567a108b6254&utm_campaign=post-share-selection&utm_medium=web&aspectRatio=instagram&bgColor=%230891b2&textColor=%23ffffff)_. It’s by Nadja Dwenger and Anna Gumpert, and it deals with **building administrative capacity**, specifically looking at the evidence from German Reunification. The study focuses on **secondments**: the practice of sending officials from high-capacity units (West Germany) to low-capacity ones (the newly formed East German tax administration). What’s shocking is the scale of the challenge they faced: the GDR was a socialist state where taxes were subordinate, so they basically had to build a modern tax administration from scratch starting in 1990. The researchers collected a massive, novel dataset to provide the first causal evidence on this. Their findings are quite powerful: - **ROI:** Every day a West German official spent training East German staff had a return on investment of **1.5 to 3.1** in terms of output quantity (cases assessed). - **Quality over quantity:** While the "quantity" of work converged over time as the East caught up, the quality improvements were persistent. Offices that received more support continued to have fewer errors and fewer taxpayer objections in the long run. - **Tacit knowledge:** The main mechanism here wasn't just "more hands on deck." It was the transfer of **tacit, experience-based knowledge** - the "know-how" required for complex discretionary judgments that you can't just learn from a manual. The paper also highlights a few design keys: support is more effective when it comes from offices with **strong administrative traditions** (like former Prussian regions), and **intermediate durations** (around 3 months) are optimal to avoid the "rotating door" problem of frequent turnover.