What makes a good policy analysis

A credible policy recommendation is not merely a correct idea. This session unpacks the distance between analysis that is right and a recommendation that can actually be used — a distance I only came to understand properly after moving from analysing policy outside government to working inside it.

The backbone is the four-step frame from the World Bank’s Regulatory Impact Assessment practice: define the problem and why intervention is warranted, state targets that can genuinely be checked, map the causal chain from policy to outcome, and test alternatives — including the alternative of doing nothing. To those four I add a fifth: political and institutional analysis. Who implements it, which budget line pays for it, and who stands to lose and has the power to block it. A recommendation that answers none of these is not a recommendation but a wish list.

The frame is then applied to the competition’s three subthemes. On the fiscal side, government revenue moved from 13.4% of GDP in 2000 to 13.3% in 2025 while neighbouring countries held at 20–21%; World Bank estimates put the VAT and corporate income tax gap at an average 6.4% of GDP over 2016–2021, with 58% coming from non-compliance rather than policy design — which makes raising rates the wrong instrument for the diagnosis. On the monetary side, the September 2025 placement of Rp200 trillion of government funds in five state-owned banks is read as a case: government deposits at Bank Indonesia fell from Rp451 trillion to Rp239 trillion in a single month, while what was actually weak was credit demand — working capital credit growth slowed from 12.4% to 4.4%. The most feasible instrument is not necessarily the one that fits the diagnosis. On the real side, manufacturing’s share of GDP fell from 32.0% in 2002 to 19.0% in 2024, and the downstreaming argument serves as an analytical exercise for testing the infant industry conditions and tracing who bears the cost.

The session closes with the Kuvukiland exercise, showing why a policy that is right in aggregate can still lose at the ballot box, a note on comparing mechanisms rather than outcomes when learning from other countries, and ten questions worth answering before an essay is submitted.