Idle Monetary Policy? Evidence from Monetary Policy Shocks for Pakistan
Description
Research hypothesis. The prevailing view is that monetary policy is largely ineffective in Pakistan, since interest rate changes appear not to move inflation and the economy. We hypothesise this reflects a measurement problem: central banks raise rates when they expect inflation to rise, so observed rate changes are contaminated by the forecasts that prompted them. Once purged of this systematic component, the remaining surprise should show that policy transmits normally. What the data contains. The core dataset is a meeting level record of all 50 SBP policy rate decisions from 2008 to 2019, with the SBP's own inflation projections for each meeting, drawn from Monetary Policy Statements and MPC minutes. The inflation projections are proprietary and have to be requested directly from SBP. However, we share estimated narrative shocks, quarterly macro-financial panel (industrial production, CPI, exchange rate, term spread, and private, public and total credit), a US policy surprise series, and global controls (oil and food prices, the VIX). Construction. Each decision is regressed on the rate carried into the meeting and the SBP's inflation projections. The residual is the narrative shock (see files), the part of the decision not explained by what the committee expected. Shocks are summed within each quarter; quarters with no meeting take a shock of zero, since no meeting means no unexplained change in intentions. This gives a quarterly shock series, 2008Q3 to 2019Q4. Findings. A one percentage point contractionary shock lowers industrial production by about 1.8 percent within two quarters, with no return to baseline over three years. Y/y inflation falls by 0.62 points on impact and by a further 0.91 points after two years, following an intervening rebound. The policy rate, term spread and credit responses carry the expected signs, though the latter two are estimated less precisely. Spillovers from US monetary policy are small and imprecisely estimated at the quarterly frequency. Interpretation. These results overturn the "idle policy" reading: transmission is present and economically large once decisions are cleansed of their forecast driven component. The inflation result is a rate effect, not a price level effect, since the cumulative price level response is insignificant. The credit results are directionally consistent nominal stocks, not a precisely estimated real credit channel. Use. The quarterly shock series can be used as an instrument or right hand side variable in Pakistani monetary transmission studies, in place of the raw policy rate change, which is not exogenous to the economy. SBP inflation projections are not fully recoverable for every meeting; missing values are interpolated, and results are robust to excluding the affected meetings. The package is organised in two folders: one with the meeting level code for constructing the narrative shock, and one with the quarterly panel and code estimating the local projections.
Files
Steps to reproduce
Meeting level record. We identified all 50 SBP Monetary Policy Committee decisions from 2008 to 2019 from published Monetary Policy Statements and MPC minutes, recording the meeting date and policy rate set. For each meeting we recorded the SBP's own real time inflation projection for the previous, current and next period, prepared for and available at that meeting, so the conditioning information matches what the committee had rather than later outturns. This projections dataset is proprietary to the SBP and must be obtained directly from the bank; it cannot be redistributed, so the projection column in the shared main_data file is left empty. To preserve reproducibility, the package separately provides the narrative shock series from our own estimation, so the second stage and results can be reproduced without access to the proprietary projections. Coverage is also incomplete for some meetings, particularly early in the sample; missing values were interpolated between adjacent meetings. Narrative shock construction. Following Romer and Romer (2004) and Choudhary and Limodio (2022), we regressed the policy rate change at each meeting on the pre meeting rate level, the SBP inflation projections, and output, output growth and unemployment. The residual is the narrative shock, the part not explained by the systematic rule, summed within each calendar quarter; quarters with no meeting were coded zero rather than missing, since no meeting implies no unexplained change in policy intentions. Macro-financial panel. Quarterly outcome series (industrial production, CPI, exchange rate, term spread, and private, public and total credit) were compiled from the Pakistan Bureau of Statistics and SBP releases. Global controls (oil and food prices, the VIX) and the US surprise series of Bolhuis, Das and Yao (2024) came from their published sources, aggregated to quarterly frequency as for the domestic shock. Estimation. Impulse responses were estimated using long difference local projections (Jorda and Taylor, 2025), controlling for one lag of the outcome, one lag of the shock, and global controls. Uniform confidence bands used a wild block bootstrap (block length 6, 5,000 replications), following Montiel Olea and Plagborg-Moller (2019) and Inoue, Jorda and Kuersteiner (2026), implemented in Python with the quarterly panel stored as an Excel workbook. Reproducibility. The package has two folders in two languages. The first holds the meeting level data, the empty projection column, and a Stata do file building the narrative shock; a user with SBP access populates the column, opens the do file, and edits the `path' local, with no other change required. The second holds the quarterly panel as an Excel file, our narrative shock series, Python scripts estimating the local projections and confidence bands, and a markdown readme; without SBP access, this folder alone reproduces the paper's results. Each readme names inputs, the line to edit, and expected outputs.
Institutions
- Loughborough UniversityEngland, Loughborough