Local Projections & Geopolitical Shocks Lab 1
← Teaching Lab
Explore · Estimate · Interpret · Export

GeoRisk Data Lab 1

This interactive lab teaches you Local Projections (Jordà 2005) using real Geopolitical Risk data (Caldara & Iacoviello 2022). You will explore the GPR Index, understand the LP estimator, interpret impulse response functions, and produce an analysis report—all in the browser.

1. ContextThe paper, the method, the key equation.
2. ExploreInteractive GPR data with event annotations.
3. EstimateLP impulse responses on financial variables.
4. ExportSummary, scoring, Markdown & R script.
The Paper

Caldara & Iacoviello (2022): Measuring Geopolitical Risk

Published in the American Economic Review 112(4), pp. 1194–1225, this paper constructs a monthly index of geopolitical risk (GPR) by counting the share of newspaper articles in major outlets that discuss geopolitical threats and events. The index covers 1900 to the present and decomposes into two sub-indices:

GPR Threats (GPRT)

Articles mentioning risks of war, terrorism, geopolitical tensions—capturing ex ante uncertainty about future adverse events.

GPR Acts (GPRA)

Articles reporting actual geopolitical events: military actions, terrorist attacks, escalations—measuring realized geopolitical shocks.

The key finding: an increase in the GPR index causes a decline in investment, employment, and stock market returns, while raising oil prices and capital flows toward safe-haven economies. The effect is economically significant and robust across identification strategies.

The Method

Local Projections (Jordà 2005)

Introduced in Econometrica 73(1), pp. 161–182, Local Projections (LPs) offer a flexible, robust alternative to Vector Autoregressions (VARs) for estimating impulse response functions (IRFs). The core idea is remarkably simple:

“Instead of iterating a VAR forward, estimate a separate regression for each horizon h. Each regression is just OLS.” — Jordà (2005)
yt+h = αh + βh · GPRt + Γh · Xt + εt+h     for h = 0, 1, 2, …, H

The sequence of coefficients {β0, β1, …, βH} traces out the impulse response function. At each horizon h, you run a different OLS regression where the dependent variable is y shifted forward by h periods. The shock variable (GPRt) stays at date t.

Why not a VAR?

VARs impose the same dynamic structure at every horizon. LPs are robust to misspecification of the lag structure, naturally handle nonlinearities, and provide direct standard errors.

Controls Xt

Typically include lags of the dependent variable, lags of GPR, and other macro controls (interest rates, oil prices, VIX) to isolate the causal effect of the geopolitical shock.

Inference

Standard errors must be corrected for serial correlation in the residuals (Newey-West HAC), since the projection residuals at horizon h > 0 are mechanically autocorrelated.

Identification

The GPR index captures exogenous variation in geopolitical risk because it is constructed from newspaper text, not from economic data. This provides a narrative identification strategy.

Key References
  • Jordà, Ò. (2005). “Estimation and Inference of Impulse Responses by Local Projections.” American Economic Review, 95(1), 161–182.
  • Caldara, D. & Iacoviello, M. (2022). “Measuring Geopolitical Risk.” American Economic Review, 112(4), 1194–1225.
  • Bloom, N. (2009). “The Impact of Uncertainty Shocks.” Econometrica, 77(3), 623–685.
  • Ramey, V.A. (2016). “Macroeconomic Shocks and Their Propagation.” Handbook of Macroeconomics, Vol. 2.
Interactive Data

GPR Index: January 2000 – December 2024

The chart below plots the monthly Geopolitical Risk Index (global) alongside its two components: Threats and Acts. Toggle the series to compare them. Click on any annotated event marker to see details about the geopolitical spike.

Observations. The GPR index captures distinct geopolitical regimes. The post-9/11 period (2001–2003) shows persistently elevated threat levels. The 2022 spike following Russia’s invasion of Ukraine is the largest since the Iraq War. Notice how Threats and Acts often diverge: a buildup of threats does not always culminate in acts (e.g., North Korea 2017), while some events arrive with little warning in the threat index.

Descriptive Statistics

Summary Statistics & Distribution

Mean GPR (2000–2024)
Standard Deviation
Maximum (month)
Skewness
Local Projection Results

Impulse Response Functions: GPR → Financial Variables

Below are pre-computed LP estimates of the effect of a one-standard-deviation GPR shock on four financial variables, at horizons h = 0 to 12 months. Select a dependent variable to display its IRF with 68% and 90% confidence bands (Newey-West HAC standard errors). Controls include 4 lags of the dependent variable, 4 lags of GPR, the federal funds rate, and the log of oil prices.

Comprehension Check

Interpret the Impulse Responses

Answer the following questions based on the IRFs displayed above. Your score is tracked in the status panel.

Q1. For the S&P 500, at what horizon is the negative effect of a GPR shock largest?
Q2. Does the VIX response to a GPR shock persist beyond 3 months?
Q3. Why do Local Projections use separate OLS regressions at each horizon instead of iterating a VAR?
Q4. The EUR/USD response to a GPR shock is negative. What does this mean economically?
Q5. The oil price response is positive on impact but the 90% confidence band includes zero by h = 4. What is the correct interpretation?
Score: 0 / 5
Summary of Findings

What We Learned

Using Local Projections on monthly data (2000–2024), we estimated the dynamic causal effect of geopolitical risk shocks on four financial variables. The results are broadly consistent with Caldara & Iacoviello (2022):

S&P 500
Negative on impact (β0 ≈ −1.2%), partial reversion by h = 6. Significant at 90% for h ≤ 3.
VIX
Sharp spike (+2.8 pts on impact), rapid decay. Insignificant after h = 3. Classic uncertainty shock pattern (Bloom 2009).
Oil Price
Positive on impact (+1.8%), transitory. Reflects supply-disruption fears rather than persistent demand effects.
EUR/USD
Negative (−0.6%), persistent for ~4 months. Consistent with safe-haven flows to USD assets.
“Geopolitical risk shocks cause statistically significant declines in investment, increases in uncertainty, and reallocation toward safe-haven assets.” — Caldara & Iacoviello (2022, p. 1196)
Your Performance

Lab Score

Quiz Score
0 / 5
Phases Completed
0 / 4

Complete all phases and quiz questions to receive your assessment.

Export

Markdown Report

Click the button below to generate a Markdown report summarizing your lab session. You can copy it to your clipboard or download it as a file.

Report will appear here after generation.
Replication

R Script

The following R script replicates the Local Projection exercise using the lpirfs package and actual GPR data. Download it and run it in RStudio.