Cookies sur Macro by Mark.Nous utilisons des cookies essentiels pour la connexion, la langue et la sécurité. Avec votre accord, nous ajoutons des cookies optionnels de préférences, d’analyse et de mesure publicitaire avec le pixel Meta (Facebook). Lisez notre politique relative aux cookies.
Vector autoregression - a small system where each macro variable is regressed on lags of itself and the other variables in the system.
Graph workspace
The canonical VAR visualization is a panel of impulse response functions. Each subplot traces how one variable responds over time to a one-standard-deviation shock to another variable. The IRF panel directly communicates the dynamic interactions that are the VAR's core output - more informative than a forecast cone for a multivariate model where the cross-variable dynamics are the point.
Peak response of var 1 to shock 2
The maximum effect of a one-standard-deviation shock to variable 2 on variable 1. Captures the strength of the transmission channel.
0.2862
units of variable 1
Half-life of own shock (var 1)
How many periods until the effect of a shock to variable 1 on itself decays to half its impact value, searched past the 24 plotted quarters if needed. Blank for a unit-root or explosive (largest eigenvalue modulus at least 1), even when the response dips below half on its way out, and when it never halves.
5.00
periods
FEVD: shock 2 share of var 1 at h=8
What fraction of variable 1's 8-step forecast uncertainty is due to shocks originating in variable 2.
10.70%
percent
Largest eigenvalue of A1
Governs the overall persistence of the system. Must be below 1 for stationarity. Closer to 1 means slower decay of all IRFs.
0.8732
dimensionless
Own-lag persistence (var 1): 0.8
How long a hit to output keeps mattering; higher settings make it fade more slowly.
Higher : the diagonal decays more slowly. The off-diagonal IRFs from shocks to variable 1 also persist longer.
Cross-lag (var 2 → var 1): 0.2
How much last period's inflation predicts output; positive means higher inflation predicts higher output.
Non-zero activates the off-diagonal : a shock to inflation generates a response in output. Also makes the Granger causality test from inflation to output significant.
Cross-lag (var 1 → var 2): 0.1
How much last period's output predicts inflation; positive means higher output predicts higher inflation.
Non-zero activates : a shock to output causes a response in inflation. The feedback channel between the two variables becomes visible.
Own-lag persistence (var 2): 0.6
How long a jump in inflation keeps mattering; higher settings make it fade more slowly.
Higher : decays more slowly. Combined with non-zero , high can create a hump-shaped response in output to an inflation shock.
Innovation correlation: 0.3
How much shocks to output and inflation hit at the same moment.
Higher makes the Cholesky-identified IRFs more sensitive to variable ordering. At , ordering does not matter. At , swapping the order changes IRF shapes dramatically.