macroshock
macroshock is a Python library for the estimation and analysis of SVAR (Structural Vector Autoregression) models aimed at applied macroeconometrics. It allows you to estimate restricted VARs, identify the structural system using four different methodologies, compute impulse-response functions (IRF), forecast error variance decomposition (FEVD), and forecasts, all with bootstrap confidence intervals and ready-to-use plotting.
Main Features
- Restricted VAR estimation (Lütkepohl-style GLS), with support for:
| Supports | Input |
|---|---|
| Constant, trend and quadratic trend | const |
| Exogenous variables | data_exogenous,variables_exogenous |
| Seasonal dummies (quarterly or monthly) | DUM |
| Linear restrictions on the coefficients | restrictions and more |
- Four structural identification methods:
| Method | Description |
|---|---|
short |
Short-run restrictions (Cholesky or custom restrictions via optimization) |
long |
Long-run restrictions (Cholesky or custom restrictions via optimization) |
signs |
Sign restrictions (orthogonal rotation sampling with acceptance/rejection) |
IV |
Instrumental variable identification (proxy-SVAR) |
- Impulse-response functions (IRF) with confidence bands.
- Forecast error variance decomposition (FEVD).
- Conditional forecasts with confidence bands.
- Bootstrap (normal or wild resampling) for inference.
- Built-in plots for IRF, forecasts and FEVD, publication-ready.
- Stability test of the system (characteristic polynomial roots).
Installation
pip install macroshock
Requirements
- Python >= 3.9
numpy >= 1.24.0pandas >= 2.0.0scipymatplotlib >= 3.7.0
Quick Start
import pandas as pd
from macroshock import SVAR
# df is a DataFrame with your series in columns
df = pd.read\_csv("my\_dataset.csv")
model = SVAR(
data=df,
variables=\["gdp", "inflation", "interest\_rate"],
lags=4,
const=1, # 0: no constant, 1: constant, 2: +trend, 3: +quadratic trend
method="short", # 'short', 'long', 'signs' or 'IV'
steps=24, # horizon for IRF/FEVD
horizon=8, # forecast horizon
past=12, # past periods to show in the forecast
alpha=32, # confidence interval level (32 -> 68% central band)
reps=1000, # bootstrap replications
)
# Runs the complete pipeline: Y/X construction, VAR estimation,
# structural identification and results summary
model.run(showIR=True, showFC=True, showVD=True)
# System stability check
model.stability(show=True)
Step-by-step Usage
The model.run() method executes the next python method secuence in a single line:
model.YX() # builds Y, X and the restriction matrix R
model.VAR() # estimates the reduced-form VAR and the companion matrix F
model.S() # identifies the structural system (matrix B)
model.summary() # prints the estimation summary
# Impulse response to shock 1, with "pair" bands (low/high percentile)
ir = model.ImpulseResponse(1, show=True, bands="pair")
# Forecast with "many" bands (multiple stacked confidence levels)
fc = model.Forecast(show=True, bands="many")
# Forecast error variance decomposition for variable 1
vd = model.VarianceDecomp(1, show=True)
Each of these methods returns a dictionary with the point estimate and confidence bands at different levels ('point', 'low', 'high', 2.5, 5, 95, 97.5, 0.5, 99.5), indexed by variable name.
GLS Restrictions
The GLS methodology can be input in four ways depending on the level of depth one needs the restrictions to be. In this section the example of restrictions will be a bivariate VAR with lags = 2 and variables = [ y , u ]. The restrictions are $a^{(1)}_{12}=a^{(2)}_{12}=0$, equal to saying y is an $AR(2)$ process or that u does not impact yinside the autoregression.
Standard [standard]
Standard notation with $\beta$ being the vector of all the regressors, matrix $C$ (c_matrix) being the coefficients of the linear combination and $c$ (c_vector) being the intercept of the system. The system of constraint is given by $C\beta=c$.
c_matrix = [[0,0,0,0,1,0,0,0,0,0],
[0,0,0,0,0,0,0,0,1,0]]
c_vector = [0,
0]
modelo = SVAR(data2, variables, lags,
restrictions='standard',
c_matrix=c_matrix,
c_vector=c_vector,
)
modelo.run()
Custom Zero Restrictions [zeros custom]
This notation uses a n_vars$\times$n_cols matrix (r_matrix). If r_matrix[i][j]=0 the jth regressor of the ith equation has a zero-constraint parameter and if r_matrix[i][j]=1 then the same regressor has a free parameter.
r_matrix = [[1,1,0,1,0],
[1,1,1,1,1]]
modelo = SVAR(data, variables, lags,
restrictions='zeros custom',
r_matrix=r_matrix,
)
modelo.run()
Absolut Zero Restrictions [zeros]
This notation uses a n_vars$\times$n_vars matrix (r_matrix). If r_matrix[i][j]=0 all of the jth variable lagged autoregressive parameters are zero inside the ith equation and if r_matrix[i][j]=1 then all lagged coefficients of the jth variable are free for the ith equation.
r_matrix = [[1,0],
[1,1]]
modelo = SVAR(data, variables, lags,
restrictions='zeros',
r_matrix=r_matrix,
)
modelo.run()
Custom [custom]
This notation follows Lutkepohl (2005) p.193-197 system $(R\gamma=r)$, so r_matrix and r_vector follow $R$ and $r$ definitions shown in the book.
modelo = SVAR(data, variables, lags,
restrictions='custom',
r_matrix=r_matrix,
r_vector=r_vector,
)
modelo.run()
Structural Identification
1. Short-run (Cholesky / custom restrictions)
model = SVAR(data=df, variables=variables, lags=4, method="short")
By default it uses the Cholesky decomposition. If matrix\_short is passed (a binary matrix of 1s and 0s), the B matrix satisfying the indicated zero restrictions is solved numerically.
2. Long-run (Blanchard-Quah)
model = SVAR(data=df, variables=variables, lags=4, method="long")
By default it uses the Cholesky decomposition. If matrix\_long is passed (a binary matrix of 1s and 0s), the B matrix satisfying the indicated zero restrictions is solved numerically.
3. Sign restrictions
matrix\_signs = \[\[1, -1, 0],
\[1, 1, 0],
\[0, 0, 1]]
model = SVAR(
data=df, variables=variables, lags=4,
method="signs",
matrix\_signs=matrix\_signs,
steps\_signs=4, # horizons over which signs are checked
reps=1000, # number of accepted B matrices
)
4. Instrumental variables (Proxy-SVAR)
model = SVAR(
data=df, variables=variables, lags=4,
method="IV",
iv=\["instrument\_1"], # DataFrame column(s) with the instrument(s)
)
Package Structure
macroshock/
├── data.py # construction of Y, X and restriction matrix R
├── var.py # reduced-form VAR estimation (restricted GLS, IC, t-stats)
├── stats.py # statistical utilities (Student's t CDF and p-values)
├── identification.py # the 4 structural identification methods
├── irf.py # impulse-response functions and bands from bootstrap
├── fevd.py # forecast error variance decomposition
├── fc.py # forecast computation
├── bootstrap.py # bootstrap (normal / wild) for IR, FC and FEVD
├── plotting.py # IRF, forecast and FEVD plots
└── macroshock.py # main SVAR class (and SVEC, LP scaffolds for future development)
Roadmap
- Full documentation (Sphinx / ReadTheDocs)
- SVEC model (structural Vector Error Correction)
- Local Projections (LP)
- Historical decomposition (
HistoricalDecomp) - Unit tests and CI
Contributing
Issues and pull requests are welcome. If you find a bug or have an improvement proposal, open an issue in the project's repository.
License
This project is distributed under the MIT license. See the LICENSE file for more details.
Metadata
Release files for macroshock 1.0.7
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| macroshock-1.0.7.tar.gz | 39.5 kB | Without provenance | Details for macroshock-1.0.7.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| macroshock-1.0.7-py3-none-any.whl | Python 3 | none | any | Details for macroshock-1.0.7-py3-none-any.whl |
Total release size: 79.9 kB
Release files / macroshock-1.0.7.tar.gz
| Download URL | macroshock-1.0.7.tar.gz |
|---|---|
| Size | 39.5 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
6db2bf0f1d7da7b212314bf360b2405a37a6abde97c074c08125de07829e08bc
|
|
BLAKE2b-256 checksum How to use checksums |
ed8596df05b86f592ce3364bb9a64d8bff54e86cb9da38ff3981133369915367
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.13.9
|
Release files / macroshock-1.0.7-py3-none-any.whl
| Download URL | macroshock-1.0.7-py3-none-any.whl |
|---|---|
| Size | 40.3 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
c4d513566f23cc43429149440326a20bc5104753c6090768bdbd2bb8053dda18
|
|
BLAKE2b-256 checksum How to use checksums |
28f4903451f57b6519bd398f7716f59cd3a5ee7609b060ad6b99f5455304fee7
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.13.9
|
Security for macroshock 1.0.7
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.