Panel Toda-Yamamoto Causality Test
Panel Toda-Yamamoto Granger Non-Causality Test · Also known as: Panel TY causality test, Toda-Yamamoto panel causality, panel modified Wald causality test, panel MWALD causality
The Panel Toda-Yamamoto (PTY) causality test extends the Toda-Yamamoto modified Wald approach to panel data, allowing researchers to test Granger non-causality across multiple cross-sectional units without requiring pre-testing for cointegration or imposing a common causality direction on all units.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use the Panel Toda-Yamamoto test when your panel contains integrated or possibly cointegrated time series and you want to test directional causality without the fragility of a two-step cointegration pre-test. It suits macro panels with moderate to long time dimensions (T >= 20) and a small to moderate number of cross-sections (N up to ~30). It is particularly appropriate when units may differ in their causality relationships — the test allows heterogeneous causal patterns. Do not use it when T is very short (T < 15), as the augmented VAR over-fits severely; prefer difference-GMM-based causality tests in that case. Also avoid it when all series are clearly I(0), where a standard panel Granger test is simpler.
Strengths & limitations
- No pre-test for cointegration is required: the MWALD statistic is chi-squared regardless of integration order.
- Valid for I(0), I(1), and I(2) processes simultaneously, giving robustness against misclassification of integration order.
- Allows heterogeneous causality patterns across panel units rather than forcing a single pooled direction.
- Straightforward to implement using SUR estimation or unit-by-unit OLS with bootstrap critical values.
- Requires a moderately long time dimension; with very short T the augmented VAR is poorly identified.
- Adding d_max extra lags reduces degrees of freedom and reduces power, especially in small samples.
- Unit-by-unit inference loses panel efficiency gains if heterogeneity is limited; pooled approaches may be preferred when causality is homogeneous.
- Cross-sectional dependence in the panel must be explicitly handled — ignoring it distorts size.
Frequently asked
Do I need to test for cointegration before running the Panel Toda-Yamamoto test?
No. That is the main advantage of the Toda-Yamamoto approach: by augmenting the VAR with d_max extra lags the MWALD statistic follows a chi-squared distribution regardless of whether the series are cointegrated or not, so no cointegration pre-test is needed.
How do I choose p and d_max?
Select p, the optimal VAR lag order, using AIC or BIC for each cross-section. Set d_max to the highest integration order found across the panel — typically 1, and rarely 2 for economic time series. The total lag length in each augmented VAR is then p + d_max.
What if causality differs across panel units?
The panel Toda-Yamamoto approach naturally accommodates this. Test each unit individually and inspect unit-level p-values. You can then summarise whether causality is found in a majority of units or only in specific subgroups, rather than imposing a single pooled conclusion.
How does this differ from the standard panel Granger causality test?
A standard panel Granger causality test uses first-differenced or stationary data and risks size distortion if integration or cointegration is misspecified. The Toda-Yamamoto variant operates on levels with augmented lags, sidestepping that fragility at the cost of slightly lower power due to the additional lags.
Is the Panel Toda-Yamamoto test valid with cross-sectional dependence?
Not automatically. When cross-sections are dependent (e.g., countries linked by trade or financial flows), standard chi-squared critical values are distorted. Use the Konya (2006) bootstrap-SUR procedure or a CD-robust variant to obtain correct critical values.
Sources
- Toda, H. Y., & Yamamoto, T. (1995). Statistical inference in vector autoregressions with possibly integrated processes. Journal of Econometrics, 66(1-2), 225-250. DOI: 10.1016/0304-4076(94)01616-8 ↗
- Konya, L. (2006). Exports and growth: Granger causality analysis on OECD countries with a panel data approach. Economic Modelling, 23(6), 978-992. DOI: 10.1016/j.econmod.2006.04.008 ↗
How to cite this page
ScholarGate. (2026, June 3). Panel Toda-Yamamoto Granger Non-Causality Test. ScholarGate. https://scholargate.app/en/econometrics/panel-toda-yamamoto-causality
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Granger Causality TestEconometrics↔ compare
- Panel Granger CausalityEconometrics↔ compare
- Panel Johansen CointegrationEconometrics↔ compare
- Panel VECMEconometrics↔ compare
- Toda-Yamamoto causality testEconometrics↔ compare