Synthetic Control Method

The core mechanism involves solving a constrained optimization problem to find a weighted average of donor pool units that minimizes the pre-intervention prediction error. The resulting synthetic control serves as the counterfactual—what would have happened absent the intervention. The causal effect is the post-intervention divergence between the actual treated unit and its synthetic twin.

• Weights are typically non-negative and sum to one
• Pre-intervention outcome and predictor variables are matched
• Extends traditional difference-in-differences by relaxing parallel trends assumptions

Because the method typically involves a single treated unit, standard large-sample inference is unavailable. Instead, placebo tests (or permutation tests) are used: the synthetic control method is applied iteratively to every untreated unit in the donor pool as if it were treated. The magnitude of the estimated effect for the actual treated unit is then ranked against this distribution of placebo effects to calculate a pseudo p-value.

• In-space placebos: apply treatment to each control unit
• In-time placebos: shift the intervention date to a pre-treatment period
• Leave-one-out robustness checks validate sensitivity to donor composition

The validity of the synthetic control hinges on the donor pool—the set of untreated units used to construct the counterfactual. Units must not be affected by the intervention (no spillover effects) and should share similar structural characteristics. The donor pool must be sufficiently large to allow a good pre-intervention fit, but restricted enough to avoid overfitting.

• Exclude units exposed to similar interventions
• Include units with comparable institutional or operational contexts
• Use predictor variables that are unaffected by the intervention itself

In supply chain contexts, the method is used to quantify the impact of a localized disruption—such as a port closure, factory fire, or supplier bankruptcy—on a specific metric like lead time or cost. The treated unit is the disrupted node; the synthetic control is built from similar, unaffected nodes. This isolates the disruption's causal effect from broader market trends.

• Example: Estimating the cost impact of a single supplier failure
• Controls for seasonality and macroeconomic fluctuations
• Provides a counterfactual for root cause identification engines

The method requires a sufficiently long pre-intervention period to establish a good fit. Poor pre-treatment fit indicates the synthetic control is not a credible counterfactual. The method also assumes the intervention has no effect on donor units and that the relationship between predictors and outcomes remains stable over time.

• Pre-intervention RMSE should be close to zero
• Sensitivity analyses test robustness to weight restrictions
• Not suitable for interventions affecting the entire system simultaneously

While the Synthetic Control Method is often presented as a purely data-driven technique, it can be formalized within the Structural Causal Model (SCM) framework. The weights implicitly encode assumptions about the data-generating process. Integrating SCM logic allows analysts to explicitly model latent confounders and test the sensitivity of results to violations of causal assumptions.

• Bridges reduced-form and structural approaches
• Complements Directed Acyclic Graphs for assumption mapping
• Enables formal counterfactual reasoning about the disruption

A quasi-experimental design that compares the change in outcomes over time between a treated unit and an untreated control group. Unlike the Synthetic Control Method, which constructs a weighted counterfactual, DiD relies on the parallel trends assumption—the idea that treated and control units would have followed similar trajectories absent the intervention. It is widely used in policy evaluation and supply chain disruption analysis.

A formal framework defining causal relationships using endogenous variables, exogenous noise terms, and structural equations. SCMs represent the data-generating mechanism of a system, enabling counterfactual reasoning. In supply chain contexts, an SCM can model how a port closure (intervention) propagates through inventory levels and delivery times, providing the theoretical backbone for methods like the Synthetic Control.

The process of estimating what would have happened to an outcome if a specific treatment had been different, given observed factual data. The Synthetic Control Method is a prime tool for counterfactual reasoning in comparative case studies. In supply chains, this answers questions like: 'What would our on-time delivery rate have been if we hadn't switched to this new logistics provider?'

A set of three inference rules developed by Judea Pearl for transforming interventional distributions (do-operator) into observational distributions. This mathematical toolkit allows analysts to determine if a causal effect can be estimated from non-experimental data and which variables must be adjusted for. It provides the formal identification logic that underpins methods like backdoor adjustment and front-door criterion.

An adaptation of the random forest algorithm designed to estimate heterogeneous treatment effects. It recursively partitions the feature space to identify subgroups with different causal responses. In supply chain risk management, a causal forest can reveal that a supplier diversification policy reduces disruption impact only for high-volume, low-complexity product categories, enabling precise targeting.