EconBase
← Back to paper

Uncertainty Quantification in Synthetic Controls with Staggered Treatment Adoption

Extracted main text — title through conclusion, appendix excluded. This is what our citation measures are computed over, published so the extraction can be checked by eye.

105,972 characters · 23 sections · 19 citation commands

Rendered from LaTeX for readability, not typeset faithfully. Citation keys are highlighted; maths is left as source; figures, tables and equation environments are summarised rather than reproduced; unrecognised commands are greyed out so nothing is silently dropped. Email addresses are removed.

Supplement to “Uncertainty Quantification in Synthetic Controls with Staggered Treatment Adoption”

abstractThis supplement contains all proofs, additional results, other technical details about data preparation and estimation, and the code used to reproduce the results.

Notation

We summarize the notation used throughout the paper in the following table.

table[table omitted — 2,747 chars of source]

Other Strategies for Uncertainty Quantification

Out-of-Sample Error

In Section 4.2 we discuss the approach for quantifying the out-of-sample uncertainty based on the non-asymptotic bounds. We briefly describe two strategies below.

itemize[leftmargin=*] • Location-scale model. Suppose that $u_{it}=\mathbb{E}[u_{it}|\mathscr{H}]+(\mathbb{V}[u_{it}|\mathscr{H}])^{1/2}\nu_{it}$ with $\nu_{it}$ statistically independent of $\mathscr{H}$. The bounds on $u_{it}$ can now be set to $\underline{M\mkern-4mu}\mkern4mu _{\mathtt{out}}=\mathbb{E}[u_{it}|\mathscr{H}]+(\mathbb{V}[u_{it}|\mathscr{H}])^{1/2}\mathfrak{c}_\nu(\alpha_{\mathtt{out}}/2)$ and $\overline{\mkern-3muM\mkern-1.5mu}_{\mathtt{out}}=\mathbb{E}[u_{it}|\mathscr{H}]+(\mathbb{V}[u_{it}|\mathscr{H}])^{1/2}\mathfrak{c}_\nu(1-\alpha_{\mathtt{out}}/2)$ where $\mathfrak{c}_\nu(\alpha_{\mathtt{out}}/2)$ and $\mathfrak{c}_\nu(1-\alpha_{\mathtt{out}}/2)$ are $\alpha_{\mathtt{out}}/2$ and $(1-\alpha_{\mathtt{out}}/2)$ quantiles of $\nu_{it}$, respectively, and $\alpha_{\mathtt{out}}$ is the desired pre-specified level. • Quantile regression. We can determine the $\alpha_{\mathtt{out}}/2$ and $(1-\alpha_{\mathtt{out}}/2)$ conditional quantiles of $u_{it}|\mathscr{H}$. Consequently, another possibility is to employ quantile regression methods to estimate those quantities using pre-treatment data.

Simultaneous Prediction Intervals

Section 4.4 constructs prediction intervals with simultaneous coverage. We briefly describe two other common approaches below.

itemize[leftmargin=*] • Bonferroni-type correction. There is a large literature on Bonferroni corrections that can be used to construct multiple prediction intervals with simultaneous coverage. For example, consider a simple correction strategy: for each $k=0, \ldots, L$, use any strategy described in Section 4.2 or Section (ref) to construct a prediction interval for $u_{i(T_i+k)}$ that has a coverage probability at least $1-(\alpha_{\mathtt{out}}/(L+1))$. Then, the simultaneous coverage probability of the $L+1$ prediction intervals $\{\tilde{\mathcal{I}}_{k}: 0\leq k\leq L\}$ is at least $1-\alpha_{\mathtt{out}}$. Some other more sophisticated corrections are also available in the literature (see, e.g., Ravishanker-Hochberg-Melnick_1987_TCMA). For instance, the second-order Bonferroni-type bound implies that \[ \begin{split} &\mathbb{P}\big[u_{i(T_i+k)}\in\tilde{\mathcal{I}}_{k} \text{ for all } 0\leq k\leq L\,\big|\, \mathscr{H}\big]\geq 1-\sum_{k=0}^L p_k+\sum_{k=0}^{L-1} p_{k,k+1},\quad \text{where}\\ &p_k=\mathbb{P}(u_{i(T_i+k)}\in\tilde{\mathcal{I}}_{k}|\mathscr{H}), \quad p_{k,k+1}=\mathbb{P}(u_{i(T_i+k)}\in\tilde{\mathcal{I}}_{k},\; u_{i(T_i+k+1)}\in\tilde{\mathcal{I}}_{k+1}|\mathscr{H}). \end{split} \] Then, one can construct the prediction intervals $\tilde{\mathcal{I}}_{k}$'s with corresponding coverage probabilities $p_k$ and $p_{k,k+1}$ such that $1-\sum_{k=0}^L p_k+\sum_{k=0}^{L-1} p_{k,k+1}\geq 1-\alpha_{\mathtt{out}}$. Such bounds are usually sharper, but their implementation requires the modeling of the dependence of $(u_{it}, u_{i(t+1)})$ conditional on $\mathscr{H}$ and is computationally more burdensome. • Scheff\'{e}-type intervals. An alternative approach is to construct Scheff\'{e}-type simultaneous prediction intervals, though stronger distributional assumptions need to be made. For instance, assume that $(u_{iT_i}, \cdots, u_{i(T_i+L)})'$ jointly follows a conditional Gaussian distribution with mean zero and variance $\Sigma_{\mathscr{H}}$. Then, $$(u_{iT_i}, \cdots, u_{i(T_i+L)})\Sigma_{\mathscr{H}}^{-1}(u_{iT_i}, \cdots, u_{i(T_i+L)})'\sim\chi^2_{L+1},$$ where $\chi^2_{L+1}$ is $\chi^2$ distribution with $L+1$ degrees of freedom. The sequence of prediction intervals $\tilde{\mathcal{I}}_k=\Big[-\sigma_{\mathscr{H},kk}\sqrt{\chi^2_{L+1}(1-\alpha_{\mathtt{out}})},\; \sigma_{\mathscr{H},kk}\sqrt{\chi^2_{L+1}(1-\alpha_{\mathtt{out}})}\,\Big]$ have the simultaneous coverage probability at least $1-\alpha_{\mathtt{out}}$, where $\sigma^2_{\mathscr{H},kk}$ is the $k$-th diagonal element of $\Sigma_{\mathscr{H}}$ and $\chi^2_{L+1}(1-\alpha_{\mathtt{out}})$ is the $(1-\alpha_{\mathtt{out}})$-quantile of $\chi^2$ distribution with $L+1$ degrees of freedom.

Alternative Bounds for In-sample Error

When the causal predictand of interest depends on treatment effects on multiple treated units, such as the time-specific unit-averaged predictand $\tau_{\mathcal{Q}k}$, there is an alternative method for in-sample uncertainty quantification that may yield tighter prediction intervals. It relies on the fact that if the SC weights $\widehat{\boldsymbol{\beta}}^{[i]}$ for each ever-treated unit $i$ are obtained through a separate optimization process using data $\mathbf{A}^{[i]}$, $\mathbf{B}^{[i]}$ and $\mathbf{C}^{[i]}$, then, as outlined in Remark 1, a sequence of restrictions must be obeyed: for each $i\in\mathcal{E}$,

equation[equation omitted — 433 chars of source]

Then, analogously to (6.5) in the main text, for any predictand $\tau$ defined before, we can set

equation[equation omitted — 530 chars of source]

conditional on the data, with $\tilde{\mathcal{M}}_{\mathbf{G}}^\star=\{\boldsymbol{\delta}=(\boldsymbol{\delta}^{[1]\prime},\cdots, \boldsymbol{\delta}^{[J_1]\prime})': \; \boldsymbol{\delta}^{[i]}\in\Delta^{[i]^\star},\; \boldsymbol{\delta}^{[i]\prime}\widehat\mathbf{Q}^{[i]}\boldsymbol{\delta}^{[i]}- 2(\mathbf{G}^{[i]\star})'\boldsymbol{\delta}^{[i]}\leq 0,\; i\in\mathcal{E}\}$ and $\mathbf{G}^\star=(\mathbf{G}^{[1]\prime},\cdots, \mathbf{G}^{[J_1]\prime})'|\mathsf{Data}\sim \mathsf{N}(\mathbf{0},\widehat{\boldsymbol{\Sigma}})$. Again, $\Delta^{[i]}$ is replaced with its feasible (enlarged) version $\Delta^{[i]\star}$, and $\widehat{\boldsymbol{\gamma}}^{[i]}-\boldsymbol{\gamma}^{[i]}$ is replaced with the approximating Gaussian vector $\mathbf{G}^{[i]\star}$. As a reminder, this strategy is not equivalent to doing in-sample uncertainty quantification for each time-specific unit-specific prediction $\widehat{\tau}_{ik}$ and then combining their bounds together, since the whole vector $\mathbf{G}^\star$ has the same covariance as $\widehat{\boldsymbol{\gamma}}=(\widehat{\boldsymbol{\gamma}}^{[1]\prime}, \cdots, \widehat{\boldsymbol{\gamma}}^{[J_1]\prime})'$, thus maintaining the correlation structure among different treated units.

Technically, the above optimization procedure for constructing SC weights is equivalent to a special case of (6.1), where the weighting matrix $\mathbf{V}$ is block diagonal, taking the form $\mathbf{V}=\mathsf{diag}(\mathbf{V}^{[1]}, \cdots, \mathbf{V}^{[J_1]})$, and the feasible set $\Delta=\mathcal{W}\times \mathcal{R}$ is the Cartesian product of $\Delta^{[i]}=\mathcal{W}^{[i]}\times \mathcal{R}^{[i]}$ for each subvector $\boldsymbol{\beta}^{[i]}$ (hence, there is no cross-treated-unit constraint). The general in-sample uncertainty quantification strategy presented earlier in (6.5) of the main text relies on the fact that $(\widehat{\boldsymbol{\beta}}-\boldsymbol{\beta}_0)'\widehat{\mathbf{Q}}(\widehat{\boldsymbol{\beta}}-\boldsymbol{\beta}_0)- 2(\widehat{\boldsymbol{\gamma}}-\boldsymbol{\gamma})'(\widehat{\boldsymbol{\beta}}-\boldsymbol{\beta}_0)= \sum_{i\in\mathcal{E}}[(\widehat{\boldsymbol{\beta}}^{[i]}-\boldsymbol{\beta}_0^{[i]})'\widehat{\mathbf{Q}}^{[i]}(\widehat{\boldsymbol{\beta}}^{[i]}-\boldsymbol{\beta}_0^{[i]})- 2(\widehat{\boldsymbol{\gamma}}^{[i]}-\boldsymbol{\gamma}^{[i]})'(\widehat{\boldsymbol{\beta}}^{[i]}-\boldsymbol{\beta}_0^{[i]})]\leq 0$ and $\widehat{\boldsymbol{\beta}}-\boldsymbol{\beta}_0\in\Delta=\times_{i\in\mathcal{E}}\Delta^{[i]}$, which are immediately implied by (ref). In this sense, the restriction (ref) is stricter, making the alternative bounds in (ref) tighter (or at least no looser) than those in (6.5) in the main text. Table (ref) quantifies the gains from this restriction in terms of the length of the prediction intervals in our leading empirical application.

table[table omitted — 3,441 chars of source]

Due to the complexity of the feasibility set $\tilde{\mathcal{M}}_{\mathbf{G}}^\star$, the alternative bounds (ref) may not be theoretically justified using the same argument for (6.5) in the main text. However, other techniques, such as strong approximations, can be employed to establish the validity of (ref), albeit at the expense of additional technical complexity. We do not pursue further justification of (ref) and leave it for future research.

Linear Factor Model Justification

There are different ways to justify the synthetic control method, including the cointegrated system motivated by our empirical application. In this section, we briefly discuss an alternative justification that assumes the data are generated through a linear factor model: \[ Y_{it}(\infty)=\lambda_t\mu_i + \nu_{it}, \qquad 1\leq i\leq N,\; 1\leq t\leq T. \] For simplicity, assume that only the first unit is treated, the common factor $\lambda_t$ is i.i.d. over $1\leq t\leq T$, $\nu_{it}$ is i.i.d. across $1\leq i\leq N$ and over $1\leq t\leq T$, and $\{\lambda_t:1\leq t\leq T\}$ and $\{\nu_{it}:1\leq i\leq N, 1\leq t\leq T\}$ are independent of each other. In addition, we assume that $\mathbb{E}[\nu_{it}]=0$, the factor loadings $\{\mu_i:1\leq i\leq N\}$ are fixed, and only the pre-intervention outcomes are used in the SC construction. Then, $\mathscr{H}=\{(Y_{2t}(\infty), \cdots, Y_{Nt}(\infty)): 1\leq t\leq T\}$. Accordingly, $\mathbf{w}_0$ is given by the following expression: \[ \mathbf{w}_0=\underset{\mathbf{w}\in\mathcal{W}}{\operatorname*{arg\,min}}\; \frac{1}{T_0}\sum_{t=1}^{T_0}\mathbb{E}\Big[\Big((\mu_1-\boldsymbol{\mu}_c'\mathbf{w})\lambda_t+(\nu_{1t}-\bm{\nu}_{t,c}'\mathbf{w})\Big)^2\Big|\mathscr{H}\Big], \] where $\boldsymbol{\mu}_c=(\mu_2, \cdots, \mu_N)'$ and $\bm{\nu}_{t,c}=(\nu_{2t}, \cdots, \nu_{Nt})'$, assuming the expectation (exists and) is finite. Define $\mathscr{H}_t=\{Y_{2t}(\infty), \cdots, Y_{Nt}(\infty)\}$. Then, $\mathbf{w}_0$ can be further written as

align*[align* omitted — 504 chars of source]

assuming these expectations are finite. Given our assumptions on $\lambda_t$ and $\nu_{it}$, we expect that $$\frac{1}{T_0}\sum_{t=1}^{T_0}\mathbb{E}[\lambda_t^2|\mathscr{H}_t] \approx \mathbb{E}[\lambda_t^2], \quad \frac{1}{T_0}\sum_{t=1}^{T_0}\mathbb{E}[\bm{\nu}_{t,c}\bm{\nu}_{t,c}'|\mathscr{H}_t] \approx \mathbb{E}[\nu_{1t}^2]\mathbf{I}_{N-1}, \quad\text{and}\quad \frac{1}{T_0}\sum_{t=1}^{T_0}\mathbb{E}[\lambda_t\bm{\nu}_{t,c}'|\mathscr{H}_t]\approx \bm{0} $$ with high probability when $T_0$ is large, which can be shown under additional moment conditions. As a consequence, our expression for $\mathbf{w}_0$ is similar to that in Ferman-Pinto_2021_wp.

Then, the in-sample error for the prediction of the TSUS effect $\tau_{1k}$ is given by \[ \mathsf{InErr}(\tau_{1k})=-\mathbf{Y}_{\mathcal{N}t}'(\widehat{\mathbf{w}}-\mathbf{w}_0), \] which represents the error in estimating the weights. The out-of-sample error is given by \[ \mathsf{OutErr}(\tau_{1k})=Y_{1(T_1+k)}-\mathbf{Y}_{\mathcal{N}(T_1+k)}'\mathbf{w}_0= \nu_{1(T_1+k)}-\bm{\nu}_{T_1+k,c}'\mathbf{w}_0+ (\mu_1-\boldsymbol{\mu}_c'\mathbf{w}_0)\lambda_{T_1+k}. \] The first term $\nu_{1(T_1+k)}$ represents the treated unit's innovation in the post-treatment period $T_1+k$; the second term $\bm{\nu}_{T_1+k,c}'\mathbf{w}_0$ represents the weighted average of the innovations for the donor units in the post-treatment period $T_1+k$; and the third term $(\mu_1-\boldsymbol{\mu}_c'\mathbf{w}_0)\lambda_{T_1+k}$ can be thought of as the impact of the “bias” of the SC weights in $T_1+k$, if one thinks the target weight in this context is the one recovering the true factor loading $\mu_1$ of the treated unit. Given the conditions imposed, our method is still applicable to this scenario, and importantly, the “bias” due to $(\mu_1-\boldsymbol{\mu}_c'\mathbf{w}_0)\lambda_{T_1+k}$ is taken into account.

Proofs

Proof of Corollary 1

proofWe only need to verify the conditions in Theorem 1. The proof is divided into several steps. Step 0: We first give several useful facts. The SC construction considered in this corollary is a special case of that considered in Theorem 1, where the weighting matrix is an identity matrix, and there is only one feature to be matched (GDP per capita) and no additional covariates. As shown in the main paper, for any $\widehat\tau\in\{\widehat\tau_{ik}, \widehat\tau_{i\cdot}, \widehat\tau_{\mathcal{Q}k}, \widehat\tau_{\cdot\cdot}\}$, the in-sample error can always be expressed as $\mathbf{p}_\tau'(\widehat{\mathbf{w}}-\mathbf{w}_0)$. Note that $\Delta^{[i]}$ and $\Delta^{[i]\star}$ are convex in this case, and thus $\times_{i\in\mathcal{S}}\Delta^{[i]}$ and $\times_{i\in\mathcal{S}}\Delta^{[i]\star}$ (for some set of treated units $\mathcal{S}$) are all convex as well. For any $\kappa$, define $\mathcal{A}_\kappa= \{\boldsymbol{\xi}: \sup_{\boldsymbol{\delta}\in\mathcal{M}_{\boldsymbol{\xi}}}\mathbf{p}_\tau'\boldsymbol{\delta}\leq\kappa\}$, which is convex by Lemma 2 of Cattaneo-Feng-Titiunik_2021_JASA. Let $d_\beta=J_0J_1$, and $\mathfrak{C}_1$, $\mathfrak{C}_2$, $\cdots$ denote some constant independent of $T_0$. Throughout this proof, $\|\cdot\|$ denotes the L2 norm for vectors and the operator norm for matrices. Step 1: We want to verify condition (i) in Theorem 1. Without loss of generality, we only consider the upper bound in this step, and the lower bound follows similarly. So our goal is to bound $|\mathbb{P}(\widehat{\boldsymbol{\gamma}}-\boldsymbol{\gamma}\in\mathcal{A}_\kappa|\mathscr{H})-\mathbb{P}(\mathbf{G}\in\mathcal{A}_\kappa|\mathscr{H})|$. Note that by Assumption 1, $u_{it}$ is independent conditional on $\mathscr{H}$. To simplify expressions, assume in this proof that $\boldsymbol{\gamma}=0$, and define $\tilde{u}_{it}=u_{it}$ if $t\leq T_i-1$ and $\tilde{u}_{it}=0$ if $t>T_i-1$. So we can write $\widehat{\boldsymbol{\gamma}}=(\sum_{t=1}^{T_1-1}\mathbf{Y}_{\mathcal{N}t}'u_{1t}, \cdots, \sum_{t=1}^{T_{J_1}-1}\mathbf{Y}_{\mathcal{N}t}'u_{J_1t})'= \sum_{t=1}^{T_{J_1}}(\mathbf{I}_{J_1}\otimes\mathbf{Y}_{\mathcal{N}t})\tilde{u}_t$ where $\tilde{u}_t=(u_{1t}, \cdots, u_{J_1t})'$. Applying the Berry-Esseen theorem for convex sets Raivc_2019_Bernoulli, \[ |\mathbb{P}(\widehat{\boldsymbol{\gamma}}-\boldsymbol{\gamma}\in\mathcal{A}_\kappa|\mathscr{H})- \mathbb{P}(\mathbf{G}\in\mathcal{A}_\kappa|\mathscr{H})| \leq 42(d_\beta^{1/4}+16)\sqrt{J_1}\|\boldsymbol{\Sigma}^{-\frac{1}{2}}\|^3\sum_{t=1}^{T_{J_1}} \mathbb{E}\Big[\sum_{i=1}^{J_1}\Big\|\mathbf{Y}_{\mathcal{N}t}u_{it}\Big\|^3\Big|\mathscr{H}\Big]. \] Given condition (i) in the corollary, we only need to bound $\sum_{t=1}^{T_{J_1}} \|\mathbf{Y}_{\mathcal{N}t}\|^3$ and $\|\boldsymbol{\Sigma}^{-1}\|$. First note that \[ \frac{1}{T_0^3}\sum_{t=1}^{T_{J_1}}\|\mathbf{Y}_{\mathcal{N}t}\|^3 \leq\frac{\sqrt{J_0}}{T_0^{3/2}}\sum_{t=1}^{T_{J_1}}\sum_{j\in\mathcal{N}}|Y_{jt}/\sqrt{T_0}|^3, \] By Assumption 1, each $\mathbf{Y}_{\mathcal{N}t}$ can be understood as a multivariate partial sum process indexed by $t$. By strong approximation of partial sum processes (e.g., Chang-Park-Song_2006_JoE), \[ \mathbb{P}\Big(\max_{1\leq t\leq T_0} \|T_0^{-1/2}\mathbf{Y}_{\mathcal{N}t}-\tilde\mathbf{G}(t/T_0)\| \geq T_0^{-0.1} \Big)\leq \mathfrak{C}_1T_0^{-1}, \] where $\tilde\mathbf{G}(\cdot)$ is a $J_0$-dimensional Brownian motion on $[0,1]$ with the variance $\mathbb{E}[\mathbf{v}_t\mathbf{v}_t']$. On the other hand, it is well known that for each $1\leq j\leq J_0$, for any $m>0$, \[ \mathbb{P}(\max_{0\leq r\leq 1}|\tilde{G}_{j}(r)|>m)\leq 2\mathbb{P}(\max_{0\leq r\leq 1}\tilde{G}_{j}(r)>m)=2\mathbb{P}(|G_{j}(1)|>m), \] where $\tilde{G}_{j}(\cdot)$ is the $j$-th element of $\mathbf{G}(\cdot)$. Using the tail bound for Gaussian distributions, we can set $m=\sqrt{2\log (2J_0T_0)\sigma_{\max}^2}$ where $\sigma_{\max}^2$ is the largest variance of $\{v_{jt}:1\leq j\leq J_0\}$, which leads to $\max_{0\leq r\leq 1}|G_{j}(r)|\leq m$ with probability over $\mathscr{H}$ at least $1-(J_0T_0)^{-1}$. Therefore, \[ \frac{1}{T_0^3}\sum_{t=1}^{T_0}\|\mathbf{Y}_{\mathcal{N}t}\|^3 \leq J_0^{3/2}T_0^{-1/2}\Big(\sqrt{2\log (2J_0T_0)\sigma^2_{\max}}+ T_0^{-0.1}\Big)^3 \] with probability over $\mathscr{H}$ at least $1-T_0^{-1}-\mathfrak{C}_1T_0^{-1}$. In addition, note that $\frac{1}{T_0^3}\sum_{t=T_0+1}^{T_{J_1}}\|\mathbf{Y}_{\mathcal{N}t}\|^3\leq \frac{4}{T_0^3}\sum_{t=T_0+1}^{T_{J_1}}(\|\mathbf{Y}_{\mathcal{N}T_0}\|^3+\|\sum_{s=T_0+1}^{t}\mathbf{v}_s\|^3)$. Then, by the previous result and sub-Gaussianity of $\mathbf{v}_t$, we conclude that with probability at least $1-\mathfrak{C}_\pi T_0^{-1}$ for some constant $\mathfrak{C}_\pi>0$, \[ \frac{1}{T_0^3}\sum_{t=1}^{T_{J_1}}\|\mathbf{Y}_{\mathcal{N}t}\|^3 \leq J_0^{3/2}(T_0^{-1/2}+\mathfrak{C}_2T_0^{-3/2}+\mathfrak{C}_3)\Big(\sqrt{2\log (2J_0T_0)\sigma^2_{\max}}+ T_0^{-0.1}\Big)^3. \] Finally, we consider $\boldsymbol{\Sigma}$. By assumption in the corollary, $s_{\min}(\boldsymbol{\Sigma})\geq\eta\widehat{\mathbf{Q}}$. Recall that $\widehat{\mathbf{Q}}=\operatorname*{diag}(\widehat{\mathbf{Q}}^{[1]}, \cdots, \widehat{\mathbf{Q}}^{[J_1]})$. Again, by strong approximation used previously, with probability over $\mathscr{H}$ at least $1-\mathfrak{C}_\pi T_0^{-1}$, $$ \bigg\|\frac{1}{T_0^2}\sum_{t=1}^{T_0}\mathbf{Y}_{\mathcal{N}t}\mathbf{Y}_{\mathcal{N}t}'- \frac{1}{T_0}\sum_{t=1}^{T_0}\tilde\mathbf{G}\Big(\frac{t}{T_0}\Big)\tilde\mathbf{G}\Big(\frac{t}{T_0}\Big)'\bigg\|\leq 2\sqrt{J_0}T_0^{-0.1}(m+T_0^{-0.1}). $$ Then, by the condition in the corollary, with probability over $\mathscr{H}$ at least $1-\pi_0-\mathfrak{C}_\pi T_0^{-1}$, $s_{\min}(\widehat{\mathbf{Q}}^{[1]})\geq \frac{2}{3}(\log T_0)^{-1/5}T_0^2$ for $T_0$ large enough. For other blocks $\widehat\mathbf{Q}^{[j]}$ with $j\neq 1$, using the previous results about the bound on $\mathbf{Y}_{\mathcal{N}t}$, we have with probability $1-\mathfrak{C}_\pi T_0^{-1}$, $\|\widehat{\mathbf{Q}}^{[j]}-\widehat{\mathbf{Q}}^{[1]}\|\leq \mathfrak{C}_4(m+T_0^{-0.1})^2T_0$. Therefore, we can conclude that with probability over $\mathscr{H}$ at least $1-\pi_0-\mathfrak{C}_\pi T_0^{-1}$, $s_{\min}(\widehat{\mathbf{Q}})\geq (\log T_0)^{-1/5}T_0^2/2$ for $T_0$ large enough. Therefore, we can take $\pi_\gamma=\pi_0+\mathfrak{C}_\pi T_0^{-1}$ and $\epsilon_{\gamma}=\mathfrak{C}_\epsilon (\log T_0)^2T_0^{-1/2}$ for some non-negative finite constant $\mathfrak{C}_\epsilon$ implied by the previous calculations. Step 2: Consider condition (ii) in Theorem 1. We have the following basic inequality hold: \[ \lambda_{\min}(\mathfrak{D}_T^{-1}\widehat{\mathbf{Q}}\mathfrak{D}_T^{-1})\|\mathfrak{D}_T\boldsymbol{\delta}\|^2\leq \boldsymbol{\delta}\widehat{\mathbf{Q}}\boldsymbol{\delta}\leq 2\mathbf{G}'\boldsymbol{\delta}\leq 2\|\mathfrak{D}_T^{-1}\mathbf{G}\|\|\mathfrak{D}_T\boldsymbol{\delta}\|. \] where $\mathfrak{D}_T=\operatorname*{diag}((T_1-1)\mathbf{I}_{T_1-1}, \cdots, (T_{J_1}-1)\mathbf{I}_{T_{J_1}-1})$. By condition (ii) imposed in the corollary and the argument given in Step 1, $s_{\min}(\mathfrak{D}_T^{-1}\widehat{\mathbf{Q}}\mathfrak{D}_T^{-1})\geq(\log T_0)^{-1/5}/2$ and $\lambda_{\max}(\mathfrak{D}_T^{-1}\widehat{\mathbf{Q}}\mathfrak{D}_T^{-1})\leq 2(\log T_0)^{1/5}$, with probability over $\mathscr{H}$ at least $1-\pi_\gamma$. Then, by the Gaussian tail bound, we can take $\varpi_{\delta}^{\star}=\mathfrak{C}_5(\log T_0)^{0.9}/T_0$, $\pi_\delta^{\star}=\pi_\gamma$, and $\epsilon_{\delta}^{\star}=T_0^{-1}$. Step 3: Consider condition (iii) in Theorem 1. Given the specific choices of $\varrho_\ell^{[i]}$'s and the argument in Step 2, the conditions in Lemma 1 are satisfied for $T_0$ large enough, and we can differentiate the binding and nonbinding constraints with high probability. For each treated unit $i$, an L1-L2 constraint is imposed on the corresponding SC weights. The set $\Delta^{[i]}$ then can be written as $\Delta^{[i]}=\Delta_{\tt l}^{[i]}\cap \Delta_{\tt nl}^{[i]}$, where $\Delta_{\tt l}^{[i]}$ and $\Delta_{\tt nl}^{[i]}$ denote the feasibility sets defined by the L1 (simplex) and L2 (ridge) constraints respectively. Similarly, $\widehat{\Delta}$ in (6.7) for this special case can be written as $\widehat{\Delta}^{[i]}=\widehat{\Delta}_{\tt l}^{[i]}\cap \widehat{\Delta}_{\tt nl}^{[i]}$. The L1 constraint is linear, and thus by Lemma 1, $\Delta^{[i]}_{\tt l}\cap\mathcal{B}(\bm{0},\varpi_\delta^\star)\subseteq\widehat{\Delta}_{\tt l}^{[i]}$. For the L2 constraint, Lemma 1 implies that $\Delta^{[i]}_{\tt nl}\cap\mathcal{B}(\bm{0},\varpi_\delta^\star)\subseteq\{\boldsymbol{\delta}:\mathrm{dist}(\boldsymbol{\delta}, \widehat{\Delta}_{\tt nl}^{[i]}\leq \mathfrak{C}(\varpi_\delta^\star)^2)\}$ for some constant $\mathfrak{C}>0$. Then, by the specified adjustment for the L2 constraint, condition (iii) holds for every $\Delta^{[i]\star}$ as well as $\times_{i\in\mathcal{S}}\Delta^{[i]\star}$, for $\mathcal{S}=\mathcal{Q}$ or $\mathcal{E}$. In this case, we can set $\pi_\Delta^\star=\pi_\gamma$ and $\epsilon_\Delta^\star=T_0^{-1}+\epsilon_\gamma$. Step 4: Finally, consider condition (iv) in Theorem 1. Note that \[ \operatorname*{tr}\Big[(\boldsymbol{\Sigma}^{-1/2}\widehat{\boldsymbol{\Sigma}}\boldsymbol{\Sigma}^{-1/2}-\mathbf{I}_{d_\beta})^2\Big] \leq d_\beta s_{\min}(\boldsymbol{\Sigma})^{-2}\|\widehat{\boldsymbol{\Sigma}}-\boldsymbol{\Sigma}\|^2. \] Since $\lambda_{\min}(\boldsymbol{\Sigma})\geq \eta\lambda_{\min}(\widehat{\mathbf{Q}})\geq\eta(\log T_0)^{-1/5}T_0^2/2$, with probability over $\mathscr{H}$ at least $1-\pi_{\delta}^{\star}$, it follows that $\pi_\gamma^\star=\pi_\gamma$, $\epsilon_{\gamma,2}^\star=\epsilon_{\Sigma,2}^\star$ and $\epsilon_{\gamma,1}^\star=2\sqrt{d_\beta}\epsilon_{\Sigma,1}^\star(\log T_0)^{1/5}/(T_0^2\eta)$. Given all results above, we finally set $\pi=\pi_\gamma$ and $\epsilon=2\mathfrak{C}_\epsilon (\log T_0)^2T_0^{-1/2} +4\sqrt{d_\beta}\epsilon_{\Sigma,1}^\star(\log T_0)^{1/5}/(T_0^2\eta) +\epsilon_{\Sigma,2}^\star+3T_0^{-1}$. Then, the proof is complete.

Proof of Theorem 1

proofLet \begin{align*} &\ell(\boldsymbol{\delta})=\boldsymbol{\delta}'\widehat{\mathbf{Q}}\boldsymbol{\delta}-2\mathbf{G}'\boldsymbol{\delta}\quad with \mathbf{G}|\mathscr{H}\sim\mathsf{N}(\bm{0}, \boldsymbol{\Sigma}),\\ &\ell^\star(\boldsymbol{\delta})=\boldsymbol{\delta}'\widehat{\mathbf{Q}}\boldsymbol{\delta}-2(\mathbf{G}^\star)'\boldsymbol{\delta}\quad with \mathbf{G}^\star|\mathsf{Data}\sim\mathsf{N}(\bm{0}, \widehat\boldsymbol{\Sigma}). \end{align*} Accordingly, define \[ \begin{split} &\bar{\varsigma}^{\star}=\sup\Big\{-\mathbf{p}_{\tau}'\boldsymbol{\delta}: \boldsymbol{\delta}\in\Delta^{\star},\,\ell^{\star}(\boldsymbol{\delta})\leq 0\Big\}\\ &\bar{\varsigma}_\mathtt{r}^{\mathsf{int}}=\sup\Big\{-\mathbf{p}_{\tau}'\boldsymbol{\delta}: \boldsymbol{\delta}\in\Delta,\,\|\boldsymbol{\delta}\|\leq \varpi_{\delta}^{\star},\, \ell^{\star}(\boldsymbol{\delta})\leq 0 \Big\},\\ &\bar{\varsigma}_\mathtt{r}=\sup\Big\{-\mathbf{p}_{\tau}'\boldsymbol{\delta}: \boldsymbol{\delta}\in\Delta,\,\|\boldsymbol{\delta}\|\leq \varpi_{\delta}^{\star},\, \ell(\boldsymbol{\delta})\leq 0 \Big\}. \end{split} \] The subscript $\mathtt{r}$ indicates the quantity is a supremum over a further restricted region for $\boldsymbol{\delta}$ (due to the constraint $\|\boldsymbol{\delta}\|\leq \varpi_\delta^\star$). For any $\alpha_0\in[0,1]$, let $\bar{\mathfrak{c}}_{\mathtt{r}}(\alpha_0)$ be the $\alpha_0$-quantile of $\bar{\varsigma}_{\mathtt{r}}$ conditional on $\mathscr{H}$. Similarly, define \[ \begin{split} &\underline{\varsigma}^{\star}:=\inf\Big\{-\mathbf{p}_{\tau}'\boldsymbol{\delta}: \boldsymbol{\delta}\in\Delta^{\star},\,\ell^{\star}(\boldsymbol{\delta})\leq 0\Big\}\\ &\underline{\varsigma}_\mathtt{r}^{\mathsf{int}}=\inf\Big\{-\mathbf{p}_{\tau}'\boldsymbol{\delta}: \boldsymbol{\delta}\in\Delta,\,\|\boldsymbol{\delta}\|\leq \varpi_{\delta}^{\star},\, \ell^{\star}(\boldsymbol{\delta})\leq 0 \Big\},\\ &\underline{\varsigma}_\mathtt{r}=\inf\Big\{-\mathbf{p}_{\tau}'\boldsymbol{\delta}: \boldsymbol{\delta}\in\Delta,\,\|\boldsymbol{\delta}\|\leq \varpi_{\delta}^{\star},\, \ell(\boldsymbol{\delta})\leq 0 \Big\}. \end{split} \] Let $\underline{\mathfrak{c}}_{\mathtt{r}}(\alpha_0)$ be the $\alpha_0$-quantile of $\underline{\varsigma}_{\mathtt{r}}$ conditional on $\mathscr{H}$. Let $\mathbb{P}_1=\mathsf{N}(\bm{0},\boldsymbol{\Sigma})$ and $\mathbb{P}_2=\mathsf{N}(\bm{0},\widehat{\boldsymbol{\Sigma}})$. By condition (iv), on an event with $\mathbb{P}$-probability at least $1-\pi_\gamma^\star$, with $\mathbb{P}(\cdot|\mathscr{H})$-probability at least $1-\epsilon_{\gamma,2}^\star$, the Kullback-Leibler divergence $\mathbb{KL}(\mathbb{P}_1, \mathbb{P}_2)\leq 2(\epsilon_{\gamma,1}^\star)^2$, and by Pinsker’s inequality, this implies that for any $\kappa'\leq \kappa$, \[ |\mathbb{P}^{\star}(\bar{\varsigma}_{\mathtt{r}}^{\mathsf{int}}\leq \kappa)-\mathbb{P}^{\star}(\bar{\varsigma}_{\mathtt{r}} \leq\kappa)| \leq \epsilon_{\gamma,1}^{\star}\quad\text{and}\quad |\mathbb{P}^{\star}(\underline{\varsigma}_{\mathtt{r}}^{\mathsf{int}}\geq \kappa')-\mathbb{P}^{\star}(\underline{\varsigma}_{\mathtt{r}} \geq\kappa')| \leq \epsilon_{\gamma,1}^{\star}. \] On the other hand, note that by condition (iii), on an event with $\mathbb{P}$-probability over $\mathscr{H}$ at least $1-\pi_\Delta^\star$, with $\mathbb{P}(\cdot|\mathscr{H})$-probability at least $1-\epsilon_\Delta^\star$, the event $\{\bar{\varsigma}^{\star}\leq\kappa\}$ implies that \[ \sup\Big\{-\mathbf{p}_{\tau}'\boldsymbol{\delta}: \boldsymbol{\delta}\in\Delta,\,\|\boldsymbol{\delta}\|\leq \varpi_{\delta}^{\star},\, \ell^{\star}(\boldsymbol{\delta})\leq 0 \Big\}\leq \sup\Big\{-\mathbf{p}_{\tau}'\boldsymbol{\delta}: \boldsymbol{\delta}\in\Delta^{\star},\, \ell^{\star}(\boldsymbol{\delta})\leq 0 \Big\}\leq\kappa \] and $\{\underline{\varsigma}^{\star}\geq \kappa'\}$ implies that \[ \inf\Big\{-\mathbf{p}_{\tau}'\boldsymbol{\delta}: \boldsymbol{\delta}\in\Delta,\,\|\boldsymbol{\delta}\|\leq \varpi_{\delta}^{\star},\, \ell^{\star}(\boldsymbol{\delta})\leq 0 \Big\}\geq \inf\Big\{-\mathbf{p}_{\tau}'\boldsymbol{\delta}: \boldsymbol{\delta}\in\Delta^\star,\, \ell^{\star}(\boldsymbol{\delta})\leq 0 \Big\} \geq \kappa'. \] Therefore, \[ \mathbb{P}^{\star}(\bar{\varsigma}^{\star}\leq \kappa)\leq \mathbb{P}^{\star}(\bar{\varsigma}_{\mathtt{r}}^{\mathsf{int}}\leq \kappa)\quad \text{and}\quad \mathbb{P}^{\star}(\underline{\varsigma}^{\star}\geq \kappa')\leq \mathbb{P}^{\star}(\underline{\varsigma}_{\mathtt{r}}^{\mathsf{int}}\geq \kappa'). \] Then, by definitions of $\bar{\varsigma}^{\star}$ and $\underline{\varsigma}^{\star}$, on an event with $\mathbb{P}$-probability over $\mathscr{H}$ at least $1-\pi_\gamma^\star-\pi_\Delta^\star$, with $\mathbb{P}(\cdot|\mathscr{H})$-probability at least $1-\epsilon_{\gamma,2}^\star-\epsilon_\Delta^\star$, we have \begin{align*} &1-\alpha_{\mathtt{in}}/2\leq \mathbb{P}^{\star}\Big( \bar{\varsigma}^{\star}\leq \overline{\mkern-3muM\mkern-1.5mu}_{\mathtt{in}}(\tau)\Big) \leq \mathbb{P}^\star\Big( \bar{\varsigma}_{\mathtt{r}}\leq \overline{\mkern-3muM\mkern-1.5mu}_{\mathtt{in}}(\tau)\Big)+ \epsilon_{\gamma,1}^\star\quad and\\ &1-\alpha_{\mathtt{in}}/2\leq \mathbb{P}^{\star}\Big( \varsigma^{\star}\geq M\mkern-4mu\mkern4mu _{\mathtt{in}}(\tau)\Big) \leq \mathbb{P}^\star\Big( \varsigma_{\mathtt{r}}\geq \underline{M\mkern-4mu}\mkern4mu _{\mathtt{in}}(\tau)\Big)+ \epsilon_{\gamma,1}^\star. \end{align*} Also, by condition (ii), we have with $\mathbb{P}$-probability over $\mathscr{H}$ at least $1-\pi_\delta^\star$, $$ \bar{\mathfrak{c}}_{\mathtt{r}}(1-\alpha_{\mathtt{in}}/2-\epsilon_{\gamma,1}^\star)\geq \bar{\mathfrak{c}}(1-\alpha_{\mathtt{in}}/2-\epsilon_{\gamma,1}^\star-\epsilon_\delta^\star)\quad\text{and}\quad \underline{\mathfrak{c}}_{\mathtt{r}}(\alpha_{\mathtt{in}}/2+\epsilon_{\gamma,1}^\star)\leq \underline{\mathfrak{c}}(\alpha_{\mathtt{in}}/2+\epsilon_{\gamma,1}^\star+\epsilon_\delta^\star). $$ Using condition (i) and all results above, we conclude that with $\mathbb{P}$-probability over $\mathscr{H}$ at least $1-\pi_\gamma-\pi_\gamma^\star-\pi_\Delta^\star-\pi_\delta^\star$, \[ \begin{split} &\mathbb{P}\Big( \underline{M\mkern-4mu}\mkern4mu _{\mathtt{in}}(\tau) \leq -\mathbf{p}_{\tau}'(\widehat{\boldsymbol{\beta}}-\boldsymbol{\beta}_0) \leq \overline{\mkern-3muM\mkern-1.5mu}_{\mathtt{in}}(\tau)\Big|\mathscr{H}\Big)\\ \geq&\, \mathbb{P}\Big( \underline{\mathfrak{c}}_{\mathtt{r}}(\alpha_{\mathtt{in}}/2+\epsilon_{\gamma,1}^{\star}) \leq-\mathbf{p}_\tau'(\widehat{\boldsymbol{\beta}}-\boldsymbol{\beta}_0)\leq \bar{\mathfrak{c}}_{\mathtt{r}}(1-\alpha_{\mathtt{in}}/2-\epsilon_{\gamma,1}^{\star})\Big|\mathscr{H}\Big)-\epsilon_{\gamma,2}^{\star}-\epsilon_{\Delta}^{\star}\\ \geq&\, \mathbb{P}\Big( \underline{\mathfrak{c}}(\alpha_{\mathtt{in}}/2+\epsilon_{\gamma,1}^{\star}+ \epsilon_{\delta}^{\star})\leq -\mathbf{p}_\tau'(\widehat{\boldsymbol{\beta}}-\boldsymbol{\beta}_0)\leq \bar{\mathfrak{c}}(1-\alpha_{\mathtt{in}}/2-\epsilon_{\gamma,1}^{\star}-\epsilon_{\delta}^{\star})\Big|\mathscr{H}\Big)-\epsilon_{\gamma,2}^{\star}-\epsilon_{\Delta}^{\star}\\ \geq&\, 1-\alpha_{\mathtt{in}}-2\epsilon_{\gamma,1}^{\star}-2\epsilon_{\delta}^{\star}-\epsilon_{\gamma}-\epsilon_{\gamma,2}^{\star}-\epsilon_{\Delta}^{\star}. \end{split} \] Finally, by condition (v), we immediately have $\mathbb{P}(\underline{M\mkern-4mu}\mkern4mu _{\mathtt{out}}(\tau)\leq \mathsf{OutErr}(\tau)\leq \overline{\mkern-3muM\mkern-1.5mu}_{\mathtt{out}}(\tau))\geq 1-\alpha_{\mathtt{out}}$. Then the proof is complete.

Verification of Condition (i) in Theorem 1

As explained in the main paper, by convexity of the constraint set $\mathcal{W}\times\mathcal{R}$ and the optimality of $\widehat{\boldsymbol{\beta}}$, $$ \inf_{\boldsymbol{\delta}\in\mathcal{M}_{\widehat\boldsymbol{\gamma}-\boldsymbol{\gamma}}}-\mathbf{p}_\tau'\boldsymbol{\delta} \leq -\mathbf{p}_\tau'(\widehat\boldsymbol{\beta}-\boldsymbol{\beta}_0)\leq \sup_{\boldsymbol{\delta}\in\mathcal{M}_{\widehat\boldsymbol{\gamma}-\boldsymbol{\gamma}}}-\mathbf{p}_\tau'\boldsymbol{\delta}, $$ where $\mathcal{M}_{\widehat{\boldsymbol{\gamma}}-\boldsymbol{\gamma}}= \{\boldsymbol{\delta}\in\Delta:\boldsymbol{\delta}'\widehat{\mathbf{Q}}\boldsymbol{\delta}- 2(\widehat{\boldsymbol{\gamma}}-\boldsymbol{\gamma})'\boldsymbol{\delta}\}$. Thus, condition (i) in Theorem 1 indeed requires that $\widehat\boldsymbol{\gamma}-\boldsymbol{\gamma}$ can be approximated by a Gaussian vector $\mathbf{G}$. Corollary 1 in the paper provided a verification of this condition in the special case of cointegrated data. In this section, we provide a more general way to verify condition (i) by imposing a conditional independence assumption on the pseudo-true residuals. The extension that allows for weakly dependent errors can be established using the idea of Theorem A in Cattaneo-Feng-Titiunik_2021_JASA. For simplicity, we assume that only $T_0$ pre-treatment periods are used to obtain the weights. Also, we write $\mathbf{U}^{[i]}=(u_{it,1}, \cdots, u_{it,M})'$, which is the vector of pseudo-true residuals corresponding to the treated unit $i$.

lemAssume $\mathcal{W}$ and $\mathcal{R}$ are convex, $\widehat{\boldsymbol{\beta}}$ in Equation (6.1) and $\boldsymbol{\beta}_0$ in Equation (6.3) exist, and $\mathscr{H}=\sigma(\mathbf{B}, \mathbf{C}, \mathbf{p}_\tau)$. In addition, for some finite nonnegative constants, the following conditions hold: \begin{enumerate}[label=(\roman*)] • $\mathbf{u}_t=(u_{1t,1}, \cdots, u_{1t,M}, \cdots, u_{J_1t,1}, \cdots, u_{J_1t,M})$ is independent over $t$ conditional on $\mathscr{H}$; • $\mathbb{P}(\sum_{t=1}^{T_0}\mathbb{E}[\|\sum_{j=1}^{J_1}\sum_{l=1}^M\tilde{\mathbf{z}}_{t,l}^{[j]} (u_{jt,l}-\mathbb{E}[u_{jt,l}|\mathscr{H}])\|_2^3|\mathscr{H}]\geq \epsilon_\gamma(84(d_\beta^{1/4}+16))^{-1})\geq 1-\pi_\gamma$ where $\tilde{\mathbf{z}}_{t,l}^{[j]}$ is the $((j-1)T_0M+(l-1)T_0+t)$-th column of $\boldsymbol{\Sigma}^{-1/2}\mathbf{Z}'$. \end{enumerate} Then, with $\mathbb{P}$-probability over $\mathscr{H}$ at least $1-\pi_\gamma$, \[ \mathbb{P}\Big(\underline{\mathfrak{c}}(\alpha_0)\leq -\mathbf{p}_\tau'(\widehat{\boldsymbol{\beta}}-\boldsymbol{\beta}_0)\leq \bar{\mathfrak{c}}(1-\alpha_0)|\mathscr{H}\Big) \geq 1-2\alpha_0-\epsilon_\gamma. \]
proofDefine $\mathcal{M}_{\boldsymbol{\xi}}=\{\boldsymbol{\delta}\in\Delta:\boldsymbol{\delta}'\widehat{\mathbf{Q}}\boldsymbol{\delta}-2\boldsymbol{\xi}'\boldsymbol{\delta}\}$. Fix $\widehat{\mathbf{Q}}$ and $\mathbf{p}_\tau$. By Lemma 2 of Cattaneo-Feng-Titiunik_2021_JASA, for any $\kappa$, $\mathcal{A}_\kappa:=\{\boldsymbol{\xi}\in\mathbb{R}^{d_\beta}:\sup_{\boldsymbol{\delta}\in\mathcal{M}_\xi}-\mathbf{p}_\tau'\boldsymbol{\delta} \leq \kappa\}$ and $\mathcal{A}_\kappa'=\{\boldsymbol{\xi}\in\mathbb{R}^{d_\beta}:\inf_{\boldsymbol{\delta}\in\mathcal{M}_\xi}-\mathbf{p}_\tau'\boldsymbol{\delta} \geq \kappa\}$ are convex. By Berry-Esseen Theorem for convex sets Raivc_2019_Bernoulli, \[ |\mathbb{P}(\widehat{\boldsymbol{\gamma}}-\boldsymbol{\gamma}\in\mathcal{A}_\kappa|\mathscr{H})- \mathbb{P}(\mathbf{G}\in\mathcal{A}_\kappa|\mathscr{H})| \leq 42(d_\beta^{1/4}+16)\sum_{t=1}^{T_0}\mathbb{E}\Big[\Big\|\sum_{j=1}^{J_1}\sum_{l=1}^M \tilde{\mathbf{z}}_{t,l}^{[j]}\tilde{u}_{jt,l}\Big\|^3|\mathscr{H}\Big], \] where $\tilde{u}_{jt,l}=u_{jt,l}-\mathbb{E}[u_{jt,l}|\mathscr{H}]$. By condition (ii), with $\mathbb{P}$-probability over $\mathscr{H}$ at least $1-\pi_\gamma$, \[ |\mathbb{P}(\widehat{\boldsymbol{\gamma}}-\boldsymbol{\gamma}\in\mathcal{A}_\kappa|\mathscr{H})- \mathbb{P}(\mathbf{G}\in\mathcal{A}_\kappa|\mathscr{H})|\leq \epsilon_\gamma/2. \] Then, for any $\kappa$, with $\mathbb{P}$-probability over $\mathscr{H}$ at least $1-\pi_\gamma$, \[ \mathbb{P}(-\mathbf{p}_\tau'(\widehat\boldsymbol{\beta}-\boldsymbol{\beta}_0)\leq \kappa|\mathscr{H})\geq \mathbb{P}(\widehat{\boldsymbol{\gamma}}-\boldsymbol{\gamma}\in\mathcal{A}_\kappa|\mathscr{H}) \geq\mathbb{P}(\mathbf{G}\in\mathcal{A}_\kappa|\mathscr{H})-\epsilon_\gamma/2. \] Similarly, we can show for any $\kappa$, \[ \mathbb{P}(-\mathbf{p}_\tau'(\widehat\boldsymbol{\beta}-\boldsymbol{\beta}_0)\geq \kappa|\mathscr{H})\geq \mathbb{P}(\widehat{\boldsymbol{\gamma}}-\boldsymbol{\gamma}\in\mathcal{A}_\kappa'|\mathscr{H}) \geq\mathbb{P}(\mathbf{G}\in\mathcal{A}_\kappa'|\mathscr{H})-\epsilon_\gamma/2. \] Therefore, with $\mathbb{P}$-probability over $\mathscr{H}$ at least $1-\pi_\gamma$, \[ \mathbb{P}\Big(\underline{\mathfrak{c}}(\alpha_0)\leq -\mathbf{p}_\tau'(\widehat{\boldsymbol{\beta}}-\boldsymbol{\beta}_0)\leq \bar{\mathfrak{c}}(1-\alpha_0)\Big|\mathscr{H}\Big) \geq 1-2\alpha_0-\epsilon_\gamma. \] Then the proof is complete.

Proof of Lemma 1

proofIn this proof, the constant $\mathfrak{C}>0$ is a generic constant that is independent of $T_0$ and may be different in different uses. Note that $\bm{m}_{=}(\boldsymbol{\beta}_{0})=\bm{0}$ and $\bm{m}_{\leq}(\boldsymbol{\beta}_{0})\leq \bm{0}$. For $\mathfrak{c}:=\max_{1\leq\ell\leq d_{\leq}}\sup_{\boldsymbol{\beta}\in\mathcal{B}(\boldsymbol{\beta}_0,\varpi_\delta^\star)}\|\frac{\partial}{\partial\boldsymbol{\beta}}m_{\leq,\ell}(\boldsymbol{\beta})\|$, we have $\max_{1\leq\ell\leq d_{\leq}} |m_{\leq,\ell}(\widehat\boldsymbol{\beta})-m_{\leq,\ell}(\boldsymbol{\beta}_0)|\leq \mathfrak{c}\varpi_{\delta}^\star$ with $\mathbb{P}(\cdot|\mathscr{H})$-probability at least $1-\epsilon_{\Delta}^\star$, on an event with $\mathbb{P}$-probability over $\mathscr{H}$ at least $1-\pi_\Delta^\star$. Note that if the $\ell$-th inequality constraint is binding, i.e., $m_{\leq,\ell}(\boldsymbol{\beta}_0)=0$, then $m_{\leq,\ell}(\widehat{\boldsymbol{\beta}})=\frac{\partial}{\partial\boldsymbol{\beta}'}m_{\leq,\ell}(\tilde\boldsymbol{\beta})(\widehat\boldsymbol{\beta}-\boldsymbol{\beta}_0)$ for some $\tilde{\boldsymbol{\beta}}$ between $\boldsymbol{\beta}_0$ and $\widehat\boldsymbol{\beta}$. By the condition imposed on the tuning parameters $\varrho_\ell$'s, on an event with $\mathbb{P}$-probability over $\mathscr{H}$ at least $1-\pi_\Delta^\star$, with $\mathbb{P}(\cdot|\mathscr{H})$-probability at least $1-\epsilon_\Delta^\star$, $\mathcal{A}$ coincides with the set of indices for the binding inequality constraints. Without loss of generality, we assume $\bm{m}_{\leq}(\boldsymbol{\beta}_0)=\bm{0}$ hereafter. Otherwise, the non-binding constraints can be dropped, and the proof can proceed the same way as described below. Define $\Gamma_{=}(\boldsymbol{\beta})=\frac{\partial}{\partial \boldsymbol{\beta}'}\bm{m}_{=}(\boldsymbol{\beta})$ and $\Gamma_{\leq}(\boldsymbol{\beta})=\frac{\partial}{\partial \boldsymbol{\beta}'}\bm{m}_{\leq}(\boldsymbol{\beta})$. Let \[ \Gamma(\boldsymbol{\beta})=\Big(\Gamma_{=}'(\boldsymbol{\beta}), \Gamma_{\leq}'(\boldsymbol{\beta}),\Gamma_{c}'(\boldsymbol{\beta}_0)\Big)',\quad \Gamma^0=\Gamma(\boldsymbol{\beta}_0),\quad \Gamma^\star=\Gamma(\widehat\boldsymbol{\beta}), \] where $\Gamma_c(\boldsymbol{\beta}_0)$ is chosen such that $\Gamma(\boldsymbol{\beta}_0)$ is non-degenerate. By conditions (i) and (ii) imposed in the lemma, $\|\Gamma^0-\Gamma^\star\|\leq \mathfrak{C}\|\widehat\boldsymbol{\beta}-\boldsymbol{\beta}_0\|$ with $\mathbb{P}(\cdot|\mathscr{H})$-probability at least $1-\epsilon_\Delta^\star$, on an event with $\mathbb{P}$-probability at least $1-\pi_\Delta^\star$. Let $$\bm{m}^+(\cdot)=\Big( \bm{m}_{=}'(\cdot),\;\bm{m}_{\leq}'(\cdot), \;(\cdot-\boldsymbol{\beta}_0)'\times\Gamma_c(\boldsymbol{\beta}_0)'\Big)'. $$ Then, $\bm{m}^+(\boldsymbol{\beta}_0)=\bm{0}$. For each $\boldsymbol{\beta}$ in the neighborhood around $\boldsymbol{\beta}_0$ such that $\boldsymbol{\beta}-\boldsymbol{\beta}_0\in\Delta\cap\mathcal{B}(0,\varpi_\delta^\star)$, define \[ \boldsymbol{\lambda}^0=(\Gamma^{ 0})^{-1}\Big(\bm{m}^{+}(\boldsymbol{\beta})-\bm{m}^+(\boldsymbol{\beta}_0)\Big). \] Thus, $\Gamma^0_{=}\boldsymbol{\lambda}^0=\bm{0}$, $\Gamma^0_{\leq}\boldsymbol{\lambda}^0\leq \bm{0}$. Note that by Taylor's expansion, \[ \|\boldsymbol{\lambda}^0-(\Gamma^0)^{-1}\Gamma^0(\boldsymbol{\beta}-\boldsymbol{\beta}_0)\|\leq \mathfrak{C}\|\boldsymbol{\beta}-\boldsymbol{\beta}_0\|^2, \] implying that $\|\boldsymbol{\lambda}^0-(\boldsymbol{\beta}-\boldsymbol{\beta}_0)\| \leq \mathfrak{C}(\varpi_\delta^\star)^2 $ with $\mathbb{P}(\cdot|\mathscr{H})$-probability at least $1-\epsilon_\Delta^\star$, on an event with $\mathbb{P}$-probability at least $1-\pi_\Delta^\star$. Next, define $\tilde{\bm{m}}(\cdot)=\bm{m}^+(\widehat\boldsymbol{\beta}+\cdot)-\bm{m}^+(\widehat\boldsymbol{\beta})$ and $\tilde{\boldsymbol{\beta}}=\boldsymbol{\phi}^\star+\widehat\boldsymbol{\beta}$ for $\boldsymbol{\phi}^\star$ defined below: \[ \boldsymbol{\phi}^\star:=\tilde{\bm{m}}^{-1}\Big(\Gamma^\star(\boldsymbol{\lambda}^0-(\Gamma^\star)^{-1} (\Gamma^\star-\Gamma^0)\boldsymbol{\lambda}^0)\Big). \] By Taylor's expansion, \[ \begin{split} \boldsymbol{\phi}^\star &=\tilde{\bm{m}}^{-1}(\bm{0})+ \Big[\frac{\partial}{\partial \boldsymbol{\phi}'}\tilde{\bm{m}}(\bm{0})\Big]^{-1} \Gamma^\star\Big(\boldsymbol{\lambda}^0-(\Gamma^\star)^{-1} (\Gamma^\star-\Gamma^0)\boldsymbol{\lambda}^0\Big)+\mathfrak{Re}\\ &=\boldsymbol{\lambda}^0-(\Gamma^\star)^{-1}(\Gamma^\star-\Gamma^0)\boldsymbol{\lambda}^0+\mathfrak{Re}, \end{split} \] where $\|(\Gamma^\star)^{-1}(\Gamma^\star-\Gamma^0)\boldsymbol{\lambda}^0+\mathfrak{Re}\|\leq \mathfrak{C}\|\boldsymbol{\lambda}^0\|^2$ with $\mathbb{P}(\cdot|\mathscr{H})$-probability at least $1-\epsilon_\Delta^\star$, on an event with $\mathbb{P}$-probability over $\mathscr{H}$ at least $1-\pi_\Delta^\star$. That is, we actually find $\tilde{\boldsymbol{\beta}}$ such that $\|(\tilde{\boldsymbol{\beta}}-\widehat\boldsymbol{\beta})-\boldsymbol{\lambda}^0\|\leq \mathfrak{C}(\varpi_\delta^\star)^2$. Note that $$ \bm{m}^+(\widehat\boldsymbol{\beta}+\boldsymbol{\phi}^\star)-\bm{m}^+(\widehat\boldsymbol{\beta})= \tilde{\bm{m}}(\boldsymbol{\phi}^\star)= \tilde{\bm{m}}\Big(\tilde{\bm{m}}^{-1}\Big(\Gamma^\star(\boldsymbol{\lambda}^0-(\Gamma^\star)^{-1} (\Gamma^\star-\Gamma^0)\boldsymbol{\lambda}^0)\Big)\Big).$$ Thus, $\bm{m}_{=}(\tilde{\boldsymbol{\beta}})= \bm{0}$ and $\bm{m}_{\leq}(\tilde{\boldsymbol{\beta}})\leq\bm{m}_{\leq}(\widehat{\boldsymbol{\beta}})$, i.e., $\boldsymbol{\phi}^\star\in\widehat{\Delta}$. This shows that $\mathrm{dist}(\Delta\cap\mathcal{B}(\bm{0}, \varpi_\delta^\star),\widehat{\Delta})\leq \varpi_\Delta^\star$ with $\mathbb{P}(\cdot|\mathscr{H})$-probability at least $1-\epsilon_\Delta^\star$, on an event with $\mathbb{P}$-probability over $\mathscr{H}$ at least $1-\pi_\Delta^\star$. Then, the desired conclusion holds by definition of the $\varepsilon$-enlargement $\widehat{\Delta}_\varepsilon$. Note that when constraints are linear, $\Gamma^0=\Gamma^\star$ and the second-order derivative of $\bm{m}^+(\cdot)$ is exactly zero. So the above calculation implies that $\boldsymbol{\beta}-\boldsymbol{\beta}_0=\boldsymbol{\lambda}_0=\boldsymbol{\phi}^\star$. In the above, we make use of the fact that $\Gamma^\star$ is non-degenerate, i.e., its smallest eigenvalue is bounded away from zero. Note that by assumptions on the constraints, it is feasible to construct $\Gamma_c$ such that $\Gamma^0$ is non-degenerate (with high probability over $\mathscr{H}$). Then, by Weyl's inequality, \[ s_{\min}(\Gamma^\star(\Gamma^\star)')\geq s_{\min}(\Gamma^0(\Gamma^0)')-\mathfrak{C}\|\widehat\boldsymbol{\beta}-\boldsymbol{\beta}_0\|, \] implying $s_{\min}(\Gamma^\star)\geq s_{\min}(\Gamma^0)-\mathfrak{C}\|\widehat\boldsymbol{\beta}-\boldsymbol{\beta}_0\|>0$ with $\mathbb{P}(\cdot|\mathscr{H})$-probability at least $1-\epsilon_\Delta^\star$, on an event with $\mathbb{P}$-probability at least $1-\pi_\Delta^\star$, where $s_{\min}(\cdot)$ here denotes the smallest eigenvalue of the symmetric matrix inside. Then, the proof is complete.

Discussion on Tuning Parameters

Determining Binding Constraints

Our proposed strategy in Section 6.2 for constructing the constraint set in the simulation relies on the tuning parameters $\varrho_\ell$'s, which are used to differentiate the binding constraints from the non-binding ones. The recommended choice described in (6.8) and (6.9) can be rationalized as follows.

For ease of notation, suppose that there is only one treated unit, and thus we can omit the superscript $[i]$ for quantities like $\boldsymbol{\beta}_0^{[i]}$ or $\varrho^{[i]}$ with no confusion. Assume the $\ell$-th constraint $m_{\leq, \ell}(\boldsymbol{\beta})\leq 0$ is binding ($m_{\leq, \ell}(\boldsymbol{\beta}_0)=0$), and $m_{\leq, \ell}(\cdot)$ is sufficiently smooth. By the mean value theorem, $|m_{\leq, \ell}(\widehat{\boldsymbol{\beta}})|=|\frac{\partial}{\partial\boldsymbol{\beta}'}m_{\leq, \ell}(\check\boldsymbol{\beta})(\widehat{\boldsymbol{\beta}}-\boldsymbol{\beta}_0)|\leq \|\frac{\partial}{\partial\boldsymbol{\beta}}m_{\leq, \ell}(\check\boldsymbol{\beta})\|_2\|\widehat{\boldsymbol{\beta}}-\boldsymbol{\beta}_0\|_2$, where $\check{\boldsymbol{\beta}}$ is some point between $\boldsymbol{\beta}_0$ and $\widehat{\boldsymbol{\beta}}$ . The parameters $\varrho_\ell^{[i]}$ are used to characterize this upper bound. For $\frac{\partial}{\partial\boldsymbol{\beta}}m_{\leq, \ell}(\check\boldsymbol{\beta})$, a natural “estimator” would be $\frac{\partial}{\partial\boldsymbol{\beta}}m_{\leq, \ell}(\widehat{\boldsymbol{\beta}})$. The difference between the two is bounded by $c_1\|\widehat{\boldsymbol{\beta}}-\boldsymbol{\beta}_0\|_2$ for some constant $c_1>0$ (with high probability). Then, it remains to characterize $\|\widehat{\boldsymbol{\beta}}-\boldsymbol{\beta}_0\|_2$.

Note that the basic inequality below always holds (deterministically) by optimality of $\widehat{\boldsymbol{\beta}}$:

equation[equation omitted — 259 chars of source]

We first discuss the denominator. Recall that $\widehat{\mathbf{Q}}=\mathbf{Z}'\mathbf{V}\mathbf{Z}$. Since an (unrestricted) intercept is often included in the SC construction, we assume the predictor variables included in $\mathbf{Z}$ are de-meaned (and the constant term is excluded from $\mathbf{Z}$). If $\mathbf{V}$ is an identity matrix and the predictor variables are (approximately) orthogonal to each other, then $s_{\min}(\widehat{\mathbf{Q}}/T_0)$ is the minimum diagonal entry of $\widehat{\mathbf{Q}}/T_0$, i.e., $\min_{1\leq j\leq J_0}\widehat{\sigma}_{b_j}^2$, which motivates the choice of the denominator in (6.9). On the other hand, recall that the numerator $\widehat{\boldsymbol{\gamma}}=\mathbf{Z}'\mathbf{u}$. As discussed in the paper, we can (conditionally) approximate $\widehat{\boldsymbol{\gamma}}-\boldsymbol{\gamma}$ by the Gaussian vector $\mathbf{G}$. Then, by Gaussian tail bound, it can be shown that, for any large $\lambda>0$, $T_0^{-1/2}\|\widehat{\boldsymbol{\gamma}}-\boldsymbol{\gamma}\|_\infty\leq (\sqrt{2\log d_\beta}+\lambda)\times\sqrt{\max_{1\leq j\leq J_0} \mathbb{V}[\widehat{\gamma}_j|\mathscr{H}]/T_0}$ with high probability. When $\mathbf{u}$ is independent of $\mathscr{H}$ and its components are independent over $t$, $\max_{1\leq j\leq J_0} \mathbb{V}[\widehat{\gamma}_j|\mathscr{H}]/T_0=\sigma_u^2\times \max_{1\leq j\leq J_0}\widehat{\sigma}_{b_j}^2$. Therefore, we can set \[ \mathcal{C}=\mathcal{C}_1:= \frac{2\sqrt{d_\beta}(\sqrt{2\log d_\beta}+\lambda)\max_{1\leq j \leq J_0}\widehat{\sigma}_{b_j}\widehat{\sigma}_{u}}{\min_{1\leq j \leq J_0}\widehat{\sigma}^2_{b_j}} \] for any large $\lambda$. In most applications, however, the SC weights are sparse due to the simplex- or lasso-type constraints. It is known from the sparse linear regression literature that the bound $\mathcal{C}_1$ can be further improved by replacing the factor $\sqrt{d_\beta}$ by $c_3\sqrt{\|\boldsymbol{\beta}_0\|_0}$, where $\|\cdot\|_0$ denotes the number of nonzeros in a vector and $c_3>0$ is some absolute constant (see, e.g., Wainwright_2019_Book). Assuming $\sqrt{\log T_0}\geq 4\sqrt{2}c_3$, we take $\lambda=\sqrt{\log d_\beta\log T_0}/(4c_3)$ and use $\|\widehat{\boldsymbol{\beta}}\|_0$ as a proxy for $\|\boldsymbol{\beta}_0\|_0$, which yields our recommended choice of $\mathcal{C}$ given in (6.9): \[ \mathcal{C}=\mathcal{C}_2:= \frac{\sqrt{\|\widehat{\boldsymbol{\beta}}\|_0\log d_\beta\log T_0}\max_{1\leq j \leq J_0}\widehat{\sigma}_{b_j}\widehat{\sigma}_{u}}{\min_{1\leq j \leq J_0}\widehat{\sigma}^2_{b_j}}. \] Furthermore, when $\widehat{\sigma}_{b_j}^2$ is roughly the same across $j$, then $\mathcal{C}_2$ can be further simplified to \[ \mathcal{C}=\mathcal{C}_3 = \frac{\sqrt{\|\widehat{\boldsymbol{\beta}}\|_0\log d_\beta\log T_0}\;\widehat{\sigma}_{u}}{\min_{1\leq j \leq J_0}\widehat{\sigma}_{b_j}}. \]

While these choices are only rules of thumb justified under specific assumptions on the data generating process, they at least have the correct order of magnitude and are valid at least when $T_0$ is large. For instance, for the cointegrated data considered in our basic setup, it can be shown that, for sufficiently small $c_4>0$, $\widehat{\sigma}^2_{b_j}\geq c_4T_0$ with high probability, and $\widehat{\sigma}_u^2$ is bounded with high probability. Therefore, the suggested $\varrho$ based on $\mathcal{C}_2$ has the order of $T_0^{-1}$, which is the well-known rate of convergence for cointegrated regression.

When SC weights are obtained by matching on both stationary and non-stationary features, the contribution of the stationary components to the concentration of $\widehat{\boldsymbol{\beta}}$ is negligible. In other words, the order of magnitude for the bound on $\|\widehat{\boldsymbol{\beta}}-\boldsymbol{\beta}\|_2$ is primarily determined by the non-stationary components. To see this, assume that we match on two features, a non-stationary $\mathbf{B}_{1}=(\mathbf{b}_{1,1}, \cdots, \mathbf{b}_{T_0,1})'\in\mathbb{R}^{T_0\times J_0}$ and a stationary $\mathbf{B}_{2}=(\mathbf{b}_{1,2}, \cdots, \mathbf{b}_{T_0,2})'\in\mathbb{R}^{T_0\times J_0}$. Consider the basic inequality (ref) again. $\widehat{\mathbf{Q}}/T_0$ can be written as $\frac{1}{T_0}\sum_{t=1}^{T_0}\mathbf{b}_{t,1}\mathbf{b}_{t,1}'+\frac{1}{T_0}\sum_{t=1}^{T_0}\mathbf{b}_{t,2}\mathbf{b}_{t,2}'=:\mathrm{I}+\mathrm{II}$. As explained previously, with high probability, $s_{\min}(\mathrm{I}/T_0)$ is bounded from below by $T_0$, up to a sufficiently small constant, whereas $\mathrm{II}/T_0$ is much smaller---under mild conditions, it is bounded with high probability. For the numerator $(\widehat{\boldsymbol{\gamma}}-\boldsymbol{\gamma})/T_0$, we can similarly decompose it into a non-stationary part and a stationary part. By the argument described before, it follows that the stationary part is much smaller than the non-stationary part, leading to a deviation bound for $\widehat{\boldsymbol{\beta}}$ on the order of $T_0^{-1}$. Therefore, we recommend applying the above formulas using the non-stationary features only in practice.

Finally, as a reminder, other choices of $\varrho$ can be proposed based on a slightly different basic inequality: $\|\widehat{\boldsymbol{\beta}}-\boldsymbol{\beta}_0\|_2\leq 2\|\widehat{\boldsymbol{\gamma}}\|_2/s_{\min}(\widehat{\mathbf{Q}})$. It also holds deterministically by optimality of $\widehat{\boldsymbol{\beta}}$. The denominator is the same as before, but the numerator $\widehat{\boldsymbol{\gamma}}$ is not demeaned. When the model is correctly specified ($\boldsymbol{\gamma}=\bm{0}$), the numerator may be small; however, when the model is misspecified, $\widehat{\boldsymbol{\gamma}}$ is not necessarily small, and it may be characterized by its sample analogue. Thus, we can set \[ \varrho=\frac{2\sqrt{d_\beta}\max_{1\leq j\leq J_o}\widehat{\sigma}^2_{b_ju}}{\min_{1\leq j\leq J_0}\widehat{\sigma}^2_{b_j}} \] where $\widehat{\sigma}^2_{b_ju}$ is the estimated covariance between the pseudo-true residual $\mathbf{u}$ and the $j$-th column of $\mathbf{B}$ (the features of the $j$-th control unit). However, compared to the alternatives described in the paper, this bound is generally loose and thus is not recommended.

Adjustment for Nonlinear Constraints

As described in the main paper, we have an additional adjustment to nonlinear constraints in the simulation (the second term on the right-hand side of (6.10)). In this section we briefly discuss the necessity of this adjustment and the justification of the proposed strategy.

Necessity of Adjustment

The constraints allowed in this paper is more general than those considered in Cattaneo-Feng-Titiunik_2021_JASA. Specifically, condition (T2.iii) in Cattaneo-Feng-Titiunik_2021_JASA requires the constraint set used in the simulation for in-sample uncertainty quantification be locally equal to the original constraint set in the synthetic control problem. In the notation of this paper, that condition means that, with some high probability, $\Delta^{\star}\cap\mathcal{B}(\bm{0},\varpi_\delta^\star)=\Delta\cap\mathcal{B}(\bm{0},\varpi_\delta^\star)$ for some (small) $\varpi_\delta^\star$-neighborhood of zero. This is generally true when constraints are formed by linear functions. To see this, suppose that there are two donor units, an L1 constraint $\mathcal{W}=\{(w_1, w_2): |w_1|+|w_2|\leq 1\}$ is imposed, and the pseudo-true value is $\mathbf{w}_0=(1,0)$. The constraint set $\Delta^\star$ used in the simulation relies on the estimated weights $\widehat{\mathbf{w}}$, which are generally close (but not exactly equal) to $\mathbf{w}_0$ with high probability. For instance, let $\widehat{\mathbf{w}}=(0.99, 0.01)$, and applying the suggested strategy described in Section 6.2, one correctly finds that two linear constraints among those derived from decomposing the original L1 constraint, $w_1+w_2\leq 1$ and $w_1-w_2\leq 1$, are binding. Thus, one defines $\Delta^\star=\{(\delta_1, \delta_2): (\delta_1+0.99)+(\delta_2+0.01)\leq 0.99+0.01, (\delta_1+0.99)-(\delta_2+0.01)\leq 0.99-0.01\}$, which is exactly the same as $\Delta=\mathcal{W}-\mathbf{w}_0$ locally around $(0,0)$.

However, such an exact equality is generally not possible when constraints are formed by nonlinear functions. Still consider the same example, but an L2 constraint $\mathcal{W}=\{(w_1, w_2): w_1^2+w_2^2\leq 1\}$ is imposed instead. Applying the same strategy, one correctly finds that the constraint should be binding and thus lets $\Delta^\star=\{(\delta_1,\delta_2):(\delta_1+0.99)^2+(\delta_2+0.01)^2\leq 0.99^2+0.01^2\}$, whereas the (centered) original constraint set is $\Delta=\{(\delta_1, \delta_2): (\delta_1+1)^2+\delta_2^2=1\}$. In other words, $\Delta$ and $\Delta^\star$ are two “circles” passing through $(0,0)$. Although they are locally close near $(0,0)$, their boundaries have different tangents at that point, and hence the two sets are not locally identical, no matter how small the neighborhood $\mathcal{B}(\bm{0},\varpi^\star)$ is. In this sense, the results in Cattaneo-Feng-Titiunik_2021_JASA are more applicable to synthetic controls with linear constraints (such as simplex and Lasso). By contrast, the local approximate equality in condition (iii) of Theorem 1 is much weaker and makes the results in this paper applicable to more general cases with possibly nonlinear constraints. Given the fact described above, when there are (binding) nonlinear constraints, we propose to enlarge the feasibility set defined by nonlinear constraints to satisfy the sufficient condition (iii) of Theorem 1.

Enlarging Nonlinear Constraint Sets

In this section we briefly discuss the justification for the adjustment in (6.10). Our strategy is motivated by the proof of Lemma 1, which characterizes the distance between $\Delta\cap\mathcal{B}(\bm{0},\varpi_\delta^\star)$ and $\widehat{\Delta}$. For simplicity, assume that there is only one treated unit (so the superscript $[i]$ can be omitted with no confusion).

Suppose that we only have binding inequality constraints $\bm{m}(\boldsymbol{\beta})\leq \bm{0}$ (so $\bm{m}(\boldsymbol{\beta}_0)=\bm{0}$), where $\bm{m}(\cdot)$ is a $d_\leq$-vector of sufficiently smooth functions. For any feasible $\boldsymbol{\beta}\in\mathbb{R}^{d_\beta}$ close to $\boldsymbol{\beta}_0$ such that $\boldsymbol{\beta}-\boldsymbol{\beta}_0\in\Delta\cap\mathcal{B}(\bm{0}, \varpi_\delta^\star)$, \[ \bm{m}(\boldsymbol{\beta})=\bm{m}(\boldsymbol{\beta})-\bm{m}(\boldsymbol{\beta}_0)=\frac{\partial}{\partial\boldsymbol{\beta}'}\bm{m}(\boldsymbol{\beta}_0)(\boldsymbol{\beta}-\boldsymbol{\beta}_0)+\frac{1}{2}(\boldsymbol{\beta}-\boldsymbol{\beta}_0)'\Big[\frac{\partial^2}{\partial\boldsymbol{\beta}\partial\boldsymbol{\beta}'}\bm{m}(\check{\boldsymbol{\beta}})\Big](\boldsymbol{\beta}-\boldsymbol{\beta}_0) \] for some point $\check{\boldsymbol{\beta}}$ between $\boldsymbol{\beta}$ and $\boldsymbol{\beta}_0$, where $\partial^2\bm{m}(\check{\boldsymbol{\beta}})/\partial\boldsymbol{\beta}\partial\boldsymbol{\beta}'$ is a $d_\beta\times d_\beta\times d_\leq$ array, with each sheet the second-order derivative matrix for one constraint function. Thus, $(\boldsymbol{\beta}-\boldsymbol{\beta}_0)'[\partial^2\bm{m}(\check{\boldsymbol{\beta}})/\partial\boldsymbol{\beta}\partial\boldsymbol{\beta}'](\boldsymbol{\beta}-\boldsymbol{\beta}_0)$ is a $d_\leq$-vector, with each element corresponding to $(\boldsymbol{\beta}-\boldsymbol{\beta}_0)'[\partial^2m_\ell(\check{\boldsymbol{\beta}})/\partial\boldsymbol{\beta}\partial\boldsymbol{\beta}'](\boldsymbol{\beta}-\boldsymbol{\beta}_0)$ for $1\leq \ell\leq d_\leq$.

Let $\Gamma_0=\frac{\partial}{\partial\boldsymbol{\beta}}\bm{m}(\boldsymbol{\beta}_0)$ and $\widehat{\Gamma}=\frac{\partial}{\partial\boldsymbol{\beta}}\bm{m}(\widehat{\boldsymbol{\beta}})$. Assume $\Gamma_0$ is invertible in the neighborhood of $\boldsymbol{\beta}_0$. (When $d_\leq< d_\beta$, we can complement $\Gamma_0$ with additional rows and construct an invertible one, which is formalized in the proof of Lemma 1). Then, \[ \Gamma_0^{-1}\bm{m}(\boldsymbol{\beta})-(\boldsymbol{\beta}-\boldsymbol{\beta}_0) =\frac{1}{2}\Gamma_0^{-1}\Big[(\boldsymbol{\beta}-\boldsymbol{\beta}_0)'\frac{\partial^2}{\partial\boldsymbol{\beta}\partial\boldsymbol{\beta}'}\bm{m}(\check{\boldsymbol{\beta}})(\boldsymbol{\beta}-\boldsymbol{\beta}_0)\Big]=:\mathscr{L} \] It can be shown that $\boldsymbol{\lambda}:=\Gamma_0^{-1}\bm{m}(\boldsymbol{\beta})$ is (approximately) in $\widehat{\Delta}$. To see this, $$ \bm{m}(\widehat{\boldsymbol{\beta}}+\boldsymbol{\lambda})\approx\bm{m}(\widehat{\boldsymbol{\beta}})+\frac{\partial}{\partial\boldsymbol{\beta}'}\bm{m}(\boldsymbol{\beta}_0)\boldsymbol{\lambda}=\bm{m}(\widehat{\boldsymbol{\beta}})+(\bm{m}(\boldsymbol{\beta})-\bm{m}(\boldsymbol{\beta}_0))\leq \bm{m}(\widehat{\boldsymbol{\beta}}). $$ This is formalized in the proof of Lemma 1. For the moment, assume $\boldsymbol{\lambda}\in\widehat{\Delta}$. Then,

align*[align* omitted — 599 chars of source]

The errors introduced by “$\approx$” are of smaller order under mild conditions. The above calculation implies that any $\boldsymbol{\beta}-\boldsymbol{\beta}_0\in\Delta\cap\mathcal{B}(\bm{0},\varpi_\delta^\star)$ (approximately) belongs to the following adjusted version of $\widehat{\Delta}$ in (6.7): \[ \Big\{\boldsymbol{\delta}: \bm{m}(\widehat{\boldsymbol{\beta}}+\boldsymbol{\delta})\leq \bm{m}(\widehat{\boldsymbol{\beta}})+\frac{1}{2}(\boldsymbol{\beta}-\boldsymbol{\beta}_0)'\frac{\partial^2} {\partial\boldsymbol{\beta}\partial\boldsymbol{\beta}'}\bm{m}(\boldsymbol{\beta}_0)(\boldsymbol{\beta}-\boldsymbol{\beta}_0)\Big\}. \] Then, for each inequality constraint $m_{\leq,\ell}(\widehat{\boldsymbol{\beta}}+\boldsymbol{\delta})\leq m_{\leq, \ell}(\widehat{\boldsymbol{\beta}})$, the “adjustment” to the upper bound is at most \[ \frac{1}{2}s_{\max}\Big(\frac{\partial^2} {\partial\boldsymbol{\beta}\partial\boldsymbol{\beta}'}m_{\leq,\ell}(\boldsymbol{\beta}_0)\Big)\|\boldsymbol{\beta}-\boldsymbol{\beta}_0\|_2^2. \] Given the high-probability bound $\varrho$ for $\|\widehat{\boldsymbol{\beta}}-\boldsymbol{\beta}_0\|_2$, we propose the adjustment in (6.10).

\noindentNotes: Our enlargement strategy used to define $\Delta^\star$ in Section 6.2 can be conceptually understood as follows. A constraint set $\Delta$ can be generally written as $\Delta=\cap_{j=1}^L A_j$ for a sequence of convex sets $A_j$. Each $A_j$ is defined by one inequality. Accordingly, $\widehat{\Delta}$ in (6.7) (with no enlargement) can be written as $\widehat{\Delta}=\cap_{j=1}^L\widehat{A}_j$.

Let $\widehat{A}_{j,\varepsilon_j}$ be an $\varepsilon_j$-enlargement of $\widehat{A}_j$. We can apply Lemma 1 to each pair of $A_j$ and $\widehat{A}_j$, and then $A_j\subseteq\widehat{A}_{j,\varepsilon_j}$ for an appropriate (small) $\varepsilon_j$. For any point $g\in\Delta$, $g\in A_j$ for every $j$. So $g\in\widehat{A}_{j,\varepsilon_j}$. Then, $g\in\cap_{j=1}^L \widehat{A}_{j,\varepsilon_j}$. Since it holds for any $g$, we have $\Delta\subseteq\Delta^\star:=\cap_{j=1}^L \widehat{A}_{j,\varepsilon_j}$. The sufficient condition (iii) in Theorem 1 holds.

Second Order Cone Programming

In this section, we first define three types of convex optimization problems that we will be relying on (for background knowledge and technical details, see Boyd_2004_BOOK). Second, we illustrate the link between these three families of convex problems. Third, we show that the optimization problems underlying the prediction/estimation and uncertainty quantification problems for SC presented in Section 6 of the main text are quadratically constrained quadratic program (QCQP) and quadratically constrained linear problem (QCLP), respectively, and show how to represent them as second-order cone program (SOCP). Finally, we provide two examples by showing how to write the L1-L2-type and Lasso-type constraints in conic form. These approaches are implemented in our companion general-purpose software \citep*{Cattaneo-Feng-Palomba-Titiunik_2025_JSS}, where we show that they lead to remarkable speed and scalability improvements.

Families of Convex Optimization Problems

QCQPs and QCLPs. A quadratically constrained quadratic program is an optimization problem of with the following form

align[align omitted — 376 chars of source]

where $\mathbf{P}_0, \mathbf{P}_1, \ldots, \mathbf{P}_m \in \mathcal{M}_{n\times n}(\operatorname{\mathbb{R}})$, $ \mathbf{q}_0,\mathbf{q}_1,\ldots,\mathbf{q}_m\in \operatorname{\mathbb{R}}^n$, $\mathbf{x}\in \operatorname{\mathbb{R}}^n$, $\mathbf{F} \in \mathcal{M}_{m \times n}(\operatorname{\mathbb{R}})$, $\mathbf{g} \in \operatorname{\mathbb{R}}^m$, and $r_0,r_1, \ldots, r_m, w \in \operatorname{\mathbb{R}}$. If all the matrices $\mathbf{P}_0,\mathbf{P}_1,\ldots, \mathbf{P}_m$ are positive semi-definite the QCQP is convex. Moreover, if $\mathbf{P}_0=\mathbf{0}$ the QCQP becomes a QCLP. For this reason, in what follows we will restrict our attention to QCQPs as they naturally embed QCLPs.

SOCPs. To define a SOCP, it is necessary to first give the definition of a second-order cone and then introduce the notion of associated generalized inequality.

\uline{Second-order cone definition.} A set $\mathcal{C}$ is called a cone if for every $\mathbf{x} \in \mathcal{C}$ and $\alpha\geq 0$ we have $\alpha \mathbf{x}\in \mathcal{C}$. A set $\mathcal{C}$ is a convex cone if it is convex and a cone, i.e. if $\forall\, \mathbf{x}_1,\mathbf{x}_2 \in \mathcal{C}, $and $\forall\, \alpha_1,\alpha_2 \geq 0$, we have \[\alpha_1 \mathbf{x}_1 + \alpha \mathbf{x}_2 \in \mathcal{C}.\] Now, consider any norm $||\cdot||$ defined on the Euclidean space $\operatorname{\mathbb{R}}^n$. The norm cone associated with the norm $||\cdot ||$ is defined to be the set \[\mathcal{C}=\{(\mathbf{x},t)\in \operatorname{\mathbb{R}}^{n+1}: ||\mathbf{x}|| \leq t\}\] and it is a convex cone by the standard properties of the norms. A second-order cone is the associated norm cone for the Euclidean norm and it is typically defined as

align*[align* omitted — 398 chars of source]

\uline{Generalized inequality.} A cone $\mathcal{C}$ is solid if it has non-empty interior and it is pointed if $\mathbf{x}\in \mathcal{C}, -\mathbf{x}\in \mathcal{C}$ implies that $\mathbf{x}=0$. We say that a cone $\mathcal{C}$ is proper if is it convex, closed, solid, and pointed. Proper cones in the Euclidean space $\operatorname{\mathbb{R}}^n$ are useful because they induce a partial ordering that enjoys almost all the properties of the basic one in $\operatorname{\mathbb{R}}$. Therefore, given a cone $\mathcal{C}\subseteq \operatorname{\mathbb{R}}^n$, we can define the generalized inequality $\preceq_\mathcal{C}$ for any two vectors $\mathbf{x},\mathbf{y}\in \operatorname{\mathbb{R}}^n$ \[\mathbf{x}\preceq_{\mathcal{C}} \mathbf{y} \iff \mathbf{y} -\mathbf{x} \in \mathcal{C}.\] From this definition we can see that quadratic constraints such as $||\mathbf{x}||_2 \leq h$ can be re-written as second-order cone constraints of the form $\mathbf{x}\preceq_\mathcal{C} h$ for some second-order cone $\mathcal{C}$. Note that if $\mathcal{C}=\operatorname{\mathbb{R}}_+^m$ then if $m=1$, $\preceq_\mathcal{C}$ is the standard inequality $\leq$ in $\operatorname{\mathbb{R}}$, whereas if $m>1$, $\preceq_\mathcal{C}$ is the component-wise inequality in $\operatorname{\mathbb{R}}^m$.

\uline{Second-order cone program.} Let $\mathcal{K}$ be a cone such that $\mathcal{K} = \mathbb{R}_+^m \times \mathcal{K}_1 \times \mathcal{K}_2 \times\cdots\times \mathcal{K}_L$ where $\mathcal{K}_l := \{(k_0,\mathbf{k}_1)\in \operatorname{\mathbb{R}}\times \operatorname{\mathbb{R}}^{l}: ||\mathbf{k}_1||_2\leq k_0\}, l= 1,\ldots, L$. Let $\preceq_\mathcal{K}$ be the generalized inequality associated with the cone $\mathcal{K}$. An optimization problem is called second-order cone program if it has the following form

align[align omitted — 293 chars of source]

Link Between QCQP and SOCPs

Any QCQP can be converted to a SOCP Boyd_2004_BOOK. In other words, we can always rewrite an optimization problem such as (ref) in the form of (ref). First, we present the general result and then we explain all the necessary steps to reformulate QCQPs as SOCPs. Without loss of generality, assume that $w=0$ in (ref) and, to ease notation, let $m=1$ so that there is only a single quadratic inequality constraint. Moreover, given any positive semi-definite matrix $\mathbf{P}$, let $\mathbf{P}^{1/2}$ be the square root of $\mathbf{P}$, that is the unique symmetric positive semi-definite matrix $\mathbf{R}$ such that $\mathbf{R}\mathbf{R}=\mathbf{R}'\mathbf{R}=\mathbf{P}$. Then for any QCQP the following two formulations are equivalent

multicols{2} \begin{align*} & \uline{QCQP} \\ \min_{\mathbf{x}}\quad & \mathbf{x}' \mathbf{P}_0 \mathbf{x} + \mathbf{q}_0'\mathbf{x} \\ subject to \quad & \mathbf{F}\mathbf{x} = \mathbf{g}, \\ & \mathbf{x}' \mathbf{P}_1\mathbf{x} +\mathbf{q}_1'\mathbf{x} +r_1\leq 0. \\ \end{align*} \null \columnbreak \begin{align*} & \uline{SOCP} \\ \min_{\mathbf{x}, v, t, s}\quad & v + \mathbf{q}_0'\mathbf{x} \\ subject to \quad &\mathbf{F}\mathbf{x} = \mathbf{g}, \\ & t+\mathbf{q}_1'\mathbf{x} +r_1\preceq_{\operatorname{\mathbb{R}}_+} 0, \\ & \mathbf{P}^{1/2}_0\mathbf{x} \preceq_{K_{1+n}} v, \\ & \mathbf{P}^{1/2}_1\mathbf{x} \preceq_{K_{1+n}} s. \end{align*}

We can see that the logic beneath the conversion of a QCQP into a SOCP is to “linearize” all the non-linear terms appearing either in the objective function or in the inequality constraints. The “linearization” step does come at a cost, as it requires the introduction of a slack variable every time we rely on it. Indeed, above we linearized the objective function and the quadratic inequality constraint by introducing two auxiliary slack variables.

More formally, let $\mathbf{x}'\mathbf{P}\mathbf{x}$ be a symmetric positive semi-definite quadratic form and consider the constraint $\mathbf{x}'\mathbf{P}\mathbf{x}\leq y$. Then

enumerate[label=(\roman*)] • Since $\mathbf{P}$ is symmetric positive semi-definite the epigraph $\mathbf{x}' \mathbf{P} \mathbf{x}\leq y$ is a convex set and $\mathbf{P}^{1/2}$ is well-defined. • Write the inequality constraint as a constraint involving the Euclidean norm $\|\cdot\|_2$ \[y\geq \mathbf{x}'\mathbf{P}\mathbf{x} = \mathbf{x}'\mathbf{P}^{1/2}\mathbf{P}^{1/2}\mathbf{x} = ||\mathbf{P}^{1/2}\mathbf{x}||_2^2.\] • Note that \begin{align} ||\mathbf{P}^{1/2}\mathbf{x}||_2^2\leq y \iff \left\|\left[\begin{array}{c} 1-y \\ 2 \mathbf{P}^{1 / 2} \mathbf{x} \end{array}\right]\right\|_2 \leq 1+y, \end{align} which can be verified by squaring the two sides of the last inequality and expand the norm. • More is true, as the right-most inequality in (ref) defines the following second-order cone for given $\mathbf{P}^{1/2}$ \[\mathcal{C} = \left\{ \left(1-y, 2\mathbf{P}^{1/2}\mathbf{x}, 1+y \right): \left\|\left[\begin{array}{c} 1-y \\ 2 \mathbf{P}^{1 / 2} \mathbf{x} \end{array}\right]\right\|_2 \leq 1+y \right\},\] which in turn induces the generalized inequality $\mathbf{P}^{1/2}\mathbf{x} \preceq_\mathcal{C} y$.

Specific Synthetic Control Problems as SOCPs

We illustrate the approach described above for the L1-L2 constraint and the Lasso constraint. To ease notation, we do not consider the regularization to the local geometry of $\Delta$. Note that simplex, ridge, or least squares are particular cases of L1-L2. Throughout this section, we assume that $T_i\equiv T_0$ for all $i\in\mathcal{E}$.

\paragraph{L1-L2-type $\mathcal{W}$.} Consider first the prediction/estimation SC optimization problem, which relies on the following program:

align[align omitted — 485 chars of source]

where, as always, $\geq$ is understood as a component-wise inequality for vectors ($\mathbf{w}\in\operatorname{\mathbb{R}}^{J_0\cdot J_1}$). First, notice that the non-convex constraints $||\mathbf{w}^{[i]}||_1=1,i=1,\ldots,J_1$ can be replaced with the convex constraints $\mathbf{1}' \mathbf{w}^{[i]}=1,i=1,\ldots,J_1$ because of the non-negativity constraint on the elements of $\mathbf{w}$. Then, we can cast (ref) as a SOCP as follows

align*[align* omitted — 801 chars of source]

where $\mathcal{K}=\mathcal{C}_1\times\mathcal{C}_2\times\mathcal{C}_3^{J_1}\times\mathcal{C}_4^{J_1}=\mathbb{R}_+^{J_0\cdot J_1} \times \mathcal{K}_{\tilde{T}\cdot M+1}\times \operatorname{\mathbb{R}}_+^{J_1} \times \mathcal{K}_{J_0 +1}^{J_1}$ is the conic constraint for this program.

For uncertainty quantification, we need to solve the optimization problem underlying (6.5) in the main text. We discuss the lower bound only. Recalling that $\boldsymbol{\beta}=(\mathbf{w}',\mathbf{r}')'$, we have

align[align omitted — 688 chars of source]

We can cast the SC optimization problem in (ref) in conic form as follows:

align*[align* omitted — 930 chars of source]

where $\mathcal{K}=\mathcal{C}_1\times\mathcal{C}_2\times\mathcal{C}_3\times\mathcal{C}_4\times\mathcal{C}_5=\operatorname{\mathbb{R}}_+\times \operatorname{\mathbb{R}}_+^{J_0\cdot J_1}\times\operatorname{\mathbb{R}}_+^{J_1}\times \mathcal{K}_{1+J_0}^{J_1} \times \mathcal{K}_{1+(J_0+KM)\cdot J_1}$ is the conic constraint for this program, $\mathbf{a}=-2('\mathbf{Q}\widehat{\boldsymbol{\beta}} + \mathbf{G}^\star)'$, and $f=\widehat{\boldsymbol{\beta}}'\mathbf{Q}\widehat{\boldsymbol{\beta}} + 2\mathbf{G}^\star\widehat{\boldsymbol{\beta}}$.

\paragraph{Lasso-type $\mathcal{W}$.} We show how to write the QCQP as a SOCP when $\mathcal{W}$ has a lasso-type constraint. In this case, the SC weight construction (3.1) has the form:

align[align omitted — 325 chars of source]

We can write the optimization problem in (ref) as a SOCP of the following form

align*[align* omitted — 688 chars of source]

where $\mathcal{K}=\mathcal{C}_1\times\mathcal{C}_2^{J_1}\times\mathcal{C}_3\times\mathcal{C}_4=\mathcal{K}_{1+T_0\cdot M \cdot J_1}\times \operatorname{\mathbb{R}}_+^{J_1}\times \mathbb{R}_+^{J_0\cdot J_1}\times \mathbb{R}_+^{J_0\cdot J_1}$ is the conic constraint for this program and $\mathbf{z}:=(\mathbf{z}_1',\ldots,\mathbf{z}_{J_1}')'$.

For uncertainty quantification, we need to solve the optimization problem underlying (6.5) in the main text. Here we discuss the lower bound only for brevity. Recalling that $\boldsymbol{\beta}=(\mathbf{w}',\mathbf{r}')'$, we have

align[align omitted — 525 chars of source]

We can cast the SC optimization problem in (ref) in conic form as follows:

align*[align* omitted — 795 chars of source]

where $\mathcal{K}=\mathcal{C}_1\times\mathcal{C}_2^{J_1}\times\mathcal{C}_3\times\mathcal{C}_4\times\mathcal{C}_5=\operatorname{\mathbb{R}}_+\times\operatorname{\mathbb{R}}_+^{J_1}\times \mathbb{R}_+^{J_0\cdot J_1}\times \mathbb{R}_+^{J_0\cdot J_1} \times \mathcal{K}_{1+(J_0+KM)\cdot J_1}$ is the conic constraint for this program, $\mathbf{a}=-2('\mathbf{Q}\widehat{\boldsymbol{\beta}} + \mathbf{G}^\star)'$, and $f=\widehat{\boldsymbol{\beta}}'\mathbf{Q}\widehat{\boldsymbol{\beta}} + 2\mathbf{G}^\star\widehat{\boldsymbol{\beta}}$.

Data Preparation and Software Implementation

In this section, we first describe the variables in the Billmeier-Nannicini_2013_RESTAT (BN, henceforth) dataset and then go through the details of our empirical specification.

Data Description

The original BN dataset contains data on some economic and political variables for 180 countries, over a period of time spanning from 1960 to 2005.\footnote{We downloaded the dataset from the Harvard Dataverse at \url{https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/28699}.} In detail, the variables available in the dataset are:

itemize• real GDP per capita in 2002 US dollars. • enrollment rate in secondary schooling. • population growth. • yearly inflation rate. • the investment ratio (the investment of a country as a percentage of GDP). • an indicator that captures whether the country is a democracy (1) or not (0). • an indicator that captures whether the economy of the country is considered closed (0) or not (1) as developed in Sachs-Warner_1995_BP (henceforth, Sachs-Warner indicator). In particular, the indicator takes value 0 if any of the following conditions is verified: \begin{enumerate}[label=\roman*)] • the average tariff is above 40%; • non-tariff barriers are imposed on a volume of imports larger than 40%; • the country has a socialist economic system; • the exchange rate black market premium is above 20%; • state monopolies control most of the country exports. \end{enumerate}

Despite having six candidate variables to match on, we end up matching on at most two variables because of missing data. Specifically, we always match on real GDP per-capita, whereas in the robustness exercise in Supplemental Appendix (ref) we also match on investment-to-GDP ratio.

A second difference of our final dataset from the original one used in BN is the final pool of countries - treated and donors - on which we conduct the analysis. In particular, we adopt the following criteria to select the countries to be included in our final dataset as either donors or treated units:

enumerate• We restrict the analysis to countries in Sub-Saharan Africa. We define this group by excluding from the analysis North African countries according to the United Nations (UN) classification (Algeria, Egypt, Libya, Morocco, Sudan, Tunisia, and Western Sahara) as well as any additional country that is not covered by the UN classification but is a member of the Arab League (Djibouti, Mauritania, Somalia). • As in Bratton1VandeWalle1997-book, we exclude Namibia because the Sachs-Werner indicator is missing for such a country.

Table (ref) shows the final set of countries we select, together with their treatment date.

table[table omitted — 1,448 chars of source]

Implementation Details

In this section, we describe all the details of the empirical application presented in the main text.

\paragraph{Constraint type.} Our preferred specification uses the L1-L2 constraint, i.e., \[\mathcal{W}_{\mathtt{L1-L2}} = \bigtimes_{i=1}^{J_1} \left\{\mathbf{w}^{[i]} \in \mathbb{R}_+^{J_0}:||\mathbf{w}^{[i]} ||_1 = 1, ||\mathbf{w}^{[i]} ||_2 \leq Q^{[i]}\right\},\] whereas the results in the Supplemental Appendix Section (ref) and (ref) use simplex- and Ridge-type constraints of the form \[\mathcal{W}_{\mathtt{S}} = \bigtimes_{i=1}^{J_1} \left\{\mathbf{w}^{[i]} \in \mathbb{R}_+^{J_0}:||\mathbf{w}^{[i]} ||_1 = 1\right\},\qquad \mathcal{W}_{\mathtt{R}} = \bigtimes_{i=1}^{J_1} \left\{\mathbf{w}^{[i]} \in \mathbb{R}^{J_0}: ||\mathbf{w}^{[i]} ||_2 \leq Q^{[i]}\right\}.\] Table (ref) shows the effective values for $Q^{[i]}, i=1,\ldots, J_1$ that we compute in our empirical application. Further below we explain in greater detail how these regularization parameters are computed in practice.

\paragraph{Selected features.} Our main specification uses only one feature ($M=1$)--the logarithm of real GDP per-capita--, uses the identify weighting matrix, and includes a constant term, that is \[\mathbf{B}^{[i]} =

bmatrix[bmatrix omitted — 61 chars of source]

, \quad \mathbf{C}^{[i]} = \mathbf{1}_{T_i}, \quad \mathcal{R}=\operatorname{\mathbb{R}}^{J_1}, \quad \mathbf{V}^{[i]} = \mathbf{I}_{T_i}, \quad i = 1,\ldots, J_1,\] where $\mathbf{Y}_j = (Y_{j1},\ldots, Y_{jT_i})', j=1,\ldots,J_0$ is the pre-treatment log-GDP per-capita of the $j$-th donor, $\mathbf{1}_{T_i}$ is a $T_i\times 1$ vector of ones, and $\mathbf{I}_{T_i}$ is the $T_i\times T_i$ identify matrix.

In Supplemental Appendix Section (ref), we present results using two features ($M=2$), where we also match on the investment-to-GDP ratio and control for a feature-specific constant term and linear trend, i.e. \[\mathbf{B}^{[i]} =

bmatrix[bmatrix omitted — 113 chars of source]

, \quad \mathbf{C}^{[i]} =

bmatrix[bmatrix omitted — 74 chars of source]

, \quad \mathcal{R}=\operatorname{\mathbb{R}}^{2\cdot J_1}, \quad \mathbf{V}^{[i]} =

bmatrix[bmatrix omitted — 92 chars of source]

,\quad i = 1,\ldots, J_1,\] where $\mathbf{IR}_j = (IR_{j1},\ldots, IR_{jT_i})', j=1,\ldots,J_0$ is the pre-treatment investment-to-GDP ratio of the $j$-th donor, $\mathbf{S}^{[i]}_Y = \mathrm{diag}(\widehat{\sigma}^{-1}_{Y,1},\ldots, \widehat{\sigma}^{-1}_{Y,T_i})$, $\mathbf{S}^{[i]}_{IR} = \mathrm{diag}(\widehat{\sigma}_{IR,1}^{-1},\ldots, \widehat{\sigma}^{-1}_{IR,T_i}),$ with \[\widehat{\sigma}_{W,t} = \left(\frac{1}{J_0-1} \sum_{j=1}^{J_0}(W_{jt} - \overline{W}_t)^2\right)^{1/2}, \quad \overline{W}_t = \frac{1}{J_0}\sum_{i=1}^{J_0}W_{jt}, \quad t=1,\ldots, T_i, \quad W\in\{Y,IR\},\] and $\mathrm{diag}(\mathbf{x})$ yields a square diagonal matrix with the elements of $\mathbf{x}$ on its main diagonal.

\paragraph{Tuning parameters.} Regarding the choice of $Q^{[i]}$, it is well-established that the Ridge regression problem can be equivalently expressed both as an unconstrained penalized optimization problem and as a constrained optimization problem. For simplicity, assume \( \mathbf{C} \) is not included and \( M = 1 \). The two Ridge-type optimization formulations are as follows: \[ \widehat{\mathbf{w}}^{[i]} = \arg \min_{\mathbf{w}^{[i]} \in \mathcal{W}} (\mathbf{A}^{[i]} - \mathbf{B}^{[i]}\mathbf{w}^{[i]})' \mathbf{V}^{[i]} (\mathbf{A}^{[i]} - \mathbf{B}^{[i]}\mathbf{w}^{[i]}) + \lambda^{[i]} ||\mathbf{w}^{[i]}||_2^2, \] where \( \lambda^{[i]} \geq 0 \) is a regularization parameter, and \[ \widehat{\mathbf{w}}^{[i]} = \arg \min_{\mathbf{w}^{[i]} \in \mathcal{W}, ||\mathbf{w}^{[i]}||_2^2 \leq \left(Q^{[i]}\right)^2} (\mathbf{A}^{[i]} - \mathbf{B}^{[i]}\mathbf{w}^{[i]})' \mathbf{V}^{[i]} (\mathbf{A}^{[i]} - \mathbf{B}^{[i]}\mathbf{w}^{[i]}), \] where \( Q^{[i]} \geq 0 \) is an explicit upper bound on the norm of \( \mathbf{w}^{[i]} \). Under the assumption of Gaussian errors, an optimal choice of the regularization parameter \( \lambda^{[i]} \) for risk minimization, as suggested by Hoerl-Kannard-Kent_1975_ridge, is: \[ \lambda^{[i]} = \frac{J_0 (\widehat{\sigma}^{[i]}_{\text{OLS}})^2}{\|\widehat{\mathbf{w}}^{[i]}_{\text{OLS}}\|_2^2}, \] where \( (\widehat{\sigma}^{[i]}_{\text{OLS}})^2 \) and \( \widehat{\mathbf{w}}^{[i]}_{\text{OLS}} \) are the estimates of the residual variance and the coefficients from the ordinary least squares (OLS) regression of $\mathbf{A}^{[i]}$ onto $\mathbf{B}^{[i]}$, respectively. Given the two optimization problems above, there exists a one-to-one correspondence between \( \lambda^{[i]} \) and \( Q^{[i]} \). For example, assuming the columns of \( \mathbf{B}^{[i]} \) are orthonormal, the closed-form solution for the Ridge estimator is: \[ \widehat{\mathbf{w}}^{[i]} = (\mathbf{I} + \lambda^{[i]} \mathbf{I})^{-1} \widehat{\mathbf{w}}^{[i]}_{\text{OLS}}, \] and if the constraint on the \( \ell_2 \)-norm is active, we have \( Q^{[i]} = \|\widehat{\mathbf{w}}^{[i]}\|_2 = \|\widehat{\mathbf{w}}^{[i]}_{\text{OLS}}\|_2 / (1 + \lambda^{[i]}) \). When more than one feature is considered (i.e., \( M > 1 \)), we compute the constraint size \( Q^{[i]}_\ell \) for each feature \( \ell = 1, \ldots, M \), and then choose \( Q^{[i]} \) as the most restrictive constraint to promote shrinkage of \( \mathbf{w}^{[i]} \): \[ Q^{[i]} := \min_{\ell=1, \ldots, M} Q^{[i]}_\ell. \]

table[table omitted — 1,663 chars of source]

\paragraph{In-sample Uncertainty.} In order to quantify the in-sample uncertainty from estimating the SC weights, we need to construct the bounds $\underline{M\mkern-4mu}\mkern4mu _{\mathtt{in}}$ and $\overline{\mkern-3muM\mkern-1.5mu}_{\mathtt{in}}$ on $\mathbf{p}_\tau'(\widehat\boldsymbol{\beta}-\boldsymbol{\beta}_0)$. The following strategy is adopted. First, we treat the synthetic control weights as possibly misspecified, thus estimating both the first and second conditional moments of the pseudo-true residuals $\mathbf{u}$. The conditional first moment $\mathbb{E}[\mathbf{u}\,|\,\mathscr{H}]$ is estimated feature-by-feature using a linear-in-parameters regression of the residual $\widehat{\mathbf{u}} = \mathbf{A} - \mathbf{B}\widehat{\mathbf{w}} -\mathbf{C}\widehat{\mathbf{r}}$ on $\mathbf{B}$ and the first lag of $\mathbf{B}$, whereas the conditional second moment $\mathbb{V}[\mathbf{u}\,|\, \mathscr{H}]$ is estimated with an HC1-type estimator. We then draw $S=200$ i.i.d. random vectors from the Gaussian distribution $\mathsf{N}(0,\widehat{\boldsymbol{\Sigma}})$, conditional on the data, to simulate the criterion function ${\ell^\star_{(s)}(\boldsymbol{\beta}-\boldsymbol{\beta}_0):=(\boldsymbol{\beta}-\boldsymbol{\beta}_0)'\widehat{\mathbf{Q}}(\boldsymbol{\beta}-\boldsymbol{\beta}_0)-2\mathbf{G}_{(s)}'(\boldsymbol{\beta}-\boldsymbol{\beta}_0)}$, $s=1,\ldots, 200$, and solve the following optimization problems

align*[align* omitted — 569 chars of source]

where $\Delta^\star$ is constructed as explained in Section 6.1. Finally, $\underline{M\mkern-4mu}\mkern4mu _{\mathtt{in}}$ is the $(\alpha_1/2)-$quantile of $\{l_{(s)}\}_{s=1}^S$ and $\overline{\mkern-3muM\mkern-1.5mu}_{\mathtt{in}}$ is the $(1-\alpha_1/2)-$quantile of $\{u_{(s)}\}_{s=1}^S$, where $\alpha_1$ is set to 0.05.

\paragraph{Out-of-sample Uncertainty.} In order to quantify the out-of-sample uncertainty from the stochastic error in the post-treatment period, we need to construct the bounds $\underline{M\mkern-4mu}\mkern4mu _{\mathtt{out}}$ and $\overline{\mkern-3muM\mkern-1.5mu}_{\mathtt{out}}$ on the out-of-sample error $e_\tau$ (associated with the $\tau$ prediction). We employ the non-asymptotic bounds described in (6.6), assuming that $e_\tau-\mathbb{E}[e_\tau|\mathscr{H}]$ is sub-Gaussian conditional on $\mathscr{H}$. Then, we take \[\underline{M\mkern-4mu}\mkern4mu _{\mathtt{out}}:=\mathbb{E}[e_\tau|\mathscr{H}]-\sqrt{2\sigma_{\mathscr{H}}^2\log(2/\alpha_2)} \qquad\text{and}\qquad \overline{\mkern-3muM\mkern-1.5mu}_{\mathtt{out}}:=\mathbb{E}[e_\tau|\mathscr{H}]+\sqrt{2\sigma_{\mathscr{H}}^2\log(2/\alpha_2)},\] We set $\alpha_2=0.05$, and the conditional mean $\mathbb{E}[e_\tau|\mathscr{H}]$ and the sub-Gaussian parameter $\sigma_{\mathscr{H}}$ are parametrized and estimated by a linear-in-parameters regression of the pre-treatment residuals on $\mathbf{B}$.

Finally, the prediction intervals for the counterfactual outcome and the treatment effect of interest are given by \[\left[\mathbf{p}_\tau'\widehat{\boldsymbol{\beta}}-\overline{\mkern-3muM\mkern-1.5mu}_{\mathtt{in}}+\underline{M\mkern-4mu}\mkern4mu _{\mathtt{out}}; \: \mathbf{p}_\tau'\widehat{\boldsymbol{\beta}}-\underline{M\mkern-4mu}\mkern4mu _{\mathtt{in}}+\overline{\mkern-3muM\mkern-1.5mu}_{\mathtt{out}}\right] \quad\text{and} \quad \left[\widehat{\tau}+\underline{M\mkern-4mu}\mkern4mu _{\mathtt{in}}-\overline{\mkern-3muM\mkern-1.5mu}_{\mathtt{out}};\: \widehat{\tau}+\overline{\mkern-3muM\mkern-1.5mu}_{\mathtt{in}}-\underline{M\mkern-4mu}\mkern4mu _{\mathtt{out}}\right],\] respectively.

\paragraph{Other assumptions.} Throughout all our specifications, we maintain the assumptions that (i) there is no anticipation of the treatment and (ii) $\mathbf{A}$ and $\mathbf{B}$ form a cointegrated system. When (i) is relaxed to allow for anticipation results remain qualitatively the same.