Table A1.

Pearson’s correlations between the model variables in one of the simulated datasets, showing some of the time points used for model-fitting.

	IVx	IVy	X100	X101	X125	X150	Y100	Y101	Y125	Y150
IVx	1
IVy	0.2500	1
X100	0.2834	0.2227	1
X101	0.2834	0.2227	0.8490	1
X125	0.2834	0.2227	0.1581	0.1640	1
X150	0.2834	0.2227	0.1087	0.1092	0.1581	1
Y100	0.2227	0.2834	0.6877	0.6904	0.1532	0.1038	1
Y101	0.2227	0.2834	0.6904	0.6877	0.1591	0.1042	0.8490	1
Y125	0.2227	0.2834	0.1532	0.1591	0.6877	0.1532	0.1581	0.1640	1
Y150	0.2227	0.2834	0.1038	0.1042	0.1532	0.6877	0.1087	0.1092	0.1581	1

Note. The simulated time-series was stationary over the time points used for model-fitting (i.e., from $T=100$ to $T=150$), as indicated by the correlations and between the instrumental variable (IVx) and X_i ($r=0.2834$), IVx and Y_i ($r=0.2227$), and the cross-sectional correlations between X_i and Y_i ($r=0.6877$). The shown correlations are based on a time-series with the direct effect on IVx on X, $b_X=0.08;$ the direct effect on IVy on Y, $b_Y=0.08;$ the first-order autoregressive coefficient (AR1) for X, $b_{X2X1}=0.7;$ the AR1 for Y, $b_{Y2Y1}=0.7;$ the first-order causal effect of X on Y, $b_{YX}=0.2;$ the first-order causal effect of Y on X, $b_{XY}=0.2;$ the cross-sectional correlation between the residuals of X and Y, $r_{\mathit{exy}}=0.2;$ and the correlation of IVx and IVy, $r_{IV}=0.25.$