Econ Papers Daily / 期刊论文
估算加利福尼亚州高速公路系统中驾驶员的时间成本函数
Estimating drivers’ time cost function for California’s freeway system
摘要
This paper estimates drivers’ time cost function, a causal relationship between traffic volume and travel time, representing the California freeway system as a whole. For identification, we instrument for traffic density using California’s March 2020 COVID-19 lockdown — a large shock to travel demand — interacted with county-level work-from-home potential, which captures each area’s exposure to the shock. For estimation, we use freeway–county-level aggregates of traffic variables, constructed from the extensive network of traffic sensors on California freeways. We find that OLS estimates are substantially larger than IV estimates. After ruling out alternative explanations, we attribute this gap primarily to a reverse-causality bias that follows mechanically from the capacity constraint of freeways: observed traffic volume is censored at capacity, steepening the OLS slope. Based on our preferred IV estimate, optimal congestion tolls average about 10 cents per vehicle-mile under typical conditions, rising to about 36 cents under extreme congestion. Such tolling would yield annual welfare gains of up to $0.74 billion.
中文摘要
本文估计了驾驶员的时间成本函数，即交通流量与通行时间之间的因果关系，并将加利福尼亚州高速公路系统作为一个整体加以刻画。为进行识别，我们利用加利福尼亚州2020年3月COVID-19封锁——一次对出行需求的巨大冲击——与县级居家办公潜力的交互项作为交通密度的工具变量，其中县级居家办公潜力刻画了各地区对该冲击的暴露程度。为进行估计，我们使用由加利福尼亚州高速公路上广泛分布的交通传感器网络构建的高速公路—县级交通变量加总数据。我们发现，OLS估计值显著大于IV估计值。在排除其他替代解释后，我们将这一差距主要归因于一种反向因果偏误，该偏误在机制上源于高速公路的容量约束：观测到的交通流量在容量处被截断，从而使OLS斜率变陡。基于我们偏好的IV估计，在典型条件下，最优拥堵收费平均约为每车英里10美分，在极端拥堵下升至约36美分。此类收费将带来每年最高7.4亿美元的福利收益。