Supply Dynamics: The Case of U.S. Broilers and Turkeys
Supply Dynamics: The Case of U.S. Broilers and Turkeys
复制标题
供应动态：美国肉鸡和火鸡的案例
DOI：
10.2307/1240650
复制
发表时间：
1982
期刊：
影响因子：
--
通讯作者：
Stanley R. Johnson
中科院分区：
文献类型：
--
作者：
J. Chavas;Stanley R. Johnson
供给动态在农业经济学中受到了相当多的关注（Nerlove 1979；Askari 和 Cummings）。许多计量经济学模型都包含了允许期望状态和实际状态不同的调整机制。通常，调整过程被指定为分布式滞后模型（Griliches）。常见的期望公式是 Nerlove 自适应模型，它假设在每个时期预期价格的变化量与最近观察到的预测误差成比例，这意味着分布滞后呈几何下降（Nerlove 1958）。该模型在计量经济建模中变得流行，部分原因是它与最优预测中的统计理论紧密结合。导致相同简化形式（误差项结构除外）的另一个假设是部分调整模型（Griliches，Fromm）。它指出，在任何特定时间段内，仅完成“所需”调整的某些固定部分。它假设存在一些调整成本来解释观察到的响应惯性。更复杂的期望公式或部分调整机制会导致更复杂的滞后结构。不幸的是，经济理论对模型构建者的调整过程几乎没有提供指导。具体形式通常由每个研究者任意选择。已经应用了一些“灵活的”滞后分布（例如 Almon、Shiller）。然而，仅从数据中区分各种可能的滞后模型存在统计上的困难。格里利奇写道
Supply dynamics have received considerable attention in agricultural economics (Nerlove 1979; Askari and Cummings). Many econometric models incorporate adjustment mechanisms that allow desired and actual states to be different. Usually, the adjustment process is specified as a distributed lag model (Griliches). A common expectation formula is the Nerlove adaptive model, which assumes that at each period the expected price is changed by an amount proportional to the most recently observed forecast error, implying a geometrically declining distributed lag (Nerlove 1958). This model has become popular in econometric modeling partly because it ties in well with statistical theory in optimal prediction. Another hypothesis leading to the same reduced form (except for the error term structure) is the partial adjustment model (Griliches, Fromm). It states that only some fixed fraction of the "desired" adjustment is accomplished within any particular time period. It assumes the existence of some adjustment costs to explain the observed response inertia. More complex expectation formulations or partial adjustment mechanisms lead to more complicated lag structures. Unfortunately, economic theory provides little guidance to model builders on their adjustment processes. The particular forms are normally selected arbitrarily by each investigator. Some "flexible" lag distributions have been applied (e.g., Almon, Shiller). However, there are statistical difficulties in distinguishing among various plausible lag models from the data alone. Griliches writes