Prices and market shares are the de

Prices and market shares are the dependent variables in all of the hypotheses, but since prices and market shares are codetermined by the inter-action of demand and supply equations, the separate effects of demand and supply must be separated by means of a simultaneous-equation estimation (Judge at al„ 1982: 337-366). For this reason, both Robinson and Fornell (1985) and Robinson (1988) used a simultaneous- equation methodology in their estimates of the magnitude of first-mover and early-mover advan-
Likewisc, this study also uses a simultaneous- equation methodology, namely Hatanaka’s (1976) procedure for estimating simultaneous-equation systems with both lagged-dependent variables as regressors and serially autocorrelated error terms. A first-order autoregressive (AR1) error term is included in each equation of the model because serially autocorrelated errors are likely to occur in regression models using time series data that exhibit gradual period-to-period changes. A lagged dependent variable is included as a regressor in each equation of the model in order to guard against the possibility of having a ‘reverse-causality’ effect that biases the results. A common problem in regressions of economic data is determining the direction of causality in the relationships between variables. In a regression of variable Y on variable X, it is not always clear whether the observed slope coef¬ficient represents the effect of X on Y, the effect of Y on X, or some combination of both. If Y has a causal effect on X, but X does not have a causal effect on Y, then a regression of Y on X will yield a significant slope that one might misinterpret as the effect of X on Y even though it is actually due to the effect of Y on X. One common approach to solving this problem, embodied in the Granger-Sims causality test (Granger, 1969; Sims, 1972; Doan, 1992: section 6.4), is to conclude that a causal relationship between two variables exists only if the coef¬ficient is statistically significant when a lagged- dependent variable is also included as an inde¬pendent variable in the regression. However, the usual estimation procedures for regression with an autocorrelated error term (e.g., Prais-Winsten, Cochrane-Orcutt, Hildreth-Lu, Yule-Walker, and Beach-MacKinnon) are inconsistent and therefore inappropriate when one of the regressors is a lagged-dependent variable. The only known consistent and asymptotically efficient procedures for estimating a simultaneous-equation system with both a lagged-dependent regressor and a serially autocorrelated error term are describedby Hatanaka (1976), so Hatanaka’s procedure was used in this study

Kết quả (Việt) 1:

Giá cả và thị phần đang biến phụ thuộc trong tất cả những giả thuyết, nhưng kể từ khi giá cả và thị trường cổ phiếu được codetermined bởi các hành động giữa nhu cầu và nguồn cung cấp các phương trình, những tác động riêng biệt của nhu cầu và nguồn cung cấp phải được tách bằng phương tiện của một ước lượng phương trình đồng thời (thẩm phán tại al"năm 1982: 337-366). Vì lý do này, cả Robinson và Fornell (1985) và Robinson (1988) sử dụng một phương pháp đồng thời phương trình của họ ước tính độ lớn của đầu tiên mover và sớm triển khai nhanh advan-Likewisc, this study also uses a simultaneous- equation methodology, namely Hatanaka’s (1976) procedure for estimating simultaneous-equation systems with both lagged-dependent variables as regressors and serially autocorrelated error terms. A first-order autoregressive (AR1) error term is included in each equation of the model because serially autocorrelated errors are likely to occur in regression models using time series data that exhibit gradual period-to-period changes. A lagged dependent variable is included as a regressor in each equation of the model in order to guard against the possibility of having a ‘reverse-causality’ effect that biases the results. A common problem in regressions of economic data is determining the direction of causality in the relationships between variables. In a regression of variable Y on variable X, it is not always clear whether the observed slope coef¬ficient represents the effect of X on Y, the effect of Y on X, or some combination of both. If Y has a causal effect on X, but X does not have a causal effect on Y, then a regression of Y on X will yield a significant slope that one might misinterpret as the effect of X on Y even though it is actually due to the effect of Y on X. One common approach to solving this problem, embodied in the Granger-Sims causality test (Granger, 1969; Sims, 1972; Doan, 1992: section 6.4), is to conclude that a causal relationship between two variables exists only if the coef¬ficient is statistically significant when a lagged- dependent variable is also included as an inde¬pendent variable in the regression. However, the usual estimation procedures for regression with an autocorrelated error term (e.g., Prais-Winsten, Cochrane-Orcutt, Hildreth-Lu, Yule-Walker, and Beach-MacKinnon) are inconsistent and therefore inappropriate when one of the regressors is a lagged-dependent variable. The only known consistent and asymptotically efficient procedures for estimating a simultaneous-equation system with both a lagged-dependent regressor and a serially autocorrelated error term are describedby Hatanaka (1976), so Hatanaka’s procedure was used in this study