Econometric model. We conducted a difference-in-differences regression to measure the effect of increasing competi- tion in the auction stage, distinguishing the alternative mechanisms discussed in the hypotheses formulation. Specifically, we sought to estimate the causal effect of the competitive treatment condition on submitted and awarded bid prices in the auction stage and posted and transaction prices during the operation stage. Auction stage: Analysis of bids. First, we focus on studying the effect of the new FA design and the competitive treatment on the auction bids. As we explained above, bid prices correspond to the ceiling prices that the suppliers can charge during the operation of the FA. Let the index t ∈ {Old, New} represent the old (2014) and the new (2017) Food FAs, respec- tively. Let bijrt represent the bid offered in FA t by supplier j to provide product i in region r, irt including shipping.4 Define Bmed as the median bid across all supplier bids submitted in the FA for that product-region-FA combination. In calculating this median bid price, we considered alternative methods to for discarding extreme prices that could be generated by bidding mis- takes or unrealistically aggressive bidding. For the main results we used a modified Tukey rule. To make the bids for the 2014 and 2017 Food FAs comparable, all prices for 2014 were adjusted using the CPI dood price index. In addition, we normalized bid prices so that they represented bid prices per unit. See Appendix A for further details on all of this data preprocessing. The following difference-in-differences specification is used to estimate the effect of the com- petitive treatment condition —denoted by the indicator variable Compi— on bid prices: irt log(Bmed) = δr + γi + αNewt + βNewt × Compi + εirt , (1) where Newt is an indicator equal to one for the new FA and δr is a region fixed effect. The sample is comprised of all products i that were matched across the old and new FAs based on similar attributes, and the regression includes a product fixed effect γi. The coefficient α measures the average differences in bid prices between the old and new FAs for the noncompetitive baseline group. The key parameter of interest is β, the coefficient capturing the incremental change in bid prices for the auctions in the competitive treatment condition in the new FA. irt Recall that Bmed, which is the median across all submitted bids, measures whether the com- petitive treatment affects the bidding behavior of the ...
Econometric model. The dexzriptive featurex of the data on a zountry’x partizipation in MEAx over time dixplay a xtrong perxixtenze. In any given period, the number of MEAx a zountry partizipatex in hax a xtrong impazt on itx xubxequent involvement in MEAx. Xxxxx, apart from fundamental 9Results are available from the authors upon request. Table 1.1: Statixtizx of balanzed data Variable Obxervationx Mean Std. Dev. Min Max YEAR Œ93† 1983 13.†660 1960 2006 NUMBER OF MEAx (ys5) Œ93† 3†.031† 36.ŒŒŒ1 0 222 LGDP Œ93† 23.6k3† 2.132† 1k.896k 30.06†6 LPOP Œ93† 9.3836 1.Œ†36 6.208† 1Œ.089† TRADE LIBERAL Œ93† 1.3396 1.8038 –Œ.11†Œ †.2†Œ2 INVEST LIBERAL Œ93† 9.93kk 18.896† 0 131 LDC Œ93† 0.1603 0.3669 0 1 PFI Œ93† 3.8†2Œ 1.9†89 1 9.6 CO2 EMISSIONS Œ93† 3.k618 Œ.Œ†k8 –0.019k 2k.k66Œ AGRLAND Œ93† Œ2.ŒŒ08 21.3902 0.62k8 91.k8†0 ezonomiz, politizal, or environmental determinantx of MEA memberxhip, a zountry’x MEA hixtory xhould be allowed to play a role.10 Thix feature may be zaptured by the inzluxion of a lagged dependent variable in the ezonometriz model. We do xo by following Blundell, Xxxxxxxx, and Xxxxxxxxxx (2002) to model the dynamizx of the number of MEAx a zountry partizipatex in ax a linear feedbazk model (LFM). The LFM axxumex that the zonditional mean of a dependent zount variable ix linear in the hixtory of the prozexx.11 Let ys5 denote the number of MEAx zountry s, s = 1, ..., N , ix a member of in year f, f = 1, ..., F . Further, let xs5 reprexent a veztor of K explanatory variablex. The zonditional 10If history matters, cross-sectional evidence on the determinants of MEA participation is difficult to interpret since the estimated responses may reflect short-run or long-run effects. 11For a good overview article of GMM for panel count data models see Windmeijer (2008). mean in the LFM ix then defined ax E (ys5|ys5—1, xs5, vs) = rys5—1 ‡ exp (xjs5Ø) vs (1.Œ) = rys5—1 ‡ µs5us, where us ÷ exp(ys) ix a permanent xzaling faztor for the individual xpezifiz mean, and r and Ø are parameterx to be extimated. The LFM zan be motivated ax an entry–exit prozexx with the probability of exit equal to (1 — r). Note that µs5us ix non–negative, xo that the mean value for ys5 ix bounded below by rys5—1. To avoid ximultaneity biax every explanatory variable enterx in their firxt lag into our regrexxionx. By thix meanx, we are able to zanzel out the Xxxxxxx feedbazk xyxtem between the number of MEAx and trade and invextment treatiex.12 Uxing the lagged valuex of the explanatory variablex reliex...
Econometric model. A zomplization in extimating the ezonometriz model arixex ax the within group mean xzaling extimator will be inzonxixtent for xmall F , xinze the lagged dependent variable ix predetermined 22. For extimation by the generalixed method of momentx (GMM), the LFM may be quaxi–differenzed (xxx Xxxxxxxxxx, 199k) with qs5 = Σy —s5 ¡=1 µs5 r¡ ys5—¡ — Σy —s5—1 ¡=1 r¡ ys5—1—¡
Econometric model. The zount of a zountry’x ratifization of MEAx over time dixplay xtrong perxixtenze. For every zountry and in every year, the number of MEAx a zountry partizipatex in followx a quite nize MA(1) term with xlowly dezreaxing zorrelation over time and a xharp dezline in partial autozorrelation after a high value in the firxt period. Thix ix true regarding all MEAx unzlaxxified ax well ax a xpezifiz MEA zluxter. Xxxxx, a zountry’x MEA hixtory xhould definitely be uxed for explaining the zurrent number of MEAx of a zountry by inzluding a lagged dependent variable in the ezonometriz model while xolving for rexulting endogeneity. Due to thix, we apply Blundell, Xxxxxxxx, and Xxxxxxxxxx (2002) to model the dynamizx of the number of MEAx a zountry ratifiex ax a dynamiz linear feedbazk model (LFM). Thix meanx, we make uxe of the feedbazk information of lagged valuex of the dependent variable. Here the zonditional mean of a dependent zount variable ix axxumed to be linear in the hixtory of the prozexx.4 4see also Windmeijer (2008). The zonditional mean in the xtandard LFM ix defined ax E (ys5|ys5—1, xs5, vs) = rys5—1 ‡ exp (xjs5Ø) vs (2.1) = rys5—1 ‡ µs5us, where ys5 denotex the number of MEAx zountry s, s = 1, ..., N , hax ratified in year f, f = 1, ..., F . xs5 reprexentx a veztor of K explanatory variablex and us ÷ exp(ys) ix a permanent xzaling faztor for the individual xpezifiz mean. The parameterx r and Ø are to be extimated. Ax the number of MEAx ix predetermined, i.e., zorrelated with paxt xhozkx but not zurrent onex, E (xs5us5‡¡) = 0, j ≤ 0, E (xs5us5—s) = 0, ‹ ≤ 1, we zan inxtrument the zontemporaneoux valuex with itx xezond lagx and thux xolve the endogeneity problem. Azzording to Windmeijer (2008) the LFM zan alxo be motivated ax an entry–exit prozexx with the probability of exit equal to (1 — r). Moreover the mean value for ys5 ix bounded below by rys5—1 ax µs5us ix non–negative. In a xezond xtep, to imply the effeztx of zluxterx on eazh other, we extend the LFM uxing the xuperxzript s for zluxter. s5 s5—1 E .yc |yc , xs5, vsΣ = ryc ‡ zyCƒ=c ‡ exp (xj Ø) vs (2.2) s5—1 s5—1 s5 = ryc ‡ zyC c ‡ µs5us s5—1 s5—1 Here, the parameter z ix to be extimated additionally. It meaxurex the impazt of the lagged number of MEAx of all other zluxterx or the bilateral impazt of the lagged number of MEAx of one other zluxter. We zompared different extimation methodx where it turned out that, in zontext of MEAx, the rexultx of the one–xtep extimator uxing Xxxxxxxxxx moment ...
Econometric model. Since the complex feature of the air transportation industry, underlying the continuous economic fluctuation and environmental changing, the previous method for air travel forecasters like trend extrapolation have not been impressive and not well applied for the practice research. So it is necessary to bring causality into the analysis which is called econometric models and which we used for our study. This model not only predicts air travel but also determines the impact of changes within the economic and operating environment on air travel. Also these models relate the traffic to underlying economic parameters like income of passengers or more readily available proxies for them, usually this method is calibrated by multiple regression of historic data to derive elasticity of demand, for instance, the change in demand due to the one percent change in one independent variables affecting the demand. Accordingly, the multiple regression analysis enable to link future growth in air travel demand in a specific area with expected developments of causative factors (Xxxx and Xxxxxxxxxx, 2000).
Econometric model. We employed a DID analysis and adopted a probit model to estimate the marginal effect of BCCTPA on early stage of cancer at the time of Medicaid enrollment by comparing the change before and after the implementation of BCCPTA for women with breast cancer versus those with a control cancer, controlling for other related factors. The model was shown: probit(S ) = β0 + β1 BCCPTA + β 2 Breast + β3 BCCPTA * Breast + γX i + ηCi t + δTt + ε i t where S = stage of cancer at Medicaid enrollment (S=1 if it is an early stage, S=0 if it is a late stage); BCCPTA = a vector of binary variables representing the time from the implementation of BCCPTA measured by months; we categorized these monthly time intervals into five categories: initial implementation month, 3, 6, 9 months and a year post- BCCPTA implementation; Breast = a binary variable where 1 = breast cancer and 0 = control cancers; BCCPTA*Breast = is a vector of interactions of the five categories of post- BCCPTA period interacted with breast cancer. We also controlled for Xi = a vector of individual covariates, including age at enrollment, race/ethnicity, marital status, and residence, Cit= a vector of county factors which might change over time, including county with a teaching hospital, percentage of small firms (＜10 employees) and percentage of Medicaid recipients within the residency county; and Tt = year dummy to control for other time trends factors. To interpret the DID model, β1 was the estimate of the trend of enrollment in Medicaid for women with all cancers (including both breast and control cancers) before and after the implementation of BCCPTA. Because of intensive public health campaigns and mature noninvasive diagnoses techniques, breast cancer is more often found at early stage than some other cancers, for example colorectal cancer(Xxxxxxxx and Xxx 1995; Xxxxxxx 2007). We noted that, β2 was the estimate of the baseline difference between breast and control cancer patients in early stage of Medicaid enrollment. In turn, β3 was the estimate of the DID effect, which was the policy influence of the BCCPTA on the change of breast cancer versus control cancer patients into Medicaid at an early stage. Besides, we examined five policy periods since it was possible that there was a group of long-term uninsured women with cancer in the pre-BCCPTA period who, due to the new availability of coverage post-BCCPTA, would enroll in Medicaid. This potential ‘back-log’ of women with late stage of cancer could im...
