# volume 1, pg. 426, question 22

Can someone explain why A is the correct answer? just can’t wrap my head around it.

Re: Statement 1: t & F stats biased because std errors are underestimated with cond’l hetero. Since the size of the error is correlated (conditionned on) the value (level/size) of the Indy, the Indly ITSELF explains a significant portion of the squared residual. (review page 376) Re: Statement 2: True - if you leave out a key/critical variable, then the coefficients that remain will appear to explain more of the Dep variable than they really do. Remember, in mult. regr. each coeff explains the variation of the Dep NOT explained by the OTHER Indys. Leave out an important Indy (if its correlated to the Indys that are in the equation) and the remaining estimates of coeffs are biased & inconsistent and the std errors are inconsistent. (Review pp 390-391 Table 12 - without ln(Mkt cap) the coeff of ln(market makers) is -1.6629…but ln(market cap) has a signif influence on the bid-ask spread/price…when it’s included in the regression, the coeff. on ln(mkt makers) reduces to -0.2790…the influence of the #of market makers “net of” the influence of ln(mkt cap). (In general Mkt cap has a signif influence on bid-ask spreads and prices so we must include it in the regression.) Think of trying to figure out the temperature of a bucket of water sitting outside (Dep vrbl), using month of the year (Indy vrbl) but without accounting for air temperature (Indy vrbl). When you add air temp to the equation, the “influence” of the month (regr coeff) on the temp of the water will change - it will decrease. Hope I haven’t confused you more.

good example resolute.