12271403 = Introductory Econometrics-CBCS-Core B.A. (Hons.) Economics II Year 4 Name of the Paper Name of Course : : Semester i Duration : 3 Hours Maxirnum Marks: 75 aeton : 8Hours Maxirnum Marks : 75 ‘Answer any five questions out af Seven. All questions carry equal marks Q. 1. State whether the following statements are Truc or False. Give reasons foe your answer. 6x3=15 (a) In a regression model In ¥; = 0, + BX, + it By is multiplied by 100 we obtain the gowth rate estimate of Yi. 1 Ans. In ¥, = By + Bp x, + uy Lay _ combs 22 = 00 8,% Hence, True (6) In regression through’ origin models the conventionally computed r? may not be meaningful. Ans. True, Further R? for regression through origin is same as R? for regression with mean values of ¥, and X's equal to zero. | oye i | (©) Insimple regression model ¥;= 8, + p:X; + up the OLS estimators and §2 each follow normal distributi n only if u, follows normal Ans. True, one assumption of classical linear regression model is that 4, ~ NG. c*).This induces ¥, to follow normal distribution whieh in turn makes ¥, to follow normal and in'fitted regression 1 Ie. Y= B+hyx, / | 1 + By and Gy also follow normal distribution, (d) P-value of a test statistic is equal t6 the level of singnificance, Ans. True, P-value means the level of marginal significance with a statistical pothesis test representing propability of occurrence of a given event, Alternatively. P-value is used as an alternative to rejection points to provide smallest level of significance at which null hypothesis would be rejected. Thus, Pa ue is the exact level of significance of (stati (©) Ifthe estimate of slope coefficient in a bivariate regression is zero, the measure of coefficient of determination is also zero. Ans. True, because if estimate of slope coefficient in a bivariate regression is zero, then it means that thefe is no significant relationship between explanatory and explained variable and thus, coefficient of determnination is also zero, (498) , 7 496 ‘Steam: B.A. (Hons.) Economics Il Year (Sem) @. 2. (@) You have the following information : EX = 1680, 5Y'= 1110, EXY = 204200, 22° = 318400, 9805 n= 10. Assume all assumptions of CLRM are fulfilled, Obtain, & @ By and By (ii) Bxtablished 95% interval for the propulation slope cotfcy i) Re Ans. ZX= 1680, Y= 1110, EXY= 204200, EX*= 315400, £¥°= 13539, i, 2 2-2?) a @ 8 = EX, - XP i, = 208200-10068)0111) * © 1400-10068)" = M49 ~ 829 fe = 0.834 approximately) and fh = PG, =111-16800.534) = 21.23 (approx) ) 95% C.. for By => By +8.0. toons 02 0.534 + 2.206) ) Re (iri? nEXY-EX EY, kn EX? 2 xP EYE -(EY) (10)(204200)~ (1680)(11 0) * |10¢315400) - 680 Jf10093309)—110)") = 0.9688 (approx) R? = 0.938 (approx) 148sindents in @ cortain exam are known Tobe im that it is possible to increase the avernge an addi £ ith additional study material would rewatt 29. Let denote the true average serene Mi material is Used. ae Sve 280 appropriate null and alternative hypothests? ata rpote the sample AVeTAKe Score for 25 randomly £ devomsider the test procedure with test statisste % sf g2 77.5. What is the propbability distribution of real rue ? What is the probability of type I tic wHen ing procedure in (ii) what is the probability of the true average of students when additio gist ions 1 material Hy: = 75 (The claim that additional material would improve average score is not true) F Hy: > 75 (Claim is true) Ap wwe Xegbe tho 25 selected students from normal distribution “ay guandard deviation c= 9. Then the sarnple mean score ith expected value jig = and standard deviation Tere +: = 921,60 when Hy is true Hy = 75 60 only an F value that is as ie # ould fi Hy. The oe a= P (Hg is rejected when it is true) 2 pc =77.9 when XN with ug =75, 0g = 18) tof 729 = 19f 78278 “1-9(1.6n) 20.9463 P(Tyep II error when u = 80) Ply in not rejected when it is false a8 1 =80) = p(£<77.9 when 2~N with "= 60, op rte , 719-8 (Ai = 0.0357 398 (©) Ina regression mod ‘actual ¥,is equal tathe mea Ans es To prove ‘ where From normal equation (), a Now from (9) 6 oo Ye Row od BY EV +Eu, from (i), we have £u,=0 ry= ry, Dividing both sides by n gives Pak 8 (2) Consider the following simple regression mode! where price is the housing price assess and is the assessment of housing prices. The estimated equation f =~ 14.47 ~ 0.976 assess . = (16.27) (0.039) n= 88, SSR = 165644.51, 7 = 0.820 @_ How will you test the vonstraints by = 1 and b= 0in the above regression if you are given the SSR in the restricted model a Conduct the necessary test(s) at 1" level of significance and give your :) Suppose now that the estimated model is = rie = Bp +B, assess + f, lotsize + fy sqrft +, barms + u where Jotsize = the sixe of the lot sqrft = the square footage bdrms = the number of bedrooms 150 deg aCC-10: Introductory Econometrics (May 2019) 499 The R® from estimating this model using the same 88 houses is 0.829. , Test at 1% level of significance that all partial slope coefficients are equal to zero. ; 2 Ans. P= -14.47+09760 = 06.27) 0.049) n= 88 SSRyp = 16564451 | R= 0,820 P housing prices @: assement of housing prices. o Hy:B)= 0 and B,=1 (Restriction is true) HyiBo#00rfi#1 Restriction is not true). SSR of restricted model is given as « SSR = 209446,99 ‘The F-statistics is given as : (SSRy ~SSRyq)/m ig COR ESE) ee a Be“ GSon) ini) FM (AB) where m : number fresticitions (= 2) ni: number of observation (= 68) ‘es number of parameters in unrestricted model including intercept (= 2) n-K= 88-2=86 pz (209448.99-165644.51)/2 cs . a (209478.99)/80 i. = 8.99 (approx) >t | Which is greater than the critical values of F dintribittion with 2 and a6 | degrees of fredom in numerator and denominator, as { Frage € (4.98, 4.82) Cone. «. Hyis rejected as calculated F valuelstatistics lies in rejection region and thus we reject the constraints of B= 1 and fly = 0. w P= Bo+Bio+Byl+fystBotu n= 88 R= 0.529 Ho, = Bp = By =By=0 (all partial slope coefficients | . j H, = atleast one of hem is non-zero i i ‘Vo test joint hypothesis, we use F-statistics RP k=) ' 2 P° CRG “F, 151 300 eames B.A. (Hons.) Economics Il Year (Sem-4) = pe (0.829)14 * @=0.829)/(68~4) = 101.81 (approx) + Which is highly. significant. Thus we can reject Hy and conclude hay cone of the partial slope coefficient is non-zero (b) Let ¥~ N (y, 0%), Consider two independent random, of observation on X. The samples are of size n, and n, wnt Ai and X, respectively. Two estimators of the population mie proposed: : f ae Nite gp mk tml, BE Check whether these estimatars are un! Ans, X~ Nyc) sed and calaculate thy, mX+ ney and a= “Ai tmeke Let true population mean be EX) = y To check biasness EXy EX, 2G) = B]-L a Eq) = o[teetu rn Ea) + BCR) estimator ofp. . whereas, sop = oft] [mn EX n, Ea] z—.._ | ma= EH) =u \ cod estimater of { sat wi varias varti) ‘= fork, + a Veri) = a) bs saa} {anf an(X;)+ VartX,)* — 183 7 Kyte) niVard Nyce xy] t - rece. +Var(X,))] Lyfe 4} a Sagat emt) 1 ox Tym Var (f) = feel ‘Similarly, Var (fi) = Vor [hie cal eal a [ Et] mt ~ (= way Wary) Vere, )) (Ce Var (ax + b) = a* Var (X)) 7 (: Var (ox 8) =? Var OO) (= ig | [MOE Hy at yd Vay 0X ot Kyl] [Wank + Varex,)+ 1+ VartX,))+ Vani )+ VartX) 0.6 VarlX a +var(X,)| [Soca Var O24 Var + Var )~2 Cov and Cov, X= minis oacs [snd.coe 2) 0 2 wl tects lactic “Galt sete ea] val Ino® + myo" \ -( | i mre (©) For each of the following pi = Var @) = irs of dependent (¥) and 184a CC-10: Introductory Econometrics (May 2019) 503 independent variables (3), pick the most appropriate funetional form. Explain the reason for your answer 6 “"G) Y= demand for food X= price of food (ii) Y= AFC of production X= output (ii) ¥= Population in Idian X= time Ans, () Y= Bot Py Xe where u, error (or disturbance) term, Since, there is a linear relationship between demand for food and price of food. Since food is a normal good thus, p> 0 but P< 0. . a Since there is non-linear inverse relationship between AFC of production and output. Thus, appropriate form might be 4, ye por Show absolute amount of population may depend on contra Policy makers are more interested in vate of the appropriate form would be In¥= py +PX+u Q. 4. (a) In a repression of average wages (JW) on the number of employes (3) for a random sample of 30 firms, the following results were obtained : 6 Regression 1 fr 7.5 + 0.009 N t= (16.10) R* = 0.90 1 = 0.008+7.8 4 Regression 2 t 1 = (14.43) (76.58) R* = 0.99 () How would you interpret the two regressions? i) What might be the reason for transforming 2? What assumption ‘ariance in going from Regression Lto 188 504 ene: B.A. (Hons.) Economics (id) Gan you relate the slones and intercepts ofthe two mag (iv) Can you compare the R* of the twoinodels ? Give reason," w 1 @ Regression 1 W = 7.5+0.009N we 0.0c9 aN ‘Thus, the slope coeificient shows that if number of employees increases by unit then W ie., average wages will increse 0.009. The intercept ecefficient x highly siginaficant and the intercept value of 7.5 suggest that even if number ct employees fall to zero, the averge Wages would be equal to 7.5, The high value of R might suggest that the model ie a good fit. 1 [= (14.43) 07.58) ee) = -2(4,) 8 (ss highly significant, thus we ean reject null of employees is not proportional inversely to average intercept value of 0.008 suggests that if number of employers (N) increases indefinitely, the average wage pet worker will sete, down 00.8%, The slope coefficient is postive suggesting, that the rate of chante we of Fy with respect to number of employees (1) will be negative throughout, ‘This would have been done to correct the problem of heteroscedasticity. From moving regression] to regression 2, there is an assumption made about the error variance that Var (u,) = a Nt where q is any non zero real number. is the error (or disturbance) term, , (td) No, iis not possible to compare intereceptor slope coefficient between | the two models because the explained (or dependent) variable in the two models | © are not same. 188suctory Heo (MY 2019) ses a tie In regression 1, Fis computed to measure ode ae graph paper (meastired ix GSM) weed of ibe such that it does not tear off easily while ould ore the true average thickness of the new doneinsideration. The true average thickness oer ter than or equal to 20 GSM for it prvjth normal distribution. What conclusion heats 161232055 fe 2402-02 that # 2.20 GSM. ascii 7 | a= 001 adver iy or | Sine, absolute value of f statistics iby are skilled workers and their work experience over the years. 6 | {) Define dummy variables to: capture whether workers are skilled : ies below 187 Interpret the new model, ‘Ans Let Dy be the damn Sitied wenket and nom sneak How would the model in (ii) change. I 7 st Y i BA. Hons) Enon Yeu Semi) i individual asa funet ion of wor ther i Tab eed Interpret your model OTF PeHenoe and whether irgtersand thelr werkexpeionee Moreeetsgn eee silletthe | i ‘hati ued to diffrentatebetRten illed worker. Then, eae 1 stworkers is eis 0 workers non-skiled led the dummy vanable (, ken asthe reference (r be Wear Tegression model can be stated ae Y= Bo tbiD.= itn where Y, : earnings of individual 1 le D,: whether the worker is skilled of non-skilled ie., dummy vanal | ‘X,: work experience in years - Hu mt {error (or disturbance) term and Bo, By and B are regression coefficient (or reference category) ,) takes the values sAchmark) eategory. 5 where Bis used as benchmark. when nd fs denotes the lvl of earning ofan individ unskilled and is new te, net worked atleast a veer Weegee the earnings level when the worker ih hile! ter the addon of Sncome of skilled worker a compared to unsiled worker when te short af work experience is same . ‘The X denoies the adaitios ncincome fora given siled or waskled worker when be gains one additonal year of work experienc Here, it neeeed that addition in income due toinereased work oxpeients ic tomae fe siicbet ‘inskilled worker (Gi) If her is en interaction term betwoen sil ofthe worker and their work experience thon the new regreation adel becomes Y=, Bo* By Dy +82 X,+ By Dy X, + uy {2 faked worker where D.= 1p s¢nontilled worker and D,X,: interaction term ‘Tho new term ie., the intoraction term between D, and X, measures the Gifference in impact ef each additional year of work experience on earnings of Q.5 (a) Tho results ofa logarithm regression of demand for food fon price and personal disposable income is given as : log Q, = 2:34--0.31 log P, + 0.45 log ¥,+ 0.65 158 oft bee + otCC-10: Introductory Econometrics (May 2019) 507 seo: B.A, (Hols) Economies Year on a = (0.05) (0.20) (0.1.4) n= 50 R¥= 0,90 d= 1.8 where @= food consumption per capita P= food price Y= real per capita disposable income @ Just by looking at the estimated regression, do you suspect serial correlation in it? Gi) Which test do you use to confirm your suspicion and why: Gil) Outline the steps of the above mentioned test and pro conclusion on the basis of your ealculations. Ans, log Q, = 2.94 ~ 0.31 log P, + 0.45 log ¥; + 0.05 log Q..1 865 (0.05) (0.20) (0.14) n= 50 R= 0.90 a= 18 (Yes, since the model has lagged value of dependent variable as one of ite repressors. The durbyn-watson d-test’s aesumption are also not satisfied because of lagged dependent variable fi 1) We can use Durbin’s h test since In @, depends on ite lagged values. ')_ Durbin's test. I ean be used when lagged dependent variable exists. as one of explanatory variable to check autocorrelation, But only for AR(1) and not for higher order correlation, Steps involved in the test are Step 1: Estimate the model by OLS and olitain the residuale, (2a Stop 2: Estimate A trom (252) ~ > relationship. Step 3 : Construct Durbin's h-statistics ae where n : number of observations Var (f4) : variance of coefficient of lagged dependent variable, rge samples, this statistics has a normal distribution and hence reject, hypothesis. G.e., no autocorrelation) My:P= 0 Ie value of h-statisties exceeds the eritieal A-value. In this ease, we have (2228) b= (2238).01 Ca) 159 508 he of oe ial 5000.14)? } * af $s A= 7 Ry ate h~ NOI” : Since, calculated A = 6 exceeds the critical value of autocorrelation. ; (b) Suppose you are given the following regress, ¥, = Bo+BiX, +B.X? +p, Do you think the model suffers from muilticollinearity what are the possible remedies ofthe problem ? yn, Ans. ¥, = Bo +BX, + BX? +u, ‘The problem of multicollinearity exists when one ofthe indepen can be expressed as a Variable (). Since, ho problem of multicollinearity. ‘Remodial measures to solve multicollineasity 1. Dropping variables from the regression. The mulkicollinearity may be in dropping a collinear variable unto fication error. 2. Acea jes within a ; cand is sample specie hene acquiring a nw a may reduce (or ren inearity, Further, if we increase the a=... size then the severity of multicollinearity may réduce. ine 8. Rethinking the model. Modifying the functional form mightreiia covery of 4. ‘Transformation of variables, For exemple transforming apna conapmption expe nto per capital consuiaption expendten oy (c) State and prove the minimum variance property of the slape coefficient in a two variable regression model 5 Ans, Let = B+ BX + ‘Then, Gauss-Markov theorem states that if the assumptions of cassia! linear regression model (CLRM) are satisfied, then the estimates of the mo are BLUE, The ince property of estimated slope ceecet nce of fy is loss than the variance of any othe ant potev snmoavtoryHeonometic (May 2019) 509 10: I A os sor of Py aay @ 1s unbiased linear estimator estima soon ae tat Be ou = BI Y weights miased estizaator of Dy 3@) = B wy, 508) = Bs eee x) = Bt an ae fe necessary at) iii) var (200%) @- Var (W, ¥, + W, Yat act WY) WE Var (%))+ W3Variy) +... + W2Var(Y,) (since, Cov (¥, ¥) = Ov Fa £ Wo? + Wo? +..+ Wo? EW? oF LAY, -h, +h)” = oy)? + EAE + 250%, -& yk) 4 oF SOW, hy)? +07 ZA? +20"(EMia EAP) 2 2 1 = EW hy) lee 161 S10 A = hy tei FLEW, - 8)—1] Bulfrom2and Swe ger Var 3 2 2 (8) = oF Son -4) S00 = FLAW -4y? an Ws ES Var (B) = o? S0W;~k)®+ Var @) 5 Since, c? > 0; 5 (W,~ 1)? only Whang ‘Thus, we have, 2 O always thus equation (i) is minimum if and Var (8) = Var @) aed, where we assumed Var (8) is minimum, ‘Thus, Varof fy is minimum Q. 6. (a) Consider the follpwing models : 5 slog Modell: InYy =a, +a,10 X} +p) : off : Model IT: In¥; =B, +B,In.X, +p : i sua Where ¥ w Yiand X/ =, X, the w’s being constants. Establish the relationship between the two sets of regression coefficient and their standard errors. (id) Is the R? different botween the two models? ' ‘Ans. Model 1: In" = @, +0 1aX7 +4! 182si (i) Where ¥7 = WY, and X7 = 4X, and W, are constants 1. Model 1 can be rew In cy, = a+ agin (iy = In W, +n Y, = In = a+ ayIn Wy + cg ln X tut, = (0, + ae In W~In W,)+ ay In X; + 0 > In ¥,= by tas Js tranformed version of model } where By =a, +02 In W~In W, ow. comparing (ii) with model 2 depicts that since explained, explanatory And error torms in the two models are same, the slope coefficient will not change however the intercept will change when ie move from model l to modelo pe coefficient in both model represents the rspective clasticities of iit) s given as 1.c., R? depends on slope coeffie, and on their respective means only. 7: (&) Suppose the CLRM aj BEX; = ey 'd the slone coeflicient in the regression of Yand X, , Suppose now we have a regression of X and ¥,X,= 74 ¥; du, te slope coofiicient of regression on Xon Yan inverse of slope of cenrese of Yon x. Ans. Ye mX+e, CLRM i applied ' (By OLS, method, we find slope coefficient in order to minimise sum of squared arror ie. 2 explanatory variables 163 doll sit Gxaran B.A. (Hons) EcoRomics Il Year (Sem) Setting this differential equal to 0 as, 2E[(¥, - Be X) C X= 0 = -E(HM) +B EX =0 2 Mey = Xi-n% w X(V2) = 20%;-2 YD? Differentiating w.r.t yz and then setting this equal to 0 gives eX (vi) Now, regression of Yon X is N= BeXte ye = oars ' Put this value of X; in yp as (©) Using data on compensation per employee in thousands of jars (COMP) and average productivity in thousands of dollars 104soavetory Econometrics (May 2019) 513. 750 firms for the year 1958, the following {ned (t ratios in parentheses) : 8 obrii'1992.35 + 0.233 PROD; p= (1275) (2.333) pie 0.5891 jal data included heterogeneous units, {uly to be present. The Park test was performed He Grauxiliary repression were obtained : surlts 0 2 2 35.817 - 2.8099 PROD; 1 = (0.934) (- 0.667) R® = 0.0595, ry regression to check if the model ‘cedasticity. Perform the test at 5% level of eault of aunt jeteross ossible remedies of heteroscedasticity ? 6 = 50 1992. 35 + 0.233 Prodi 42.1275) ( 2.333) R= 0.5891 = 36,817 ~ 2.8099 Prodi (0.934) (0.667) = 0.0595 a (no heteroseedasticity) He (there is heteroscedasticity) ovansmuesiectvem ae += SEG) ten =~ 0.667 = 0.08 sm nnn 2 = 0 ~ 2 a8 . ical values of ¢ with two-tailed test lies between me toon as € (2.009, 2.021) | 17.549 _ 35.173 For (1988 ~ 1969: ¥j = 10.078 ~ 10.337 - #234 SO ted ss weighted"C-10: Introductory Econometrics (May 2019) sis ea aque : B.A. (Hons.) Economics II Year (Sem) : ap 3 Find veriat meymyiote? - 4» ose soee(3) Sap Fis x os Soot: Boch empte: patentee : 0.259: 2] ae Stop 3 : Check oblique asymptote | eh, [-e2s0+200212) «0259 “errr aOTBK, 17.549 ( oe ee 1 . jominator is same thus i sur {0 250 + 2062431-| = Since, degree of numerator and denominator thus thereis nook, | eal : pines A) | tet eave le varteallegyeneeels Step 4: Find intercepts [Be RCAS erin aunt units tne verelanymptce andes =o, to 2A, | F = 70.280%, + 20.624 Thus, for period 1970-1977, the graph is : _ + x ib | Since degree of numerator and denominator is same thus, nae Step 4: Find intercepts : f . ce cm 20824 ‘Thus, there and percentage chang an inverse relationship between percent unemployment n earnings for period 1958-1969. But sh period 1955-1969 only. ). The regression equation is > =17.5409/1) (cu {= 0rs-r0aa7p,-17.00{ 4s s6.1730( 1) Se= (1.4204) (1.6859) (8.3373) (9.399) neis . : ' of significant @ = 5% = 0.05 : = swore-azae( 2) na) The ditferntil inereept ie : ay [toere-toarracsare-sneing, D1 BG) = a : woors 17.540 ip=0 | (x) 167 : 168CC-10: Introductory Econometrics (May 2019) 517 10.078) vend = Tazoa | Sincecaleulated 1 is statistically significast (as t.., > 3) thus null hypothesis of no & : |p ong Seated in favour of ast sgmbconce ot, noone se-S p= Is the De (= 10.078) is significant whereas the -value of differential = fq 1 = 10:387) @ = 10.837) is 6.13 {a0 = 19387) whieh is also statistically sascant. Thus, Hy By = O is rejected i... B; is signifieant ne differential slope coefficient is J (= 38.173) and tvalue od differential api eoefcient is 4.06 which is statistically significant, thus Hy, 20 ie ie din favour of the fact that differential slope is statistically signiteant. ‘This suggests that the rate of change of hourly earnings will be different us ferent period with constant rate of unemployment, Thus, ifwe docenot de dummy variable, D,, then the model will be underft. ) The period duramy intercept coefficient of ~ 10.397 interprets that gercont change in hourly earnings is lower during 1958-1969 by about 10.5796 Weompared to period 1970-1977, The covfficient of benchmark category th {0078 states the percentage change in hourly earnings curing 1970-109 Iossewith no change in unemployinent rate, The slope cociSeient of (17-540 nee BL + 86.178) = By Qstates that with one peeent change in) the hourly ezenings changes by ~ 17.549 dining 1970-1977 and by — 17.849 + 981 delenit daring toas-i06 (by Two madels for Engel oxpondleure hinctlon are estimated § Models = destoao + 0.077 t= 1.64) R*= 0.350 F = 468.645 Model: 8, = iE Gass can.sio R 0 ‘The Festatstics is REe=v Fe + a= RDI) ki, i for each auxillary regression. * Uealelated value of Fis eresenshandhe crite! value fF yey of significance, then we reject null hypothesis, H and mighs eat collinear with other explanatory variables, o 172