Exogeneity tests and weak-identification
|
|
- Darren Byrd
- 5 years ago
- Views:
Transcription
1 Exogeneity tests and weak-identification Firmin Doko Université de Montréal Jean-Marie Dufour McGill University First version: September 2007 Revised: October 2007 his version: February 2007 Compiled: February 5, 2008, :58am his work was supported by the William Dow Chair in Political Economy (McGill University), the Canada Research Chair Program (Chair in Econometrics, Université de Montréal), the Bank of Canada (Research Fellowship), a Guggenheim Fellowship, a Konrad-Adenauer Fellowship (Alexander-von-Humboldt Foundation, Germany), the Institut de finance mathématique de Montréal (IFM2), the Canadian Network of Centres of Excellence [program on Mathematics of Information echnology and Complex Systems (MIACS)], the Natural Sciences and Engineering Research Council of Canada, the Social Sciences and Humanities Research Council of Canada, the Fonds de recherche sur la société et la culture (Québec), and the Fonds de recherche sur la nature et les technologies (Québec), and a Killam Fellowship (Canada Council for the Arts). Université de Montréal, Département de Sciences Économiques, C.P. 628, succ. Centre-ville, Montréal, QC H3C 3J7. EL: () ext. 5283; Fax: () ; f.doko.tchatoka@umontreal.ca. William Dow Professor of Economics, McGill University, Centre interuniversitaire de recherche en analyse des organisations (CIRANO), and Centre interuniversitaire de recherche en économie quantitative (CIREQ). Mailing address: Department of Economics, McGill University, Leacock Building, Room 59, 855 Sherbrooke Street West, Montréal, Québec H3A 27, Canada. EL: () ; FAX: () ; jean-marie.dufour@mcgill.ca. Web page:
2 ABSRAC In a linear regression model, one often uses Hausman and Wu procedures to access whether some set of explanatory variables are uncorrelated with the disturbances. Instrumental variable methods are typically employed to make reliable these two procedures. However, the Hausman and Wu tests are based on the questionable assumption that the instrumental variables used are strong and valid and this raises the question: what happens to these tests when some of the instruments are weak? In this paper, we focus on structural models and analyze the effects of weak instruments (or weak identification) on the Hausman and Wu exogeneity tests. We show that if the model parameters are not identified or are close to not being identifiable, both procedures are valid but not consistent against the presence of endogenous explanatory variables. In particular, if the model parameters are not identified, the power of the tests converges to their nominal level α. In other words, if the parameters are unidentified or are weakly identified, both Ordinary Least Squares (OLS) and wo Stage Least Squares (2SLS) estimators are biased but the Hausman and Wu procedures are unable to detect the bias. Key words: structural model; exogeneity tests; weak-identification; robustness to weak instruments. i
3 Contents. Introduction 2. Framework 3. Statistics 3 4. Asymptotic theory with weak instruments Simplified model Extension to structural equations with exogenous variables Monte Carlo experiment 7 6. Conclusion 2 A. Appendix: Proofs 2 List of Assumptions, Propositions and heorems 4. heorem : Asymptotic distribution of the statistics under H heorem : Asymptotic distribution of the statistics under the alternative hypothesis heorem : Asymptotic distribution of the statistics under H 0 : extended model heorem : Asymptotic distribution of the statistics under the alternative hypothesis: extended model Proof of heorem Proof of heorem Proof of heorem Proof of heorem ii
4 . Introduction Since Durbin (954) and Wu (973), it has been showed growing interest on the inference about the vector of covariance between the stochastic explanatory variables and the disturbance term of a structural equation. Revankar and Hartley (973), Farebrother (976), Hausman (978), Revankar (978), Kariya and Hodoshima (980), Richard (980), Holly and Sagan (982) proposed tests for the hypothesis of independence between a full vector of stochastic explanatory variables and a disturbance term. Hwang (980) and Smith (984) studied likelihood ratio tests, Hausman and aylor (98), Spencer and Berk (98) and Wu (983) extended the tests previously studied by Wu (973) and Hausman (978), Engle (982) derived the LM tests while Dufour (987) proposed a generalized Wald procedures for the inference on the vector of covariance between the stochastic explanatory variable and the disturbance term in structural models. A drawback with all these procedures is that they assume strong identification of the model parameters, i.e. they assume that the instruments are strong. However, the last decade shows growing interest for weak instruments problems in the econometric literature, i.e. situations where instruments are poorly correlated with endogenous explanatory variables; see the reviews of Dufour (2003) and Stock, Wright and Yogo (2002). More generally, these can be viewed as situations where model parameters are not identified or close to not being identifiable, as meant in the econometric literature [see Dufour and Hsiao (2008)]. his raises the question: what happens to these procedures when some of the instruments are weak? In other words, how robust are these inference procedures to weak-identification? In the econometric literature, little attention is focused on Hausman (978) and Wu (973, 974) tests when the instruments used are weak. In this paper, we focus on structural models and analyze the effects of weak-identification (weak instruments) on these exogeneity tests. After formulating a general asymptotic framework which allows one to study these issues in a convenient way, we show that both test procedures studied are robust to weak-identification (in term of level control), but have no power. hese results means that both test procedures are valid in the presence of weakidentification, but are not able to detect whether or not the explanatory variables are endogenous in the model. When the model is not or weakly identified, both Ordinary Least Squared (OLS) and wo Least Squared (2SLS) estimators of the structural parameters are inconsistent but the Wu and Hausman type procedures are unable to detect this inconsistency. he paper is organized as follows. Section 2 formulates the model considered. Section 3 describes briefly the statistics. Section 4 studies the behavior of the statistics when some instruments are weak. Section 5 studies the finite-samples properties of the statistics with weak-identification. We conclude in section 6. Proofs are presented in the Appendix. 2. Framework We consider the following standard simultaneous equation framework: y = Y β + γ + u, (2.) Y = Π + 2 Π 2 +, (2.2)
5 where y is a vector of observations on the dependent variable, Y is a G matrix of observations on explanatory (possibly) endogenous variables (G ), is a k matrix of observations on the included exogenous variables, 2 is a k 2 (k 2 > G) matrix of observations on the excluded variables (instruments), = [, 2 ] = [ (),..., ( ) ] is a k (k = k + k 2 ) full-column-rank matrix of instruments, u = [u,..., u ] and = [,..., ] are respectively vector and G disturbance matrices with mean zero, β and γ are G and k vectors of unknown coefficients, Π and Π 2, are k G and k 2 G matrices of unknown coefficients. he usual necessary and sufficient condition for identification of this model is rank(π 2 ) = G. We also assume that u t = t a + ε t, t =,...,, (2.3) where ε t has mean zero and variances σ 2 ε and uncorrelated with t, a is a G vector of unknown coefficients. (2.3) can be rewritten in matrix form as: u = a + ε. (2.4) We make the following generic assumptions on the asymptotic behaviour of model variables [where A > 0 for a matrix A means that A is positive definite (p.d.), and refers to limits as ]: [ [ ] [ ] p Σ 0 ] ε ε 0 σ 2 > 0, (2.5) ε [ ε ] p 0, ε L S ε N [0, Σ ε ] (2.6) L S, vec(s ) N [0, Σ Σ ], (2.7) [ p Σ Σ Σ = 2 Σ 2 Σ 2 ] ] [ u ε [ L Su S ε ] > 0, (2.8) N [0, Σ S ], (2.9) S u N [ 0, σ 2 uσ ], Sε N [ 0, σ 2 εσ ], (2.0) where Σ is G G fixed matrix and σ 2 u is the variance of u. From the above assumptions, it is easy to see that: u p 0, [ u ] [ u ] p Σ = [ σ 2 u δ δ Σ ] > 0, (2.) where δ = Σ a, σ 2 u = a Σ a + σ 2 ε, S u = S a + S ε = S (Σ δ) + S ε (2.2) Σ Y = Σ Π, Σ Y = Π Σ Π + Σ. (2.3) 2
6 It is worthwhile to note that since Σ > 0, δ = 0 if and only if a = 0. Hence, testing the exogeneity of Y is equivalent to test whether a = 0 in this model. 3. Statistics We consider in this paper the problem of testing H 0 : δ = 0 (3.) and we analyze the behavior of the?and? statistics. he Wu and Hausman test-statistics for H 0 in equations (2.)-(2.2) are given by H = Q Q/, = Q /G Q /(k 2 G), 2 = Q /G Q 2 /( k 2G), (3.2) where 3 = Q /G Q 3 /( k G), 4 = Q /G Q/( k G), (3.3) Q = (ˆβ β) [(Y (M M)Y ) (Y M Y ) ] (ˆβ β), (3.4) Q = (y Y ˆβ) M (y Y ˆβ), Q = (y Y β) (M M)(y Y β), (3.5) Q 2 = Q Q, Q 3 = (y Y β) M (y Y β), (3.6) β = [Y (M M)Y ] Y (M M)y, ˆβ = [Y M Y ] Y M y, (3.7) M = I ( ), M = I ( ), and = [, 2 ]. (3.8) Note that ˆβ is the ordinary least squares estimator of β whereas β is the instrumental variables method estimator of β. Q/ is the ordinary least squares estimator of σ 2 u. From (2.)-(??), it easy to see that [ ] plim ˆβ = β + Π (Σ Σ Σ Σ )Π + Σ δ, (3.9) where Σ = [Σ, Σ 2 ]. If Π 0, i.e the instruments are strong, we have plim β = β. (3.0) However, if the model parameters are not identified or are close to being identifiable, like the Ordinary Least Squares, the wo Stage Least Squares estimator β of β may be inconsistent. In particular, if Π = 0, we have plim β = plim ˆβ = β + Σ δ, (3.) and when δ 0, both ˆβ and β are inconsistent estimators of β. If the model parameters are identified, with some regular conditions, the asymptotic distribution of H, 3, and 4 is a χ 2 (G) under H 0. Further, if u N[0, σ 2 I ], then F (G, k 2 G) and 2 F (G, k 2G) [see, 3
7 Hausman (978), Wu (973)]. However, if the parameters are not identified or are close to not being identifiable, the distributions of these statistics may be affected. Following Staiger and Stock (997), we refer to locally weak-instrument asymptotic setup by considering a limiting sequence of Π where Π is local-to-zero. In the following section, we study the asymptotic behavior of the statistics when some instruments may be weak. 4. Asymptotic theory with weak instruments We study in this section the large-sample properties of the statistics described above. In the first subsection, we consider the case where drops from the model (no excluded exogenous instruments) and in the second subsection, we extend the model to structural equations with exogenous variables. 4.. Simplified model We consider the following form of (2.)-(2.2) where is dropped from the model: y = Y β + u, (4.) Y = Π +, (4.2) where y and u are vectors, Y and are G matrices, is a k matrix, and β, Π are respectively a G vector and a k G matrix of unknown coefficients. wo setups are considered: () Π = Π 0 where Π 0 is a k G constant matrix, and (2) Π = Π 0 / ( ), where Π 0 is a k G constant matrix. heorem 4. below summarizes the asymptotic behavior of Hausman and Wu statistics for both setups described above. For a random variable W whose distribution depends on the sample size, the notation W L + means that P [W > x] as, for any x. heorem 4. ASYMPOIC DISRIBUION OF HE SAISICS UNDER H 0. Suppose that the assumptions (2.)-(2.0), and (4.)-(4.2) hold, and let δ = 0, then H L χ 2 (G), (4.3) L 2, 3, 4 G χ2 L (k G) (G) and F (G, k G). (4.4) G In the above theorem, no restriction is imposed on the rank of Π. In particular, the result holds even if Π is not a full-column rank matrix. When δ = 0, the limiting distributions of the statistics do not depend on nuisance parameters even if the model parameters are unidentified, i.e the statistics do not show size distortion even if the parameters are not identified or are close to not being identifiable. heorem 4.2 below summarizes the asymptotic behavior of the statistics when δ 0, i.e under the alternative hypothesis. 4
8 heorem 4.2 POHESIS. Suppose that the assumptions (2.)-(2.0), and (4.)-(4.2) hold, and let δ 0. (A) If Π = Π 0 where Π 0 is a k G constant matrix and if rank(π 0 ) = G, then ASYMPOIC DISRIBUION OF HE SAISICS UNDER HE ALERNAIE HY- H, i L +, i =, 2, 3, 4. (4.5) (B) If Π = Π 0 / where Π 0 is a k G constant matrix (Π 0 = 0 is allowed), then H L S δ = σ 2 ε (Π 0 Σ H S L χ 2 (G, µ ) where µ = σ 2 ε δ Σ S ε) Σ Λ (Λ Σ Λ ) Λ Σ (Π 0 Σ δ Σ S ε), δ Σ Π 0Σ Λ (Λ Σ Λ ) Λ Σ Π 0 Σ δ, (4.6) where 2, 3, 4 L G S δ and 2, 3, 4 S G χ2 (G, µ ), (4.7) L (k 2 G) G S δ L (k G) and S Λ δ G F (G, k G; ν, λ ), (4.8) λ = σ 2 ε Λ δ = (S Σ δ + S ε) [Σ Λ (Λ Σ Λ ) Λ ](S Σ δ + S ε), (4.9) δ Σ S [Σ σ 2 ε, Σ, Σ, S, and S ε are defined in (2.)-(2.2). Λ (Λ Σ Λ ) Λ ]S Σ δ, Λ = Π 0 + Σ S. (4.0) he results of heorem 4.2 show clearly that when the model parameters are identified, the Wu and Hausman procedures are consistent against the presence of endogenous explanatory variable in the regression model. However, when the model parameters are not identified or are close to not being identifiable (i.e when the instruments are weak), both procedures are not consistent against the presence of endogenous explanatory variables. hese results mean that when the instruments are weak, both Ordinary Least Squares and wo Stage Least Squares estimators of β are inconsistent but the above tests are unable to detect this inconsistency. It is worthwhile to note that if Π 0 = 0 in (B), we have ν = λ = 0, and H L χ 2 L (G), and i G χ2 (G), i =, 2, 3, 4. In this case, the power function of the tests H, i, i =, 2, 3, 4 converge to the nominal level α. Overall, our results underscore the importance of checking for the presence of weak instruments when applying exogeneity tests Extension to structural equations with exogenous variables In this subsection, we consider the more general model (2.)-(2.2). As in the previous subsection, two setups are examined: () Π = Π 0 where Π 0 is a k G constant matrix, and (2) Π = Π 0 / ( ), where Π 0 is a k G constant matrix, k = k + k 2. heorem 4.3 and heorem 4.4 below derive the asymptotic distributions of the statistics for both setups. 5
9 heorem 4.3 ASYMPOIC DISRIBUION OF HE SAISICS UNDER H 0 : EXENDED MODEL. Suppose that the assumptions (2.)-(2.0), and let δ = 0, then H L χ 2 (G), (4.) L 2, 3, 4 G χ2 L (k 2 G) (G) and F (G, k 2 G). (4.2) G heorem 4.4 POHESIS: EXENDED MODEL. Suppose that the assumptions (2.)-(2.0) hold, and let δ 0. (A) If Π = Π 0 where Π 0 is a k G constant matrix and if rank(π 0 ) = G, then ASYMPOIC DISRIBUION OF HE SAISICS UNDER HE ALERNAIE HY- H, i L +, i =, 2, 3, 4. (4.3) (B) If Π = Π 0 / where Π 0 is a k G constant matrix (Π 0 = 0 is allowed), then where H L S δ = σ 2 [Σ δ (Λ Σ Λ Λ Σ Λ ) (Λ S u Λ S u )] ε [Λ Σ Λ Λ Σ Λ ][Σ δ (Λ Σ Λ Λ Σ Λ ) (Λ S u Λ S u )], (4.4) 2, 3, 4 L G S δ, (4.5) L σ 2 ε (k 2 G) G S δ, (4.6) Λ δ Λ δ = S uσ S u S uσ S u + [Σ δ (Λ Σ Λ Λ Σ Λ ) (Λ S u Λ S u )] [Λ Σ Λ Λ Σ Λ ][Σ δ (Λ Σ Λ Λ Σ Λ ) (Λ S u Λ S u )], (4.7) σ 2 ε = σ 2 u δ Σ δ, Λ = Π 0 + Σ S, Λ = Σ Σ Π 0 + Σ S, (4.8) S = [S, S 2 ], S u = [S u, S 2u ], and S, S u,, Σ, Σ, Σ, Σ, σ 2 ε, σ 2 u are defined in (2.)-(2.2). he results of heorem 4.3 and heorem 4.4 are similar to those in heorem 4. and heorem 4.2 of the previous subsection. Clearly, when the model parameters are not identified or are close to not being identifiable (i.e when the instruments are weak), the Wu and Hausman procedures control the size but are not consistent against the presence of endogenous explanatory variables. 6
10 5. Monte Carlo experiment We study in this section the finite-samples properties of the statistics described above when some of the instruments used are weak. We consider the following data generating process: y = Y β + u, Y = Π +, (5.9) [ ] (u t, t ) i.i.d δ N(0, Σ), where Σ = (5.20) δ Σ and Σ = [σ ij ] i,j=,,g with σ ii = for all i =,..., G. is the k fixe matrix of instruments excluded in the structural equation such that is i.i.d N(0, I ). he k G matrix Π is such that Π = ηπ 0, where η takes the value 0 (design of non identification),.00 (design of weak identification) or (design of strong identification), and Π 0 is obtained from identity matrix by keeping the first k lines and the first G columns. he number of endogenous variable G varies from to 3 whereas the number of instruments k varies from 5 to 20. he parameters values are set at β = 5 for G =, (β = 5, β 2 = 2) for G = 2 and (β = 5, β 2 = 2, β 3 = 3 2 ) for G = 3. he sample size is = 00 and the number of replications is N = 0, 000. he covariance matrix Σ is chosen as: δ {.5,.25, 0,.25,.25,.5} and Σ = for G = ; Σ = δ δ 2 δ δ δ 2 δ 2 δ δ 2 where δ j {.5,.25, 0,.25,.25,.5} (j=,2) for G = 2; and finally Σ = δ δ 2 δ 3 δ δ δ 2 δ δ 3 δ 2 δ δ 2 δ 2 δ 3 δ 3 δ δ 3 δ 2 δ 3, (5.2), (5.22) where δ j {.5,.25, 0,.25,.25,.5} (j=,2,3) for G = 3. If δ i = 0 for all j, then H 0 is satisfied and the variables Y are exogenous in the model. Otherwise, they are endogenous. he nominal level of the tests is 5%. For each value of δ j (j=,2,3), we compute the empirical rejection probability of each statistic. Remark that for a given δ 0, the computed rejection probability is the power of the test at δ. For δ = 0, the empirical rejected frequency is the level of the tests. he results are presented in ABLE -ABLE 3 for the different value of G. In the first column of the tables, we report the values of η (which is the quality of the instruments) whereas in the second columns, we report the values of k (the number of instruments). Finally in the others columns, we report for each value of δ, the rejected frequencies at nominal level of 5 % for the statistics i (i=,2,3,4) and H. he main observation from these results is that excepted the statistic 3, the size of the others statistics are not affected by weak instruments. Which means that H,, 2 and 4 tests are valid even if the model parameters are not identified or close to not being identifiable. However, the power of the tests is dramatically affected by weak instruments (approximatively equal to 5 %). his 7
11 result means that when the model parameters are not identified or close to not being identifiable, the above exogeneity tests are inconsistent against the presence of endogenous explanatory variables in the model. 8
12 ABLE : Percent Rejected at nominal level of 5 % for G= δ =.50 δ =.25 δ = 0 k H H H η = η = η = δ =.25 δ =.25 δ =.50 k H H H η = η = η =
13 ABLE 2: Percent Rejected at nominal level of 5 % for G=2 δ =.50, δ2 =.50 δ =.25, δ2 =.25 δ = 0, δ2 = 0 k H H H η = η = η = δ =.25, δ2 =.25 δ =.25, δ2 =.25 δ =.50, δ2 =.50 k H H H η = η = η =
14 ABLE 3: Percent Rejected at nominal level of 5 % for G=3 δ =.50, δ2 =.50, δ3 =.50 δ =.25, δ2 =.25, δ3 =.25 δ = 0, δ2 = 0, δ3 = 0 k H H H η = η = η = δ =.25, δ2 =.25, δ3 =.25 δ =.25, δ2 =.25, δ3 =.25 δ =.50, δ2 =.50, δ3 =.50 k H H H η = η = η =
15 6. Conclusion In this paper, we have showed that in linear structural models, when the parameters are not identified or are close to not being identifiable, the usual exogeneity tests like Hausman (978) and Wu (973) type tests are valid but not consistent against the presence of endogenous explanatory variables. In particular, if the parameters are unidentified (i.e if the instruments are irrelevant), the power of these tests converges to their nominal level α. In other words, if the instruments are weak, both Ordinary Least Squares and wo Stage Least Squares estimators are biased, but the Hausman (978) and Wu (973) procedures are unable to detect the bias. A. Appendix: Proofs PROOF OF HEOREM 4. Note first that the statistics i, i =, 2, 3, 4; can be written as = (k 2 G) Q/ H, 2 = ( k 2G) G Q G H H, (A.) and 3 = Q/ G Q 3 /( k G) H, 4 = ( k 2G) H, G (A.2) where H is defined by (3.2). Suppose that δ = 0, From (4.)-(4.2), we have M = I and M M = P = ( ). (A) If Π = Π 0 where Π 0 is a k G constant matrix, then by the assumptions (2.)- (2.0), we have and Y P Y p Π 0Σ Π 0, Y Y Q = u u u Y ( Y Y p Π 0Σ Π 0 + Σ, ) Y u p σ 2 u, Y u = Π u 0 + (Σ δ) + ε = Π u 0 + ε, Y P u = where (A.5) holds because δ = 0. So, we have ( Y ) ( (A.3) (A.4) (A.5) ) ( ) u, (A.6) Y u L Π 0 S u + S ε, Y P u L Π 0S u. (A.7) 2
16 If rank(π 0 ) = G, we have Q L [(Π 0 Σ Π 0 + Σ ) (Π 0S u + S ε ) (Π 0Σ Π 0 ) Π 0S u ] [(Π 0Σ Π 0 ) (Π 0Σ Π 0 + Σ ) ] [(Π 0Σ Π 0 + Σ ) (Π 0S u + S ε ) (Π 0Σ Π 0 ) Π 0S u ], (A.8) where Q is defined by (3.4). hus, H L σ 2 [(Π 0Σ Π 0 + Σ ) (Π 0S u + S ε ) (Π 0Σ Π 0 ) Π 0S u ] u [(Π 0Σ Π 0 ) (Π 0Σ Π 0 + Σ ) ] [(Π 0Σ Π 0 + Σ ) (Π 0S u + S ε ) (Π 0Σ Π 0 ) Π 0S u ]. (A.9) Since /delta = 0 if and only if a = 0, from (2.2), we have σ 2 u = σ 2 ε and it is easy to see from assumptions (2.)- (2.0) that [ ] { [ Π 0 S u + S ε Π 0 S N 0, σ 2 Π 0 Σ Π 0 + Σ Π 0 Σ ]} Π 0 u u Π 0 Σ Π 0 Π 0 Σ Π 0 and (Π 0Σ Π 0 + Σ ) (Π 0S u + S ε ) (Π 0Σ Π 0 )Π 0S u N { 0, σ 2 u[(π 0Σ Π 0 ) (Σ + Π 0Σ Π 0 ) ] }. (A.0) hus Q L σ 2 uχ 2 (G) and H L χ 2 (G), 2, 4 L G χ2 (G). (A.) Furthermore, if δ = 0, we have plim ( Q 3 G ) = σ2 u, so Moreover, we can write Q as 3 L G χ2 (G). Q = u P M (P Y )P u = ũ MŶ ũ, (A.2) (A.3) where ũ = P u and Ŷ = P Y. So, we have Q = ũ MŶ ũ = u P u u ( ) Y (Y P Y ) Y ( ) u, (A.4) 3
17 and if rank(π 0 ) = G, from assumptions (2.)- (2.0), Q L S u Σ S u S uπ 0 (Π 0Σ Π 0 ) Π 0S u = S uσ /2 [I k P (P P ) P ]Σ /2 S u, (A.5) where P = Σ /2 Π 0 and the matrix I k P (P P ) P is idempotent with rank k G. Further, we have S u N[0, I k ], hence σ u Σ /2 Q L σ 2 u χ 2 (k G) and L (k G) G χ 2 (G) χ 2 (k G). (A.6) Further, since Q can be written as u A u where A = [Y (Y Y ) P Y (Y P Y ) ][(Y P Y ) (Y Y ) ] [(Y Y ) Y (Y P Y ) Y P ] (A.7) and from (A.3), we have Q = u P M (P Y )P u, it is clear that A (P M (P Y )P ) = 0. So, Q and Q are asymptotically uncorrelated, and because S u and S ε are normal variables, Q and Q are asymptotically independent, i.e the numerator and the denominator of are independent. hus L (k G) F (G, k G). (A.8) G (B) If Π = Π 0 / where Π 0 is a k G constant matrix (Π 0 = 0 is allowed), then if δ = 0, we have Y P Y L Ψ = (Σ Π 0 + S ) Σ (Σ Π 0 + S ), Y Y p Σ, (A.9) Y u p δ = 0, Y P u L (Σ Π 0 + S ) Σ S u, (A.20) Q = u u u Y ( Y Y ) Y u p σ 2 u δ Σ δ = σ2 u. (A.2) Because when δ = 0 S u and S are uncorrelated, we deduce from (2.)- (2.0) that S u is independent of S. Hence, H L σ 2 S uσ u (Σ Π 0 + S )Ψ Furthermore we have (Σ Π 0 + S ) Σ S u S N(0, σ 2 uψ ), so (Σ Π 0 + S ) Σ S u. (A.22) Ψ /2 (Σ Π 0 + S ) Σ σ S u S N(0, I G ), (A.23) u i.e, H S L χ 2 (G). Since the conditional distribution of H is independent of S, we have H L χ 2 (G) and 2, 3, 4 L G χ2 (G). (A.24) 4
18 By following the same steps as the proof in (A) and conditioning on S, we get L (k G) F (G, k G). (A.25) G PROOF OF HEOREM 4.2 Suppose that δ 0. As in the proof of heorem 4., we have from (4.)-(4.2) M = I and M M = P = ( ). (A) If Π = Π 0 where Π 0 is a k G constant matrix, then by the assumptions (2.)- (2.0), we have Y P Y p Π 0Σ Π 0, Y Y p Π 0Σ Π 0 + Σ, Y u p δ 0, (A.26) Y P u = ( Y ) ( he denominator of the Hausman statistic is such that ( Q = u u u Y Y Y ) Y u ) ( ) u and its numerator can be written as [ (Y Q ) Y Y ( u Y ) ] = P Y [ (Y Y P u P Y [ (Y ) Y Y ( u Y ) ] P Y Y P u. If rank(π 0 ) = G, we have [ (Y Y ) Y u ( Y P Y p 0. (A.27) p σ 2 u δ Σ δ, (A.28) ) ] [ (Y Y P u P Y [ (Y ) Y Y u ) ( Y ) ] Y ) ( Y Y ( Y P Y (A.29) ) ] ) ] Y P u p δ (Π 0Σ Π 0 + Σ ) [(Π 0Σ Π 0 ) (Π 0Σ Π 0 + Σ ) ] (Π 0Σ Π 0 + Σ ) δ > 0. (A.30) hus, Q L +, i.e H, i L +, i =, 2, 3, 4. (A.3) 5
19 (B) If Π = Π 0 / where Π 0 is a k G constant matrix (Π 0 = 0 is allowed), then Y Y In addition, p Σ, Y P Y L (Σ Π 0 + S ) Σ (Σ Π 0 + S ), Y u p δ 0 Y P u L (Σ Π 0 + S ) Σ S u. (A.32) plim ( Q ) = plim ( Q 3 ) = σ2 ε = σ 2 u δ Σ δ = σ2 u a Σ a > 0, (A.33) and we have H L S δ = σ 2 ε where Λ = Π 0 + Σ S δ = σ 2 ε [Λ Σ Λ Σ δ Λ S u ] [Λ Σ Λ ] [Λ Σ Λ Σ δ Λ S u )], (A.34) (Π 0 Σ S. Since S u = S a + S ε = S (Σ δ) + S ε, we have δ Σ S ε) Σ Λ (Λ Σ Λ ) Λ Σ (Π 0 Σ δ Σ S ε). (A.35) Moreover, from (2.)- (2.0), S ε N(0, σ 2 εσ ) and independent of S, so, we is easily to see that S δ S L χ 2 (G, µ ), i.e H S, converge to a non central non degenerate χ 2 (G, µ ), and 2 S, 3 S, and 4 S also converge to a G χ2 (G, µ ) distribution, where µ = σ 2 ε δ Σ By proceeding as the proof in heorem 4., we have Π 0Σ Λ (Λ Σ Λ ) Λ Σ Π 0 Σ δ Q = ũ M P Y ũ L Λ δ = (S Σ δ +S ε) Σ /2 M Σ /2 (S Σ δ +S ε) σ 2 εχ 2 (k G, λ ), (A.36) where M = I k P (P P ) P, P = Σ /2 Λ and λ = σ 2 δ Σ S Σ /2 M Σ /2 S Σ δ = ε σ 2 ε δ Σ S [Σ Λ (Λ Σ Λ ) Λ ]S Σ δ. So, we get S L (k G) G F (G, k G; µ, λ ) (A.37) PROOF OF HEOREM 4.3 Suppose that δ = 0. (A) If Π = Π 0, Π 0 is a k G constant matrix and if rank(π 0 ) = G, then the proof is the same as those in heorem 4.. 6
20 (B) If Π = Π 0 /, Π 0 is a k G constant matrix (Π 0 = 0 is allowed), then we have Y (M M)Y L Ξ = (Σ Π 0 +S ) Σ (Σ Π 0 +S ) (Σ Π 0 +S ) Σ (Σ Π 0 +S ), (A.38) where we partition S as S = [S, S 2 ], S is a k G random matrix and S 2 is a k 2 G random matrix. Y P Y p 0, Y M Y = Y Y Y P Y p Σ, Y M u p δ = 0, (A.39) Y (M M)u L (Σ Π 0 + S ) Σ ( Q = u u u M Y Y M Y S u (Σ Π 0 + S ) Σ S u, ) Y M u (A.40) p σ 2 u δ Σ δ = σ2 u, (A.4) where S u is defined as S u = [S u, S 2u ], S u is the first k components of S u and S 2u its last k 2 components. Hence, H L σ 2 (Λ S u Λ S u ) Ξ (Λ S u Λ S u ), (A.42) u where Λ = Π 0 + Σ S and Λ = Σ Σ Π 0 + Σ S. So, it easily see that Ξ = Λ Σ Λ Λ Σ Λ. Since δ = 0, from assumptions (2.)- (2.0), S u is independent of S. We can use this fact to show that Λ S u Λ S u S N(0, σ 2 uξ ), i.e σ u Ξ /2 (Λ S u Λ S u ) S N(0, I G ). Since S u and S u are normally distributed and S u is not dependent on S, we have [ ] (Λ S u Λ S u ) S = [Λ, Λ Su ] S N(0, σ 2 uφ ), (A.43) where Φ = Λ Σ Λ + Λ Σ Λ Λ Σ Λ Λ Σ Λ. We now show that Φ = Ξ. hat is, Λ Σ Λ = Λ Σ Λ. It is easy to see from the expressions of Λ and Λ that Φ = Ξ if and only if Σ Σ S = S. (A.44) Moreover, we have Let partition Σ S u [ Σ = [ Σ Σ Σ Σ 2 ], Σ = ] 2. (A.45) according to the partition of Σ, i.e Σ = [ Σ Σ 2 Σ 2 Σ 22 Σ 2 Σ 2 ], Σ Σ = I k. (A.46) 7
21 From (A.46), we have Σ Σ + Σ 2 Σ 2 = I k, Σ Σ 2 + Σ 2 Σ 22 = Σ 2 Σ 2 + Σ 2 Σ = 0, Σ 2 Σ 22 + Σ 2 Σ 2 = I k2, where I k and I k2 are identities matrices with order respectively k and k 2. Now, we have [ Σ Σ Σ S = [ Σ Σ Σ 2 ] 2 ] [ ] S Σ 2 Σ 22 S 2 = (Σ Σ + Σ 2 Σ 2 )S + (Σ Σ 2 + Σ 2 Σ 22 )S 2. (A.47) (A.48) So, it is clear from (A.47) that hus [Λ, Λ ] [ Su S u distribution of H is not dependent on S, we have Σ Σ S = S, i.e Ξ = Φ. (A.49) ] S N(0, σ 2 L uξ ). Hence, H S χ 2 (G). Since the conditional H L χ 2 (G), 2, 3, 4 L G χ2 (G). By following the same steps as in heorem 4., we get (A.50) and that Q and Q are asymptotically independent, thus Q L σ 2 u χ 2 (k 2 G), (A.5) L (k 2 G) F (G, k 2 G), G (A.52) PROOF OF HEOREM 4.4 Similar to the one in heorem 4.2 References Dufour, J.-M., 987. Linear Wald methods for inference on covariances and weak exogeneity tests in structural equations. In: I. B. MacNeill, G. J. Umphrey, eds, Advances in the Statistical Sciences: Festschrift in Honour of Professor.M. Joshi s 70th Birthday. olume III, ime Series and Econometric Modelling. D. Reidel, Dordrecht, he Netherlands, pp Dufour, J.-M., Identification, weak instruments and statistical inference in econometrics. Canadian Journal of Economics 36(4),
22 Dufour, J.-M., Hsiao, C., Identification. In:. L. E. Blume, S. N. Durlauf, eds, he New Palgrave Dictionary of Economics 2nd edn, Palgrave Macmillan, Basingstoke, Hampshire, England. forthcoming. Durbin, J., 954. Errors in variables. Review of the International Statistical Institute 22, Engle, R. F., 982. A general approach to lagrange multiplier diagnostics. Journal of Econometrics 20, Farebrother, R. W., 976. A remark on the wu test. Econometrica 44, Hausman, J., 978. Specification tests in econometrics. Econometrica 46, Hausman, J., aylor, W. E., 98. A generalized specification test. Economics Letters 8, Holly, A., Sagan, D., 982. A remark on hausman s test. Econometrica 50, Hwang, H.-S., 980. est of independence between a subset of stochastic regressors and disturbances. International Economic Review 2, Kariya,., Hodoshima, H., 980. Finite-sample properties of the tests for independence in structural systems and lrt. he Quarterly Journal of Economics 3, Revankar, N. S., 978. Asymptotic relative efficiency analysis of certain tests in structural sysytems. International Economic Review 9, Revankar, N. S., Hartley, M. J., 973. An independence test and conditional unbiased predictions in the context of simultaneous equation systems. International Economic Review 4, Richard, J. F., 980. Models with several regimes and changes in exogeneity. Review of Economic Studies 7, 20. Smith, R. J., 984. A note on likelihood ratio tests for the independence between a subset of stochastic regressors and disturbances. International Economic Review 25, Spencer, D. E., Berk, K. N., 98. A limited-information specification test. Econometrica 49, Staiger, D., Stock, J. H., 997. Instrumental variables regression with weak instruments. Econometrica 65(3), Stock, J. H., Wright, J. H., Yogo, M., A survey of weak instruments and weak identification in generalized method of moments. Journal of Business and Economic Statistics 20(4), Wu, D.-M., 973. Alternative tests of independence between stochastic regressors and disturbances. Econometrica 4, Wu, D.-M., 983. A remark on a generalized specification test. Economics Letters,
Statistical models and likelihood functions
Statistical models and likelihood functions Jean-Marie Dufour McGill University First version: February 1998 This version: February 11, 2008, 12:19pm This work was supported by the William Dow Chair in
More informationExogeneity tests and weak identification
Cireq, Cirano, Départ. Sc. Economiques Université de Montréal Jean-Marie Dufour Cireq, Cirano, William Dow Professor of Economics Department of Economics Mcgill University June 20, 2008 Main Contributions
More informationAsymptotic theory for linear regression and IV estimation
Asymtotic theory for linear regression and IV estimation Jean-Marie Dufour McGill University First version: November 20 Revised: December 20 his version: December 20 Comiled: December 3, 20, : his work
More informationSpecification errors in linear regression models
Specification errors in linear regression models Jean-Marie Dufour McGill University First version: February 2002 Revised: December 2011 This version: December 2011 Compiled: December 9, 2011, 22:34 This
More informationComments on Weak instrument robust tests in GMM and the new Keynesian Phillips curve by F. Kleibergen and S. Mavroeidis
Comments on Weak instrument robust tests in GMM and the new Keynesian Phillips curve by F. Kleibergen and S. Mavroeidis Jean-Marie Dufour First version: August 2008 Revised: September 2008 This version:
More informationBinary variables in linear regression
Binary variables in linear regression Jean-Marie Dufour McGill University First version: October 1979 Revised: April 2002, July 2011, December 2011 This version: December 2011 Compiled: December 8, 2011,
More informationOn the finite-sample theory of exogeneity tests with possibly non-gaussian errors and weak identification
On the finite-sample theory of exogeneity tests with possibly non-gaussian errors and weak identification Firmin Doko Tchatoka University of Tasmania Jean-Marie Dufour McGill University First version:
More informationDistribution and quantile functions
Distribution and quantile functions Jean-Marie Dufour McGill University First version: November 1995 Revised: June 2011 This version: June 2011 Compiled: April 7, 2015, 16:45 This work was supported by
More informationIdentification-Robust Inference for Endogeneity Parameters in Linear Structural Models
SCHOOL OF ECONOMICS AND FINANCE Discussion Paper 2012-07 Identification-Robust Inference for Endogeneity Parameters in Linear Structural Models Firmin Doko Tchatoka and Jean-Marie Dufour ISSN 1443-8593
More informationNew Keynesian Phillips Curves, structural econometrics and weak identification
New Keynesian Phillips Curves, structural econometrics and weak identification Jean-Marie Dufour First version: December 2006 This version: December 7, 2006, 8:29pm This work was supported by the Canada
More informationExogeneity tests, incomplete models, weak identification and non-gaussian distributions: invariance and finite-sample distributional theory
Exogeneity tests, incomplete models, weak identification and non-gaussian distributions: invariance and finite-sample distributional theory Firmin Doko Tchatoka The University of Adelaide Jean-Marie Dufour
More informationExogeneity Tests, Incomplete Models, Weak Identification and Non-Gaussian Distributions: Invariance and Finite- Sample Distributional Theory
School of Economics Working Papers ISSN 2203-6024 Exogeneity Tests, Incomplete Models, Weak Identification and Non-Gaussian Distributions: Invariance and Finite- Sample Distributional Theory Firmin Doko
More informationExogeneity tests, weak identification, incomplete models and non-gaussian distributions: Invariance and finite-sample distributional theory
School of Economics Working Papers ISSN 2203-6024 Exogeneity tests, weak identification, incomplete models and non-gaussian distributions: Invariance and finite-sample distributional theory Firmin Doko
More informationModel selection criteria Λ
Model selection criteria Λ Jean-Marie Dufour y Université de Montréal First version: March 1991 Revised: July 1998 This version: April 7, 2002 Compiled: April 7, 2002, 4:10pm Λ This work was supported
More informationSequences and series
Sequences and series Jean-Marie Dufour McGill University First version: March 1992 Revised: January 2002, October 2016 This version: October 2016 Compiled: January 10, 2017, 15:36 This work was supported
More informationIdentification-robust inference for endogeneity parameters in linear structural models
Identification-robust inference for endogeneity parameters in linear structural models Firmin Doko Tchatoka University of Tasmania Jean-Marie Dufour McGill University January 2014 This paper is forthcoming
More informationInstrument endogeneity and identification-robust tests: some analytical results
MPRA Munich Personal RePEc Archive Instrument endogeneity and identification-robust tests: some analytical results Firmin Sabro Doko chatoka and Jean-Marie Dufour 3. May 2008 Online at htt://mra.ub.uni-muenchen.de/2963/
More informationFinite-sample inference in econometrics and statistics
Finite-sample inference in econometrics and statistics Jean-Marie Dufour First version: December 1998 This version: November 29, 2006, 4:04pm This work was supported by the Canada Research Chair Program
More informationEconometrics. Week 8. Fall Institute of Economic Studies Faculty of Social Sciences Charles University in Prague
Econometrics Week 8 Institute of Economic Studies Faculty of Social Sciences Charles University in Prague Fall 2012 1 / 25 Recommended Reading For the today Instrumental Variables Estimation and Two Stage
More informationDiscussion Paper On the Validity of Durbin-Wu-Hausman Tests for Assessing Partial Exogeneity Hypotheses with Possibly Weak Instruments
SCHOOL OF ECONOMICS AND FINANCE Discussion Paper 2012-04 On the Validity of Durbin-Wu-Hausman Tests for Assessing Partial Exogeneity Hypotheses with Possibly Weak Instruments Firmin Doko Tchatoka ISSN
More informationInstrumental Variables, Simultaneous and Systems of Equations
Chapter 6 Instrumental Variables, Simultaneous and Systems of Equations 61 Instrumental variables In the linear regression model y i = x iβ + ε i (61) we have been assuming that bf x i and ε i are uncorrelated
More informationLikelihood Ratio Based Test for the Exogeneity and the Relevance of Instrumental Variables
Likelihood Ratio Based est for the Exogeneity and the Relevance of Instrumental Variables Dukpa Kim y Yoonseok Lee z September [under revision] Abstract his paper develops a test for the exogeneity and
More informationORTHOGONALITY TESTS IN LINEAR MODELS SEUNG CHAN AHN ARIZONA STATE UNIVERSITY ABSTRACT
ORTHOGONALITY TESTS IN LINEAR MODELS SEUNG CHAN AHN ARIZONA STATE UNIVERSITY ABSTRACT This paper considers several tests of orthogonality conditions in linear models where stochastic errors may be heteroskedastic
More informationFinite Sample Performance of A Minimum Distance Estimator Under Weak Instruments
Finite Sample Performance of A Minimum Distance Estimator Under Weak Instruments Tak Wai Chau February 20, 2014 Abstract This paper investigates the nite sample performance of a minimum distance estimator
More informationSpecification Test for Instrumental Variables Regression with Many Instruments
Specification Test for Instrumental Variables Regression with Many Instruments Yoonseok Lee and Ryo Okui April 009 Preliminary; comments are welcome Abstract This paper considers specification testing
More informationApproximate Distributions of the Likelihood Ratio Statistic in a Structural Equation with Many Instruments
CIRJE-F-466 Approximate Distributions of the Likelihood Ratio Statistic in a Structural Equation with Many Instruments Yukitoshi Matsushita CIRJE, Faculty of Economics, University of Tokyo February 2007
More informationBootstrapping Heteroskedasticity Consistent Covariance Matrix Estimator
Bootstrapping Heteroskedasticity Consistent Covariance Matrix Estimator by Emmanuel Flachaire Eurequa, University Paris I Panthéon-Sorbonne December 2001 Abstract Recent results of Cribari-Neto and Zarkos
More informationA CONDITIONAL LIKELIHOOD RATIO TEST FOR STRUCTURAL MODELS. By Marcelo J. Moreira 1
Econometrica, Vol. 71, No. 4 (July, 2003), 1027 1048 A CONDITIONAL LIKELIHOOD RATIO TEST FOR STRUCTURAL MODELS By Marcelo J. Moreira 1 This paper develops a general method for constructing exactly similar
More informationEconometrics of Panel Data
Econometrics of Panel Data Jakub Mućk Meeting # 3 Jakub Mućk Econometrics of Panel Data Meeting # 3 1 / 21 Outline 1 Fixed or Random Hausman Test 2 Between Estimator 3 Coefficient of determination (R 2
More informationMarkov-Switching Models with Endogenous Explanatory Variables. Chang-Jin Kim 1
Markov-Switching Models with Endogenous Explanatory Variables by Chang-Jin Kim 1 Dept. of Economics, Korea University and Dept. of Economics, University of Washington First draft: August, 2002 This version:
More informationTests of the Present-Value Model of the Current Account: A Note
Tests of the Present-Value Model of the Current Account: A Note Hafedh Bouakez Takashi Kano March 5, 2007 Abstract Using a Monte Carlo approach, we evaluate the small-sample properties of four different
More informationEconometric Analysis of Cross Section and Panel Data
Econometric Analysis of Cross Section and Panel Data Jeffrey M. Wooldridge / The MIT Press Cambridge, Massachusetts London, England Contents Preface Acknowledgments xvii xxiii I INTRODUCTION AND BACKGROUND
More informationIncreasing the Power of Specification Tests. November 18, 2018
Increasing the Power of Specification Tests T W J A. H U A MIT November 18, 2018 A. This paper shows how to increase the power of Hausman s (1978) specification test as well as the difference test in a
More informationECON Introductory Econometrics. Lecture 16: Instrumental variables
ECON4150 - Introductory Econometrics Lecture 16: Instrumental variables Monique de Haan (moniqued@econ.uio.no) Stock and Watson Chapter 12 Lecture outline 2 OLS assumptions and when they are violated Instrumental
More informationBirkbeck Working Papers in Economics & Finance
ISSN 1745-8587 Birkbeck Working Papers in Economics & Finance Department of Economics, Mathematics and Statistics BWPEF 1809 A Note on Specification Testing in Some Structural Regression Models Walter
More informationApplied Econometrics (MSc.) Lecture 3 Instrumental Variables
Applied Econometrics (MSc.) Lecture 3 Instrumental Variables Estimation - Theory Department of Economics University of Gothenburg December 4, 2014 1/28 Why IV estimation? So far, in OLS, we assumed independence.
More informationFinite Sample and Optimal Inference in Possibly Nonstationary ARCH Models with Gaussian and Heavy-Tailed Errors
Finite Sample and Optimal Inference in Possibly Nonstationary ARCH Models with Gaussian and Heavy-Tailed Errors J and E M. I Université de Montréal University of Alicante First version: April 27th, 2004
More informationEconometrics of Panel Data
Econometrics of Panel Data Jakub Mućk Meeting # 6 Jakub Mućk Econometrics of Panel Data Meeting # 6 1 / 36 Outline 1 The First-Difference (FD) estimator 2 Dynamic panel data models 3 The Anderson and Hsiao
More informationEcon 510 B. Brown Spring 2014 Final Exam Answers
Econ 510 B. Brown Spring 2014 Final Exam Answers Answer five of the following questions. You must answer question 7. The question are weighted equally. You have 2.5 hours. You may use a calculator. Brevity
More informationInstrumental Variables Estimation in Stata
Christopher F Baum 1 Faculty Micro Resource Center Boston College March 2007 1 Thanks to Austin Nichols for the use of his material on weak instruments and Mark Schaffer for helpful comments. The standard
More informationEconometrics. Week 4. Fall Institute of Economic Studies Faculty of Social Sciences Charles University in Prague
Econometrics Week 4 Institute of Economic Studies Faculty of Social Sciences Charles University in Prague Fall 2012 1 / 23 Recommended Reading For the today Serial correlation and heteroskedasticity in
More information1 Estimation of Persistent Dynamic Panel Data. Motivation
1 Estimation of Persistent Dynamic Panel Data. Motivation Consider the following Dynamic Panel Data (DPD) model y it = y it 1 ρ + x it β + µ i + v it (1.1) with i = {1, 2,..., N} denoting the individual
More informationFINITE SAMPLE TESTS FOR ARCH EFFECTS AND VARIANCE CHANGE-POINTS IN LINEAR REGRESSIONS
FINITE SAMPLE TESTS FOR ARCH EFFECTS AND VARIANCE CHANGE-POINTS IN LINEAR REGRESSIONS Jean-Marie Dufour a, Lynda Khalaf b, Jean-Thomas Bernard c and Ian Genest d a CIRANO, CIREQ, and Département de sciences
More informationInference about Clustering and Parametric. Assumptions in Covariance Matrix Estimation
Inference about Clustering and Parametric Assumptions in Covariance Matrix Estimation Mikko Packalen y Tony Wirjanto z 26 November 2010 Abstract Selecting an estimator for the variance covariance matrix
More informationA Non-Parametric Approach of Heteroskedasticity Robust Estimation of Vector-Autoregressive (VAR) Models
Journal of Finance and Investment Analysis, vol.1, no.1, 2012, 55-67 ISSN: 2241-0988 (print version), 2241-0996 (online) International Scientific Press, 2012 A Non-Parametric Approach of Heteroskedasticity
More informationTest of hypotheses with panel data
Stochastic modeling in economics and finance November 4, 2015 Contents 1 Test for poolability of the data 2 Test for individual and time effects 3 Hausman s specification test 4 Case study Contents Test
More informationChapter 1. GMM: Basic Concepts
Chapter 1. GMM: Basic Concepts Contents 1 Motivating Examples 1 1.1 Instrumental variable estimator....................... 1 1.2 Estimating parameters in monetary policy rules.............. 2 1.3 Estimating
More informationSimultaneous Equations and Weak Instruments under Conditionally Heteroscedastic Disturbances
Simultaneous Equations and Weak Instruments under Conditionally Heteroscedastic Disturbances E M. I and G D.A. P Michigan State University Cardiff University This version: July 004. Preliminary version
More informationA better way to bootstrap pairs
A better way to bootstrap pairs Emmanuel Flachaire GREQAM - Université de la Méditerranée CORE - Université Catholique de Louvain April 999 Abstract In this paper we are interested in heteroskedastic regression
More informationNearly Unbiased Estimation in Dynamic Panel Data Models
TI 00-008/ Tinbergen Institute Discussion Paper Nearly Unbiased Estimation in Dynamic Panel Data Models Martin A. Carree Department of General Economics, Faculty of Economics, Erasmus University Rotterdam,
More informationMore efficient tests robust to heteroskedasticity of unknown form
More efficient tests robust to heteroskedasticity of unknown form Emmanuel Flachaire To cite this version: Emmanuel Flachaire. More efficient tests robust to heteroskedasticity of unknown form. Econometric
More informationChapter 2. Dynamic panel data models
Chapter 2. Dynamic panel data models School of Economics and Management - University of Geneva Christophe Hurlin, Université of Orléans University of Orléans April 2018 C. Hurlin (University of Orléans)
More informationIntroductory Econometrics
Based on the textbook by Wooldridge: : A Modern Approach Robert M. Kunst robert.kunst@univie.ac.at University of Vienna and Institute for Advanced Studies Vienna November 23, 2013 Outline Introduction
More informationDynamic Regression Models (Lect 15)
Dynamic Regression Models (Lect 15) Ragnar Nymoen University of Oslo 21 March 2013 1 / 17 HGL: Ch 9; BN: Kap 10 The HGL Ch 9 is a long chapter, and the testing for autocorrelation part we have already
More information1 Procedures robust to weak instruments
Comment on Weak instrument robust tests in GMM and the new Keynesian Phillips curve By Anna Mikusheva We are witnessing a growing awareness among applied researchers about the possibility of having weak
More informationLeast Squares Estimation of a Panel Data Model with Multifactor Error Structure and Endogenous Covariates
Least Squares Estimation of a Panel Data Model with Multifactor Error Structure and Endogenous Covariates Matthew Harding and Carlos Lamarche January 12, 2011 Abstract We propose a method for estimating
More informationStructural Estimation and Evaluation of Calvo-Style Inflation Models
Structural Estimation and Evaluation of Calvo-Style Inflation Models Jean-Marie Dufour Université de Montréal Lynda Khalaf Université Laval February 2006 Maral Kichian Bank of Canada Preliminary, Please
More informationEconometrics Master in Business and Quantitative Methods
Econometrics Master in Business and Quantitative Methods Helena Veiga Universidad Carlos III de Madrid Models with discrete dependent variables and applications of panel data methods in all fields of economics
More informationThe Role of "Leads" in the Dynamic Title of Cointegrating Regression Models. Author(s) Hayakawa, Kazuhiko; Kurozumi, Eiji
he Role of "Leads" in the Dynamic itle of Cointegrating Regression Models Author(s) Hayakawa, Kazuhiko; Kurozumi, Eiji Citation Issue 2006-12 Date ype echnical Report ext Version publisher URL http://hdl.handle.net/10086/13599
More informationThe performance of tests on endogeneity of subsets of explanatory variables scanned by simulation
Discussion Paper: 2011/13 The performance of tests on endogeneity of subsets of explanatory variables scanned by simulation Jan F. Kiviet and Milan Pleus www.ase.uva.nl/uva-econometrics Amsterdam School
More informationTesting Homogeneity Of A Large Data Set By Bootstrapping
Testing Homogeneity Of A Large Data Set By Bootstrapping 1 Morimune, K and 2 Hoshino, Y 1 Graduate School of Economics, Kyoto University Yoshida Honcho Sakyo Kyoto 606-8501, Japan. E-Mail: morimune@econ.kyoto-u.ac.jp
More informationTopic 10: Panel Data Analysis
Topic 10: Panel Data Analysis Advanced Econometrics (I) Dong Chen School of Economics, Peking University 1 Introduction Panel data combine the features of cross section data time series. Usually a panel
More informationIdentification-robust inference for endogeneity parameters in linear structural models
2014s-17 Identification-robust inference for endogeneity parameters in linear structural models Firmin Doko Tchatoka, Jean-Marie Dufour Série Scientifique Scientific Series Montréal Février 2014/February
More informationHeteroskedasticity-Robust Inference in Finite Samples
Heteroskedasticity-Robust Inference in Finite Samples Jerry Hausman and Christopher Palmer Massachusetts Institute of Technology December 011 Abstract Since the advent of heteroskedasticity-robust standard
More informationRecent Advances in the Field of Trade Theory and Policy Analysis Using Micro-Level Data
Recent Advances in the Field of Trade Theory and Policy Analysis Using Micro-Level Data July 2012 Bangkok, Thailand Cosimo Beverelli (World Trade Organization) 1 Content a) Endogeneity b) Instrumental
More informationMultiple Equation GMM with Common Coefficients: Panel Data
Multiple Equation GMM with Common Coefficients: Panel Data Eric Zivot Winter 2013 Multi-equation GMM with common coefficients Example (panel wage equation) 69 = + 69 + + 69 + 1 80 = + 80 + + 80 + 2 Note:
More informationConsistent OLS Estimation of AR(1) Dynamic Panel Data Models with Short Time Series
Consistent OLS Estimation of AR(1) Dynamic Panel Data Models with Short Time Series Kazuhiko Hayakawa Department of Economics Hitotsubashi University January 19, 006 Abstract In this paper, we examine
More informationTESTING FOR NORMALITY IN THE LINEAR REGRESSION MODEL: AN EMPIRICAL LIKELIHOOD RATIO TEST
Econometrics Working Paper EWP0402 ISSN 1485-6441 Department of Economics TESTING FOR NORMALITY IN THE LINEAR REGRESSION MODEL: AN EMPIRICAL LIKELIHOOD RATIO TEST Lauren Bin Dong & David E. A. Giles Department
More informationUsing Instrumental Variables to Find Causal Effects in Public Health
1 Using Instrumental Variables to Find Causal Effects in Public Health Antonio Trujillo, PhD John Hopkins Bloomberg School of Public Health Department of International Health Health Systems Program October
More informationIV and IV-GMM. Christopher F Baum. EC 823: Applied Econometrics. Boston College, Spring 2014
IV and IV-GMM Christopher F Baum EC 823: Applied Econometrics Boston College, Spring 2014 Christopher F Baum (BC / DIW) IV and IV-GMM Boston College, Spring 2014 1 / 1 Instrumental variables estimators
More informationPermutation tests for comparing inequality measures
Permutation tests for comparing inequality measures Jean-Marie Dufour McGill University Emmanuel Flachaire Aix-Marseille University December 2015 Lynda Khalaf Carleton University This work was supported
More informationBootstrap Approach to Comparison of Alternative Methods of Parameter Estimation of a Simultaneous Equation Model
Bootstrap Approach to Comparison of Alternative Methods of Parameter Estimation of a Simultaneous Equation Model Olubusoye, O. E., J. O. Olaomi, and O. O. Odetunde Abstract A bootstrap simulation approach
More informationOnline Appendix. j=1. φ T (ω j ) vec (EI T (ω j ) f θ0 (ω j )). vec (EI T (ω) f θ0 (ω)) = O T β+1/2) = o(1), M 1. M T (s) exp ( isω)
Online Appendix Proof of Lemma A.. he proof uses similar arguments as in Dunsmuir 979), but allowing for weak identification and selecting a subset of frequencies using W ω). It consists of two steps.
More informationBootstrap Testing in Econometrics
Presented May 29, 1999 at the CEA Annual Meeting Bootstrap Testing in Econometrics James G MacKinnon Queen s University at Kingston Introduction: Economists routinely compute test statistics of which the
More informationDiscussion of Bootstrap prediction intervals for linear, nonlinear, and nonparametric autoregressions, by Li Pan and Dimitris Politis
Discussion of Bootstrap prediction intervals for linear, nonlinear, and nonparametric autoregressions, by Li Pan and Dimitris Politis Sílvia Gonçalves and Benoit Perron Département de sciences économiques,
More informationA Course on Advanced Econometrics
A Course on Advanced Econometrics Yongmiao Hong The Ernest S. Liu Professor of Economics & International Studies Cornell University Course Introduction: Modern economies are full of uncertainties and risk.
More informationSize and Power of the RESET Test as Applied to Systems of Equations: A Bootstrap Approach
Size and Power of the RESET Test as Applied to Systems of Equations: A Bootstrap Approach Ghazi Shukur Panagiotis Mantalos International Business School Department of Statistics Jönköping University Lund
More informationSmall-sample cluster-robust variance estimators for two-stage least squares models
Small-sample cluster-robust variance estimators for two-stage least squares models ames E. Pustejovsky The University of Texas at Austin Context In randomized field trials of educational interventions,
More informationEconomics 582 Random Effects Estimation
Economics 582 Random Effects Estimation Eric Zivot May 29, 2013 Random Effects Model Hence, the model can be re-written as = x 0 β + + [x ] = 0 (no endogeneity) [ x ] = = + x 0 β + + [x ] = 0 [ x ] = 0
More informationCOMPARISON OF GMM WITH SECOND-ORDER LEAST SQUARES ESTIMATION IN NONLINEAR MODELS. Abstract
Far East J. Theo. Stat. 0() (006), 179-196 COMPARISON OF GMM WITH SECOND-ORDER LEAST SQUARES ESTIMATION IN NONLINEAR MODELS Department of Statistics University of Manitoba Winnipeg, Manitoba, Canada R3T
More informationProblem Set #6: OLS. Economics 835: Econometrics. Fall 2012
Problem Set #6: OLS Economics 835: Econometrics Fall 202 A preliminary result Suppose we have a random sample of size n on the scalar random variables (x, y) with finite means, variances, and covariance.
More informationDepartment of Econometrics and Business Statistics
ISSN 440-77X Australia Department of Econometrics and Business Statistics http://wwwbusecomonasheduau/depts/ebs/pubs/wpapers/ The Asymptotic Distribution of the LIML Estimator in a artially Identified
More informationWeak Identification in Maximum Likelihood: A Question of Information
Weak Identification in Maximum Likelihood: A Question of Information By Isaiah Andrews and Anna Mikusheva Weak identification commonly refers to the failure of classical asymptotics to provide a good approximation
More informationInstrumental Variables
Università di Pavia 2010 Instrumental Variables Eduardo Rossi Exogeneity Exogeneity Assumption: the explanatory variables which form the columns of X are exogenous. It implies that any randomness in the
More informationNecessary and sufficient conditions for nonlinear parametric function identification
Necessary and sufficient conditions for nonlinear parametric function identification Jean-Marie Dufour Xin Liang February 22, 2015, 11:34 This work was supported by the William Dow Chair in Political Economy
More informationHeteroskedasticity. Part VII. Heteroskedasticity
Part VII Heteroskedasticity As of Oct 15, 2015 1 Heteroskedasticity Consequences Heteroskedasticity-robust inference Testing for Heteroskedasticity Weighted Least Squares (WLS) Feasible generalized Least
More informationThe Case Against JIVE
The Case Against JIVE Related literature, Two comments and One reply PhD. student Freddy Rojas Cama Econometrics Theory II Rutgers University November 14th, 2011 Literature 1 Literature 2 Key de nitions
More informationAdvanced Econometrics
Advanced Econometrics Dr. Andrea Beccarini Center for Quantitative Economics Winter 2013/2014 Andrea Beccarini (CQE) Econometrics Winter 2013/2014 1 / 156 General information Aims and prerequisites Objective:
More informationIntroduction to Estimation Methods for Time Series models. Lecture 1
Introduction to Estimation Methods for Time Series models Lecture 1 Fulvio Corsi SNS Pisa Fulvio Corsi Introduction to Estimation () Methods for Time Series models Lecture 1 SNS Pisa 1 / 19 Estimation
More informationEconometrics II - EXAM Answer each question in separate sheets in three hours
Econometrics II - EXAM Answer each question in separate sheets in three hours. Let u and u be jointly Gaussian and independent of z in all the equations. a Investigate the identification of the following
More informationDEPARTMENT OF ECONOMICS
ISSN 0819-64 ISBN 0 7340 616 1 THE UNIVERSITY OF MELBOURNE DEPARTMENT OF ECONOMICS RESEARCH PAPER NUMBER 959 FEBRUARY 006 TESTING FOR RATE-DEPENDENCE AND ASYMMETRY IN INFLATION UNCERTAINTY: EVIDENCE FROM
More informationSimultaneous Equation Models
Simultaneous Equation Models Suppose we are given the model y 1 Y 1 1 X 1 1 u 1 where E X 1 u 1 0 but E Y 1 u 1 0 We can often think of Y 1 (and more, say Y 1 )asbeing determined as part of a system of
More informationLecture 8: Instrumental Variables Estimation
Lecture Notes on Advanced Econometrics Lecture 8: Instrumental Variables Estimation Endogenous Variables Consider a population model: y α y + β + β x + β x +... + β x + u i i i i k ik i Takashi Yamano
More informationBootstrapping heteroskedastic regression models: wild bootstrap vs. pairs bootstrap
Bootstrapping heteroskedastic regression models: wild bootstrap vs. pairs bootstrap Emmanuel Flachaire To cite this version: Emmanuel Flachaire. Bootstrapping heteroskedastic regression models: wild bootstrap
More informationLeast Squares Estimation-Finite-Sample Properties
Least Squares Estimation-Finite-Sample Properties Ping Yu School of Economics and Finance The University of Hong Kong Ping Yu (HKU) Finite-Sample 1 / 29 Terminology and Assumptions 1 Terminology and Assumptions
More informationReference Groups and Individual Deprivation
2004-10 Reference Groups and Individual Deprivation BOSSERT, Walter D'AMBROSIO, Conchita Département de sciences économiques Université de Montréal Faculté des arts et des sciences C.P. 6128, succursale
More informationA Test of Cointegration Rank Based Title Component Analysis.
A Test of Cointegration Rank Based Title Component Analysis Author(s) Chigira, Hiroaki Citation Issue 2006-01 Date Type Technical Report Text Version publisher URL http://hdl.handle.net/10086/13683 Right
More informationThe Estimation of Simultaneous Equation Models under Conditional Heteroscedasticity
The Estimation of Simultaneous Equation Models under Conditional Heteroscedasticity E M. I and G D.A. P University of Alicante Cardiff University This version: April 2004. Preliminary and to be completed
More informationIV Estimation and its Limitations: Weak Instruments and Weakly Endogeneous Regressors
IV Estimation and its Limitations: Weak Instruments and Weakly Endogeneous Regressors Laura Mayoral IAE, Barcelona GSE and University of Gothenburg Gothenburg, May 2015 Roadmap Deviations from the standard
More informationxtseqreg: Sequential (two-stage) estimation of linear panel data models
xtseqreg: Sequential (two-stage) estimation of linear panel data models and some pitfalls in the estimation of dynamic panel models Sebastian Kripfganz University of Exeter Business School, Department
More information