Case 1:12-cv RMC-DST-RLW Document 221 Filed 06/23/12 Page 1 of 4 IN THE UNITED STATES DISTRICT COURT FOR THE DISTRICT OF COLUMBIA

Transcription

1 Case 1:12-cv RMC-DST-RLW Document 221 Filed 06/23/12 Page 1 of 4 IN THE UNITED STATES DISTRICT COURT FOR THE DISTRICT OF COLUMBIA STATE OF TEXAS, v. Plaintiff, ERIC H. HOLDER, JR., in his official capacity as Attorney General of the United States, ERIC KENNIE, et al., Defendant. Defendant-Intervenors, TEXAS STATE CONFERENCE OF NAACP BRANCHES, et al., Defendant-Intervenors, TEXAS LEAGUE OF YOUNG VOTERS EDUCATION FUND, et al., Defendant-Intervenors. TEXAS LEGISLATIVE BLACK CAUCUS, et al., VICTORIA RODRIGUEZ, et al., Defendant-Intervenors, Defendant-Intervenors. ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) CASE NO. 1:12-CV (RMC-DST-RLW) Three-Judge Court ATTORNEY GENERAL S MOTION TO EXCLUDE THE EXPERT TESTIMONY OF DR. ALAN SAGER Pursuant to Federal Rule of Evidence 702 and Federal Rule of Civil Procedure 37(c), Defendant Eric H. Holder Jr. respectfully moves for an order excluding the expert testimony of

2 Case 1:12-cv RMC-DST-RLW Document 221 Filed 06/23/12 Page 2 of 4 Dr. Alan Sager. For the reasons set out in the Attorney General s Memorandum of Points and Authorities, the Court should exclude Dr. Sager s June 1, 2012 report from evidence, along with any testimony that Dr. Sager might otherwise offer concerning the contents of that report. Dr. Sager s opinion is based on a database match that he neither performed nor directed, and he lacks crucial knowledge concerning the process by which the State conducted the match. As a result, Dr. Sager s proposed testimony rests on unreliable speculation. Moreover Dr. Sager s testimony will not aid the Court as the finder of fact because it is based on incorrect assumptions concerning who may cast a ballot at a polling place under both current Texas law and SB 14. Finally, Dr. Sager failed to disclose the facts and data that he considered in forming his opinion, which has irreparably prejudiced the State. The Attorney General therefore respectfully requests that the Court enter the attached Proposed Order Excluding Expert Testimony. Date: June 23, 2012 RONALD C. MACHEN, JR. United States Attorney District of Columbia Respectfully submitted, THOMAS E. PEREZ Assistant Attorney General Civil Rights Division /s/ Daniel J. Freeman T. CHRISTIAN HERREN, JR. MEREDITH BELL-PLATTS ELIZABETH S. WESTFALL BRUCE I. GEAR SPENCER FISHER JENNIFER L. MARANZANO RISA BERKOWER DANIEL J. FREEMAN Attorneys Voting Section, Civil Rights Division U.S. Department of Justice 950 Pennsylvania Avenue, N.W. Washington, D.C

3 Case 1:12-cv RMC-DST-RLW Document 221 Filed 06/23/12 Page 3 of 4 CERTIFICATE OF SERVICE I hereby certify that on June 23, 2012, I served a true and correct copy of the foregoing via the Court s ECF system on the following counsel of record: Jonathan Franklin Mitchell Adam W. Aston Matthew Hamilton Frederick Patrick Kinney Sweeten Office of the Attorney General of Texas jonathan.mitchell@oag.state.tx.us adam.aston@oag.state.tx.us matthew.frederick@oag.state.tx.us Patrick.sweeten@texasattorneygeneral.gov Adam K. Mortara John M. Hughes Bartlit Beck Herman Palenchar & Scott LLP adam.mortara@bartlit-beck.com john.hughes@bartlit-beck.com Counsel for Plaintiff John Tanner john.k.tanner@gmail.com Nancy G. Abudu M. Laughlin McDonald Katie O Connor Arthur B. Spitzer American Civil Liberties Union nabudu@aclu.org lmcdonald@aclu.org koconnor@aclu.org artspitzer@gmail.com Debo P. Adegbile Leah C. Aden Elise C. Boddie Ryan Haygood Dale E. Ho Natasha Korgaonkar NAACP Legal Defense and Education Fund dadegbile@naacpldf.org laden@naacpldf.org eboddie@naacpldf.org rhaygood@naacpldf.org dho@naacpldf.org nkorgaonkar@naacpldf.org Michael Birney de Leeuw Douglas H. Flaum Adam M. Harris Fried, Frank, Harris, Shriver & Jacobson douglas.flaum@friedfrank.com adam.harris@friedfrank.com michael.deleeuw@friedfrank.com Counsel for Texas League of Young Voters Intervenors Counsel Texas Legislative Black Caucus Intervenors

4 Case 1:12-cv RMC-DST-RLW Document 221 Filed 06/23/12 Page 4 of 4 Jon M. Greenbaum Mark A. Posner Lawyers Committee for Civil Rights mposner@lawyerscommittee.org jgreenbaum@lawyerscommittee.org Ezra David Rosenberg Michelle Hart Yeary Dechert LLP ezra.rosenberg@dechert.com michelle.yeary@dechert.com Robert Stephen Notzon Robert@notzonlaw.com Gary L. Bledsoe Law Office of Gary L. Bledsoe and Associates garybledsoe@sbcglobal.net Nina Perales Amy Pederson Mexican American Legal Defense & Educational Fund, Inc. nperales@maldef.org apederson@maldef.org Counsel for Rodriguez Intervenors J. Gerald Hebert hebert@voterlaw.com Chad W. Dunn Brazil & Dunn chad@brazilanddunn.com Counsel for Kennie Intervenors Myrna Perez Wendy Robin Weiser Ian Arthur Vandewalker The Brennan Center for Justice myrna.perez@nyu.edu wendy.weiser@nyu.edu ian.vandewalker@nyu.edu Counsel for NAACP Intervenors s/ Daniel J. Freeman DANIEL J. FREEMAN Attorney, Voting Section Civil Rights Division U.S. Department of Justice 950 Pennsylvania Avenue, N.W. Washington, D.C

5 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 1 of 14 IN THE UNITED STATES DISTRICT COURT FOR THE DISTRICT OF COLUMBIA STATE OF TEXAS, v. Plaintiff, ERIC H. HOLDER, JR., in his official capacity as Attorney General of the United States, ERIC KENNIE, et al., Defendant. Defendant-Intervenors, TEXAS STATE CONFERENCE OF NAACP BRANCHES, et al., Defendant-Intervenors, TEXAS LEAGUE OF YOUNG VOTERS EDUCATION FUND, et al., Defendant-Intervenors. TEXAS LEGISLATIVE BLACK CAUCUS, et al., VICTORIA RODRIGUEZ, et al., Defendant-Intervenors, Defendant-Intervenors. ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) CASE NO. 1:12-CV (RMC-DST-RLW) Three-Judge Court MEMORANDUM OF POINTS AND AUTHORITIES IN SUPPORT OF THE ATTORNEY GENERAL S MOTION TO EXCLUDE EXPERT TESTIMONY Dr. Thomas Sager has provided an expert report in this case concerning database matches that he neither performed nor directed. He did not understand the programming language used to 1

6 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 2 of 14 perform the matches, and he could not replicate the matches when he attempted to do so. Rather, he proceeded on the basis of pure speculation that the matches had been properly performed. Moreover, in response to the Court s direct order that the State produce a witness with knowledge concerning the matches, the State produced Dr. Sager, who did not conduct the match and could not fully explain it. As a result, Dr. Sager s June 1 report in this matter and any testimony that he might offer concerning original database matches performed by the State is unreliable and must be excluded under Rule 702. In addition, Dr. Sager has failed to disclose the data and code underlying the matches. As a result, the June 1 report and testimony must be excluded under Rule 37. For the reasons, that follow, the Attorney General requests that this Court grant thist motion to exclude his expert testimony. 1 I. BACKGROUND On or about May 22, 2012, the Office of the Texas Attorney General ( OAG ) requested that Dr. Thomas Sager provide expert testimony on behalf of the State in this matter. Thomas Sager 30(b)(6) Deposition at 25:20-26:3, June 21, 2012 ( 30(b)(6) Dep. ) (rough transcript) (Ex. 1), Specifically, the State requested that Dr. Sager analyze two database matches performed by the State in December 2011 and May 2012 each of which compared the Texas voter registration database and a database containing records related to Texas driver licenses and official state identification cards ( TDL/ID ). Sager Report 1, 4-7 (Ex. 2). Only ten days later, he signed an expert report disclosing his analysis. See id. at 12. The matches upon which Dr. Sager relies purport to determine the number of registered voters who are likely not to possess a form of state-identification required in order to vote under 1 The instant motion is timely pursuant to the unopposed proposed order submitted to the Court on June 20, See Proposed Order Extending Motion Deadline (June 20, 2012) (Doc ). 2

7 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 3 of 14 Texas photographic voter identification law, Senate Bill 14 (2011) ( SB 14 ). Dr. Sager s initial report describes the match methodology in two steps. First the match includes those records from the... SOS voter registration database that fail to match entries in the... DPS database based upon the strict requirement that a match exists only if all three of First Name, Last Name, and date of birth (DOB) are exactly identical between the SOS and DPS databases. Sager. Rep. 5. Second, the match filters out all individuals who self-reported to SOS that they had such an identification. Id. Sager performed a series of further manipulations on these no match lists in an effort to eliminate what he considered to be false negatives. First, he eliminated all individuals in the voter registration database to whom he could match a TDL/ID record by full social security number. Id. 11, 18. Next, he eliminated all individuals in the voter registration database to whom he could match a TDL/ID record by the final four digits of a social security number, along with first name and last name. Id. Finally, he performed these same two matches to the State s license-to-carry database, which had not been considered in the initial matches. Id. 12, 19. As an additional step, Dr. Sager separated two additional classes of voters. Without explaining the rationale, Dr. Sager separated those voters who will be over the age of 65 at the date of the next general election. Id. 14, 21. In addition, he segregated those voters who are on the State of Texas s suspense list for voter registration status. Id. 15, 22. Therefore, Dr. Sager broke down the December and May lists based on three variables: further match, suspense, and over 65. Id. tbls. 3, 6. He then distributed each list into one of the eight combinations of those three variables and determined the percentage of voters in each group with a Spanish surname. Id. In sum, Dr. Sager concluded that 47% of voters in the December list who could 3

8 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 4 of 14 not be matched, are under 65, and are not on the suspense list have Spanish surnames, as do 33% of that same category of voters within the May list. Id. Dr. Sager did not consider whether the driver license records to which he matched voters reflected TDL/IDs that were neither expired nor suspended. Thomas Sager Expert Deposition at 77:22-80:19, June 21, 2012 ( Expert Dep. ) (rough transcript) (Ex. 3). He also did not consider whether driver license numbers self-reported on voter registration applications reflected an accurate match to a TDL/ID that was not expired or suspended. See, e.g., Id. at 63:13-65:2. In addition, Dr. Sager did not disclose the two database matches performed by the State in December 2011 and May 2012, Sager Dep. at 3:13-17, or the Statistical Analysis System programming code which Dr. Sager wrote to analyze the match lists. Expert Dep. at 44:24-46:2. Dr. Sager testified that the code itself is a text file that would allow someone familiar with the SAS language to understand precisely the criteria Dr. Sager applied in performing his analyses. Expert Dep. at 46: II. DR. SAGER S ANALYSIS IS BASED ON CONJECTURE AND IS NOT DESIGNED TO REFLECT THE NUMBER OF REGISTERED VOTERS WHO WILL BE AFFECTED BY SB 14. Under Federal Rule of Evidence 702(c), a qualified expert is permitted to testify only if the expert s testimony is the product of reliable principles and methods. See also Daubert v. Merrell Dow Pharm., Inc., 509 U.S. 579, 589 (1993) ( [U]nder the Rules [of Evidence] the trial judge must ensure that any and all scientific testimony or evidence admitted is not only relevant, but reliable. ). This rule requires that the proffering party show by a preponderance of the evidence that the testimony has a reliable basis in the knowledge and experience of [the relevant] discipline. Kumho Tire Co., Ltd. v. Carmichael, 526 U.S. 137, 149 (1999) (quoting Daubert, 509 U.S. at 592) (internal quotations omitted) (alteration in original). 4

9 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 5 of 14 The subject of an expert s testimony must be... more than subjective belief or unsupported speculation. Daubert, 509 U.S. at As a result, [t]he Court must exclude expert testimony that is speculative, guesswork, conjecture, and the like. Boyar v. Korean Air Lines Co., Ltd., 954 F. Supp. 4, 7 (D.D.C. 1996) (citing Joy v. Bell Helicopter Textron Inc., 999 F.2d 549, 569 (D.C. Cir. 1993). Similarly, an expert s testimony must be excluded if it is based on assumptions that are so unrealistic and contradictory as to suggest bad faith or to be in essence an apples and oranges comparison. Id. (quoting Boucher v. U.S. Suzuki Motor Corp, 73 F.3d 18, 21 (2d Cir.1996)). Along similar lines, under Rule 702(a), an expert may testify only to the extent that his opinion will help the trier of fact to understand the evidence or to determine a fact in issue. This condition goes primarily to relevance. Daubert, 509 U.S. at 591. Specifically, this rule requires a valid... connection to the pertinent inquiry as a precondition to admissibility. Kumho Tire, 526 U.S. at 149. Dr. Sager s proposed testimony concerning the December and May matches is based on conjecture concerning both the underlying match and the utility of the filters that he applied, as he admitted numerous times in his deposition. See, e.g., Expert Dep. at 62:2-66:3. 2 At a fundamental level, Dr. Sager did not understand the matches that lie at the heart of his data. Moreover, Dr. Sager s methodology is inherently biased, as it attempted to search the lists 2 Q. But you don't have any evidence as to why voters were placed on the suspense list -- or in suspense status is probably a better way of phrasing it? A. Well, I was told that I may assume that there was some issue with their voter registration; that mail had been returned that was addressed to the voter, therefore, a reasonable conjecture is that perhaps they moved. Perhaps they died. Perhaps they moved out of state, or maybe became -- was ineligible for other reasons. Q. But you used the word conjecture. All of those things are just conjecture, because you have no evidence on any of those possibilities, right? [Objection omitted.] A. I have no independent evidence of that. 5

10 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 6 of 14 produced by the State for false no-matches voter records not matched to a TDL/ID record that could have been while failing to consider the obvious occurrence of false matches voter records that were matched to a TDL/ID should not have been. See Rebuttal Report of Dr. Ansolabehere 36 & tbl.1 (June 11, 2012) (Ex. 9) (describing inability to replicate the matches without sufficient data or code)finally, Dr. Sager s testimony lacks a valid connection to the inquiry in this case because the filters he applied eliminated valid voters who are not exempted from the requirements of SB 14. A. Dr. Sager Did Not Understand the State s Matches and Could Not Replicate the State s Results. Dr. Sager admits in his report that he did not compile either the December or May No Match lists. Sager Rep Moreover, Dr. Sager explains what he believes to be the basic process, manner and method by which the State ran its comparisons, although he does not disclose the basis for his understanding as to how the State conducted the match, nor does he verify the accuracy of the underlying matching procedure. Id. 8. Moreover, he did not disclose until his deposition that he lacks any knowledge concerning the compiling, cleaning, or filtering of data carried out by the Department of Public Safety ( DPS ) or Office of the Secretary of State ( SOS ) prior to the SOS conducting the December match or the OAG performing the May match. See 30(b)(6) Dep. at 42:13-43:7; see also id. at 57:25-58:04 ( Q: So as a result, you have no knowledge of any processing that might have taken place before delivery of the [voter registration] database to the Office of the Attorney General for the May match? A: Correct. ). When asked to explain the large difference in size between the December list and the May list, Dr. Sager speculated that some quality control had been imposed, but he could not explain what the process had been. See 30(b)(6) Dep. at 105:15-111:13. Dr. Sager therefore based on his 6

11 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 7 of 14 analysis on simple conjecture that the data had been properly extracted and contained all valid voters and TDL/ID records. Furthermore, the OAG utilized SQL code in conducting its match. 30(b)(6)( Dep. at 65:3-67:4. SQL is an established and widely-utilized programming language for managing statistical databases. See, e.g., Oracle, SQL Language Reference 1.1 (2010) (Ex. 4). Dr. Sager was shown the SQL processing code used in the match, but Dr. Sager, a statistician, is not very familiar with SQL. 30(b)(6) Dep. at 66:7-67:9. As a result, Dr. Sager was unable to evaluate the means by which the State had conducted its match. See id. at 67:19-20 ( If you are asking me if I can judge whether the code did that, I have to rely upon [an OAG programmer s] representation. ). He could only speculate as to the specifics of how the match had been performed, see, e.g., id. at 70:23-71:7, and could only speculate as to whether a proper method had been applied. See id. at 73: An expert assessment of matches between two databases requires knowledge of how the match was conducted, as the methodology requires numerous decisions concerning match protocol. See, e.g., Ansolabehere Rep (setting out matching protocol) (Ex. 5). Dr. Sager does not know whether the match was case-sensitive. Sager Dep. at 69:19-70:13. He does 3 Q. And you don't know whether the matching code that was used would return one match, zero matches, or three matches? A. I am not sure. Q. But Mr. Valles would know? A. Yes. Q. And A. I would presume that since the purpose was to see how many of the voter registration records remained after doing the process that in the case of one-to-many, finding a match for the one would match out that particular individual, and then the program would go on to the next person on the list. Q. But that's just a presumption? A. That's a presumption. Q. That's speculation? A. Yes. 7

12 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 8 of 14 not know whether a match of two null entries for example a simple lack of an entry for first name in two records would constitute a match. Id. at 70:23-71:12. Finally, Dr. Sager does not know whether multiple voter registration records would be removed from the database if they all matched to a single driver license record based on name and date of birth. Id. at 71:22-72:21. Simply put, the December list and the May list are not Dr. Sager s matches, and Texas has produced no expert to opine on their reliability. When the Attorney General initially noticed a deposition pursuant to Rule 30(b)(6) requesting information concerning the underlying matches, the State moved for a protective order. See Mot. for Protective Order (June 13, 2012) (Doc. 182); see also Notice of Rule 30(b)(6) Deposition Ex. A 10, 12 (June 5, 2012) (Doc ). This Court denied the State s motion and ordered the production of witnesses with knowledge concerning the match. See Order on Texas s Motion for Protective Order (June 15, 2012) (Doc. 188). Despite Dr. Sager s lack of involvement with the December and May matches, the State designated Dr. Sager to testify concerning the match. See from Patrick Sweeten, OAG, to Daniel Freeman, U.S. Department of Justice (June 15, 2012) (Ex. 6). The result of that designation is clear the witness lacked essential knowledge concerning the match and was unable to testify as to deposition topics concerning which this Court has expressly ordered the State to produce a witness. The December and May matches remain entirely unverified and are not appropriate evidence in this case. Therefore, Dr. Sager should not be permitted to testify concerning the contents of the December and May lists. Dr. Sager has no way of assessing the No Match number arrived at by the State without knowing the precise process by which the State conducted its analysis, including whether there was a modification of the database used prior to Dr. Sager s calculations. Critically, Dr. Sager recognized that replicating the State s match was a desirable item that 8

13 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 9 of 14 would have allowed him to confirm the accuracy of the data, but when he attempted to replicate the State s matches, he was unable to do so. Expert Dep. at 66:2-72:15;. Dr. Sager s knowledge of the State s methodology and the accuracy of its execution prior to his own original assessment was based almost wholly speculation and conjecture. See, e.g., 30(b)(6) Dep. at 69:19-70:23 (Q: Was the matching [by the AG s Office] case sensitive? A: Yes, it was. Q: For both December and May? A: Let me backtrack. I presume that that s the case, but since I am not a I am not that familiar with SQL, I cannot say for sure. But I presume that is the case, that it is not case sensitive.) (emphasis added). Dr. Sager conducted his analysis based on how he thought the State compiled the No Match lists by comparing entrant s First Name, Last Name, and Date of Birth, even though he had access to and was shown the precise methodology utilized by the State. See 30(b)(6) Dep. at 66:7-24. He simply lacked sufficient knowledge and expertise to understand that methodology in the manner presented. In fact, Dr. Sager recognized the flaws in his methodology during his deposition and his lack of knowledge as to how those flaws affected the data presented in his report: Q. And do you plan to complete that work, the work of replicating the matching that Texas did prior to trial in this case? A. I would like to do so. Q. And why would you like to do so? A. Well, for one thing, this deposition has raised a number of questions about it and so I would like to follow up on that. Expert Dep. at 152: Therefore under Fed. R. Evid. 702(c), the State cannot establish that Dr. Sager s testimony concerning database matches is based on reliable principles and methods. This testimony should therefore be excluded by the Court. 9

14 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 10 of 14 B. Dr. Sager Attempted to Discover False Positives While Failing to Analyze the Presence of False Negatives. Dr. Sager s methodology is also inherently biased and is therefore not a proper statistical analysis. Typically, analysis of a matching procedure will examine both false positives and false negatives. See, e.g., United States v. Mitchell, 365 F.3d 215, 229 (3d Cir. 2004). However, Dr. Sager s procedure had a single-minded focus to reduce the number of voter registration entries that could not be matched to what he presumed to be a list of valid TDL/IDs. To that end, Dr. Sager admitted that he was simply following tasks provided by the State, rather than assessing the overall validity of the match. He was given two initial tasks : (1) to examine the State s May No Match and December No Match data sets, to see if, in fact, there were plausible additional matches that could be made to those data sets ; and (2) to look at the voter registration [and] see if there were plausible additional matches. Expert Dep. at 12:05-14:02; see also id. at 15:11-16, 17: On the other hand, he acknowledged that sufficient data existed to assess the presence of expired licenses and suspended licenses in the DPS database, 30(b)(6) Dep. at 50:6-50:17, but he did not analyze whether individuals who had been successfully matches and were therefore omitted from the December or May matches in fact held t only an expired or suspended TDL/ID. As a result, any proposed testimony based on his manipulation of the December and May lists is not the product of reliable principles and methods and must be excluded. Fed. R. Evid. 702(c). C. Dr. Sager Failed to Assess the Validity of the Filters Applied to the December and May Lists. Dr. Sager also relied upon and further narrowed the state s no-match list based on a faulty understanding of the operation of SB 14 and untested hypotheses. As a result, his opinion concerning the database match bears no valid... connection to the pertinent inquiry in this 10

15 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 11 of 14 case and will provide no assistance to this Court as the trier of fact. As a basic matter, Dr. Sager admittedly did not understand the photo identification requirements under SB 14. Expert Dep. at at 36:03-36:12 (acknowledging not knowing the universe of people who might be barred from voting under SB 14 ). This ignorance of the law s requirements, along with the blind assumption that the State had properly identified filters that would remove unaffected individuals, resulted in matching voters who lack current qualifying forms of ID and excluding them from his lists of affected voters. At the outset, Dr. Sager relied upon the State s initial match, which suffers from two blatant errors. First, the list was filtered for records with an entry in the field for a DL or ID number in the VR database, regardless of whether this entry was correct at the time the voter registered or has subsequently become expired, revoked or suspended. Dr. Sager s analysis does not account for this possibility. See Sager Rep In fact, Dr. Sager admitted he does not know whether an expired license constitutes acceptable voter identification under SB 14. See Expert Dep. at 33:21-34:09. Second, the State s matching protocol compared voter registrations to the complete DL database, which contains expired, revoked, and suspended identification that cannot be used as acceptable identification at the polls under SB 14. See Sager Report at 5; 30(b)(6) Dep. at 50:6-50:17. In fact, Texas never removes individuals from its DL database (other than law enforcement aliases). See Whelan Dep. at 32:20-33:3 (Ex. 7) ( Q. What are the purge procedures for [the] DPS database? A. There [are] none. We do not purge.... Q. The database will grow inexorably? A. Yes. ). Dr. Sager compounded the State s error through two mistakes of his own. First, Dr. Sager excluded voters above the age of 65. However, there is no exemption for individuals aged 65 or older under SB 14, and such individuals will be unable to vote at the polls unless they 11

16 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 12 of 14 present the required identification. See Senate Bill 14 (Ex. 8). Moreover, Dr. Sager did not base his age filter on any independent study or reliable third source. Sager Report at 14, 21; Expert Dep. at 58:10-50:4. Instead, because this population was eligible to vote by mail, it seemed to him, a reasonable procedure for him to separate out the population, Expert Dep. at 59:3-59:4, even though he was not given any information about that population s propensity to vote by mail. Id. at 61:11-61:13. Second, Dr. Sager excluded individuals on the voter registration suspense list from his analysis. Sager Rep. 15, 22. He based this filter only on his limited understanding of what being in suspense means and assumption that the voter has moved... and might not be eligible to vote in the same precinct. Expert Dep. at 32:16-33:01. Dr. Sager had no knowledge concerning whether a suspense voter who attempts to vote is able to or not. See id. at 33:2-5. In fact, suspense voters are eligible voters. 42 U.S.C. 1973gg-6; Tex. Elec. Code Again Dr. Sager s methodology is fatally flawed and will provide no assistance to the Court as trier of fact. III. DR. SAGER FAILED TO DISCLOSE FACTS AND DATA CONSIDERED IN FORMING HIS OPINION. Under Rules 26(a)(2)(B)(i) and 26(a)(2)(B)(ii), a party must include as part of its expert disclosures a written report that contains, among other things, a complete statement of all opinions the witness will express and the basis and reasons for them and the facts or data considered by the witness in forming them. [T]he intention is that facts or data be interpreted broadly to require disclosure of any material considered by the expert, from whatever source, that contains factual ingredients. The disclosure obligation extends to any facts or data considered by the expert by the expert in forming the opinions to be expressed, not only those relied upon by the expert. 12

17 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 13 of 14 Fed. R. Civ. P. 26, 2010 Amendments Comm. Notes. Where an expert fails to disclose information, he may not use that information at trial unless the failure to disclose was substantially justified or harmless. Fed. R. Civ. P. 37(c)(1). Dr. Sager did not disclose the underlying data or programming used to perform the database matches performed by the State in December 2011 and May See Expert Dep. at 3: These data were unquestionably considered by Dr. Sager in his analysis, even if he presented only the list of the match. Moreover, Dr. Sager never disclosed the SAS programming code that he wrote to analyze the match lists, id. at 44:24-46:2. The absence of these disclosures was in no way justified by the State. Nor can these failings be considered harmless, as they have precluded the Attorney General from replicating the State s initial database matches or Dr. Sager s analysis. See Ansolabehere Rebuttal 43 (describing inability to replicate the matches without sufficient data or code). Accordingly, Dr. Sager s June 1 Report and any testimony concerning the subject of the report should be excluded. CONCLUSION For the reasons set out above, this Court should grant the Attorney General s motion to exclude the June 1 Report offered by Dr. Sager, as well as any testimony concerning the contents of that report. 13

18 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 14 of 14 Date: June 23, 2012 RONALD C. MACHEN, JR. United States Attorney District of Columbia Respectfully submitted, THOMAS E. PEREZ Assistant Attorney General Civil Rights Division /s/ Daniel J. Freeman T. CHRISTIAN HERREN, JR. MEREDITH BELL-PLATTS ELIZABETH S. WESTFALL BRUCE I. GEAR JENNIFER L. MARANZANO RISA BERKOWER DANIEL J. FREEMAN Attorneys Voting Section, Civil Rights Division U.S. Department of Justice 950 Pennsylvania Avenue, N.W. Washington, D.C

19 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 1 of 2 IN THE UNITED STATES DISTRICT COURT FOR THE DISTRICT OF COLUMBIA STATE OF TEXAS, v. Plaintiff, ERIC H. HOLDER, JR., in his official capacity as Attorney General of the United States, ERIC KENNIE, et al., Defendant. Defendant-Intervenors, TEXAS STATE CONFERENCE OF NAACP BRANCHES, et al., Defendant-Intervenors, TEXAS LEAGUE OF YOUNG VOTERS EDUCATION FUND, et al., Defendant-Intervenors. TEXAS LEGISLATIVE BLACK CAUCUS, et al., VICTORIA RODRIGUEZ, et al., Defendant-Intervenors, Defendant-Intervenors. ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) CASE NO. 1:12-CV (RMC-DST-RLW) Three-Judge Court [PROPOSED] ORDER EXCLUDING EXPERT TESTIMONY Upon consideration of the Attorney General s Motion to Exclude Expert Testimony (Doc. ), the State of Texas s response thereto (Doc. ), and the Attorney General s Reply Memorandum in Further Support, it is hereby ORDERED that:

20 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 2 of 2 1. The Attorney General s motion to exclude expert testimony is GRANTED. 2. The June 1, 2012 Expert Report of Dr. Alan Sager is hereby excluded from evidence. 3. Dr. Alan Sager may not testify concerning opinions disclosed in his June 1, 2012 Expert Report. SO ORDERED. Date: June, 2012 DAVID S. TATEL United States Circuit Judge ROSEMARY M. COLLYER United States District Judge ROBERT L. WILKINS United States District Judge

21 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 1 of : :59 2 CASE NAME: TEXAS V. HOLDER, ET AL 09:59 3 WITNESS: TEXAS 30(b)(6) THOMAS W. SAGER, Ph.D. 09:59 4 DATE OF DEPOSITION: JUNE 21, :59 5 WARNING! 09:59 6 This unedited rough draft of the proceedings was 09:59 7 produced in instant form and is not certified. The 09:59 8 rough draft transcript may not be cited or used in any 09:59 9 way or at any time to rebut or contradict the certified 09:59 10 transcription of proceedings. There will be 09:59 11 discrepancies in this form and the final form, because 09:59 12 this instant form has not been edited, proofread, 09:59 13 corrected, finalized, indexed, bound or certified. 09:59 14 There will also be a discrepancy in page numbers 09:59 15 appearing on the unedited rough draft and the edited, 09:59 16 proofread, corrected and certified final. 10:00 17 THE COURT REPORTER: If all counsel will 10:00 18 make their announcements. The time is 10:00 a.m. 10:00 19 MR. SELLS: My name is Bryan Sells, and I 10:00 20 am counsel for the defendant, Eric Holder, Attorney 10:00 21 General of the United States. 10:00 22 MR. BRISSENDEN: Reynolds Brissenden from 10:00 23 the Texas Attorney General's Office on behalf of the 10:00 24 State of Texas. 10:00 25 MR. SWEETEN: I'm Patrick Sweeten on

22 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 2 of :00 1 behalf the State of Texas and on behalf of the witness. 10:00 2 MR. HARRIS: This is Adam Harris from 10:00 3 Fried, Frank, Harris, Shriver & Jacobson, LLP, on behalf 10:00 4 of Defendant Intervenors, the Texas League of Young 10:00 5 Voters Education Fund, and with me on the line is my 10:00 6 colleague, Deuel Ross, D-e-u-e-l, last name, Ross, 10:01 7 R-o-s-s. 09:57 8 THOMAS SAGER, Ph.D., 09:57 9 having been first duly sworn to testify the truth, the 09:57 10 whole truth, and nothing but the truth, testified as 09:57 11 follows: 10:01 12 EXAMINATION 10:01 13 BY MR. SELLS: 10:01 14 Q. Good morning, Professor Sager. Thank you for 10:01 15 coming. Sorry you had a bit of a problem with the 10:01 16 security or a delay in getting through, but it's there 10:01 17 for a reason. 10:01 18 A. Good morning, Mr. Sells. 10:01 19 Q. Thank you. I've already introduced myself 10:01 20 prior to going on the record, but again, I am Bryan 10:01 21 Sells with the United States Department of Justice in 10:01 22 Washington, D.C., and I represent Eric Holder, the 10:01 23 Attorney General of the United States, in this matter. 10:01 24 Would you please state your name for the 10:01 25 record.

23 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 3 of :01 1 A. Thomas William Sager, S-a-g-e-r. 10:01 2 Q. And have you ever been deposed before, 10:01 3 Professor Sager? 10:01 4 A. Yes. 10:01 5 Q. When was the last time? 10:01 6 A. Gee, I think that was a couple of years ago. I 10:02 7 have testified since then, without having given a 10:02 8 deposition as a preliminary. 10:02 9 Q. Okay. Well, since it's been a few years since 10:02 10 you've given a deposition, I'll go over some guidelines 10:02 11 and ground rules. I'll do them fairly quickly, though, 10:02 12 since you have been deposed before. 10:02 13 You are under oath, subject to penalties 10:02 14 of perjury. I will be asking questions, you will be 10:02 15 answering the questions, and the court reporter will be 10:02 16 transcribing your answers. Do you understand that so 10:02 17 far? 10:02 18 A. Yes. 10:02 19 Q. Okay. Because the court reporter is taking 10:02 20 down what you say, he can't record gestures and 10:02 21 nonverbal responses, so I need you to give a verbal 10:02 22 response to all of my questions. Is that okay? 10:02 23 A. Yes. 10:02 24 Q. All right. And we may get into some technical 10:02 25 areas today, and if we do that, I may ask a question in

24 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 4 of :03 1 a way that's not clear to you, because we're not on the 10:03 2 same page technically, and if I ask something that is in 10:03 3 any way unclear, would you let me know so that I can 10:03 4 rephrase it so that it is clear to you? 10:03 5 A. Certainly. 10:03 6 Q. If you need a break, we can take a break, not 10:03 7 while a question is pending, but if you need a break for 10:03 8 any other reason, would you let me know? 10:03 9 A. Certainly. 10:03 10 Q. Great. Now, sometimes, you remember things 10:03 11 later in the day, and if that happens, would you let me 10:03 12 know so that we can add it to the record, and I'll give 10:03 13 you chances to do that as well. 10:03 14 A. Yes. 10:03 15 Q. Okay. And in a similar vein, sometimes after 10:03 16 we've been talking for a while, you'll realize that 10:03 17 something that you said earlier may have been incorrect 10:03 18 or inaccurate. If that comes up, would you also let me 10:03 19 know so we can correct the record? 10:03 20 A. Yes. 10:03 21 Q. Sometimes in answering a question, a document 10:04 22 becomes important to you that might answer the 10:04 23 question. If there's a document that you need to answer 10:04 24 a question, would you let me know that? 10:04 25 A. Yes.

25 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 5 of :04 1 Q. I may have it here, and we may be able to 10:04 2 refresh your memory on that. 10:04 3 A. Yes. 10:04 4 Q. Great. Are you on any medication or drugs of 10:04 5 any kind that would impair your ability to understand 10:04 6 and answer the questions? 10:04 7 A. No. 10:04 8 Q. Have you had anything alcoholic to drink in the 10:04 9 last eight hours? 10:04 10 A. No. 10:04 11 Q. Are you sick at all today? 10:04 12 A. No. 10:04 13 Q. Under a doctor's care for any illness? 10:04 14 A. No. 10:04 15 Q. Is there any reason you could think of why you 10:04 16 might not be able to understand and answer my questions 10:04 17 fully and completely? 10:04 18 A. Well, I hope not. Of course, I am suffering 10:04 19 from jet lag, but that goes with the nature of things. 10:04 20 Q. Okay. Well, if the jet lag becomes an issue at 10:05 21 all today, let me know. 10:05 22 A. I have my mother's milk right here, referring 10:05 23 to coffee. 10:05 24 Q. Yes. The record will reflect that Professor 10:05 25 Sager has a large coffee in front of him.

26 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 6 of :05 1 Any questions so far for me? 10:05 2 A. No. 10:05 3 Q. Okay, great. 10:05 4 Where were you born and raised, Professor 10:05 5 Sager? 10:05 6 A. I was born in the town of Centerville in the 10:05 7 state of Iowa, in the year 1946 on August 11th. 10:05 8 Q. And what's your educational background? 10:05 9 A. Well, skipping high school, I have a bachelor's 10:05 10 degree and a double major in mathematics and economics 10:05 11 from the University of Iowa in I have a master's 10:05 12 degree from the University of Iowa in statistics in 10: , and a Ph.D. in statistics from the University of 10:06 14 Iowa in :06 15 Q. So you're pretty much Iowa born, raised, and 10:06 16 educated then? 10:06 17 A. I have three degrees from Iowa, yes. 10:06 18 Q. Great. This may be demonstrating my ignorance, 10:06 19 but when I was in college, my statistics textbook was 10:06 20 by, I think, some folks from Iowa, Snedecor and Cochran? 10:06 21 A. That would be Iowa State as opposed to the 10:06 22 University of Iowa. Iowa State is located in Ames, and 10:06 23 the University of Iowa, where I got my education, is in 10:06 24 Iowa City. 10:06 25 Q. Okay. So you've heard of Snedecor and Cochran,

27 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 7 of :06 1 then? 10:06 2 A. Yes, uh-huh. 10:06 3 Q. Right. And I thought it was amusing that all 10:06 4 of the examples in this -- I was using this for social 10:06 5 science, where it was all about corn and crops and what 10:06 6 not. But statistics is statistics, isn't it? 10:06 7 A. Yes. 10:06 8 Q. All right. Tell me a little bit about your 10:07 9 employment background. 10:07 10 A. My first job, after getting the Ph.D. degree in 10: , was as an assistant professor of statistics at 10:07 12 Stanford University in California. I was there for five 10:07 13 years. And then I moved to Texas in 1978 to take 10:07 14 employment with the University of Texas at Austin where 10:07 15 I have been ever since. 10:07 16 Q. Since 1978? 10:07 17 A. Correct. 10:07 18 Q. Do you have any family here? 10:07 19 A. Yes, I have a wife. 10:07 20 Q. Any children? 10:07 21 A. No. 10:07 22 Q. What is your current position? 10:07 23 A. I am professor of statistics in the department 10:07 24 of Information, Risk, and Operations Management, which 10:08 25 we call by the acronym IROM, for short, and that's in

28 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 8 of :08 1 the McCombs School of Business at the University of 10:08 2 Texas. 10:08 3 Q. And what courses do you presently teach in your 10: on a sort of a yearly basis, or usual basis, I should 10:08 5 say? 10:08 6 A. On an ongoing basis, I teach MBA statistics and 10:08 7 Ph.D.-level statistics. 10:08 8 Q. Do you use Snedecor and Cochran for those 10:08 9 courses by any chance? 10:08 10 A. I do not. I do not. 10:08 11 Q. What textbooks do you use for those courses? 10:08 12 A. For the MBA course, for many years, we have 10:08 13 used a book by Chris -- let's see -- it's Albright, 10:08 14 Winston and Zappe. If you'd like me to spell that, I 10:09 15 will. Zappe is Z-a-p-p-e. And Winston, I think, is 10:09 16 W-i-n-s-t-o-n, and A-l-b-r-i-g-h-t. 10:09 17 Q. And you use that in the MBA course? 10:09 18 A. Correct. 10:09 19 Q. What about the Ph.D. statistics; is there a 10:09 20 common book that you use for that? 10:09 21 A. Well, it depends upon which Ph.D. course I'm 10:09 22 teaching. For the last few years, I have taught a 10:09 23 course in what's -- it's called applied mathematical 10:09 24 statistics. It sounds like an oxymoron, but it really 10:09 25 isn't. We use -- it's essentially a course in applied

29 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 9 of :09 1 econometrics, and there is a book called, I think, 10:09 2 something in the same name by an author named Ramanujan. 10:09 3 I'll take pity on the court reporter here and see if I 10:09 4 spell that. I'm not sure I can, but R-a-m-a-n-u-j-a-n, 10:10 5 I think. 10:10 6 Q. Are there any reference books in statistics 10:10 7 that you regularly use and rely upon; desk books, that 10:10 8 sort of thing? 10:10 9 A. Well, yes. There are books that I consult from 10:10 10 time to time. A wide variety. I have a full bookcase 10:10 11 of them at my office. But statistics is such a broad 10:10 12 field that it is not encompassed within a long -- within 10:10 13 a single volume. They tend to be more specialized, like 10:10 14 you'll have books on certain areas of mathematics 10:10 15 statistics. There will be areas of applications in 10:10 16 social sciences, applications in the agriculture, like 10:10 17 you mentioned Snedecor and Cochran, applications in 10:11 18 various areas. 10:11 19 Q. If you were recommending a single reference 10:11 20 work for someone like me, for use in -- for general use, 10:11 21 but primarily social sciences, what would that be? 10:11 22 A. Well, there's a book that I have found quite 10:11 23 useful, because of its practical orientation and rules 10:11 24 of thumb. It's by an author named Stevens, and I'm not 10:11 25 sure of the vintage of it. It's been, oh, at least 10

30 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 10 of :11 1 years, maybe 20 years, and I don't remember the exact 10:11 2 title. I think the author is -- I think his first name 10:11 3 is James. James Stevens. 10:11 4 Q. In the course of your duties as a professor of 10:11 5 statistics, do you also conduct research and publish 10:12 6 papers? 10:12 7 A. Yes. 10:12 8 Q. Okay. In the course of doing that, what 10:12 9 software do you use for your analysis? 10:12 10 A. Primarily SAS, and that's S-A-S. It's an 10:12 11 acronym for the Statistical Analysis System. 10:12 12 Q. Do you have any certifications in SAS? 10:12 13 A. I do not. 10:12 14 Q. Are you sufficiently familiar with it that you 10:12 15 can do programming yourself on -- 10:12 16 A. Yes, uh-huh. 10:12 17 Q. Okay. 10:12 18 A. I have, in fact, used it for 30 years, and at 10:12 19 the University of Texas, I have taught courses in it. 10:12 20 Q. Well, my understanding of SAS is that there is 10:12 21 programming language, a code, if you will, that one has 10:12 22 to learn in order to run analyses. Are you familiar 10:12 23 with that programming language and code? 10:13 24 A. Well, SAS has, as a goal, to provide a one-stop 10:13 25 shop so that a statistician can do all of the --

31 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 11 of :13 1 essentially all of the work that he needs to do within 10:13 2 confines the SAS. That includes statistical analysis. 10:13 3 It includes programming. It includes data management, 10:13 4 report writing; all sorts of things. 10:13 5 Now, SAS comes fairly close to that, 10:13 6 although there are some things that you would prefer to 10:13 7 use other programs for because of their simplicity. 10:13 8 Q. Okay. But my question goes to whether you are 10:13 9 familiar and comfortable with the programming language 10:13 10 that is part of SAS? 10:13 11 A. Yes. 10:13 12 Q. Okay. In other words, you don't farm the 10:13 13 programming out to someone else on a regular basis? 10:13 14 A. That's right. I do it myself. 10:13 15 Q. Do you have any database certifications? 10:14 16 A. I do not. 10:14 17 Q. When you are conducting research for a 10:14 18 publication, or for some other reason, if you are asking 10:14 19 others to review that research, would you normally 10:14 20 provide your SAS program along with the data so that the 10:14 21 reviewer could see what you did? 10:14 22 A. It is not a common practice, within statistics, 10:14 23 at least, to do that in work that is submitted for 10:14 24 publication. One generally presents the results and 10:14 25 cites the programming language as the tool that was used

32 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 12 of :14 1 in this analysis. 10:14 2 Q. Are you talking now about the actual 10:15 3 publication itself or submitting something for 10:15 4 publication? 10:15 5 A. Submitting something for publication. 10:15 6 Q. And you don't normally submit your code along 10:15 7 with the article? 10:15 8 A. That is not the common practice. 10:15 9 Q. Have you ever seen it done? 10:15 10 A. And wen I think back over the reviewing that I 10:15 11 have done of articles in statistics or applications of 10:15 12 statistics, I cannot recall an instance. 10:15 13 Q. Do you advise on any dissertations? 10:15 14 A. Yes. 10:15 15 Q. Do you serve on my peer review panels for any 10:15 16 journals? 10:15 17 A. Yes. 10:15 18 Q. Which journals? 10:15 19 A. Well, for instance, I am associate editor of 10:15 20 the Journal of Risk and Insurance. 10:15 21 Q. Any others? 10:15 22 A. No. I'm not at the editorial level on others, 10:16 23 but I provide reviewing of articles for a wide variety 10:16 24 of journals. 10:16 25 Q. Now, if you received a dissertation draft or a

33 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 13 of :16 1 draft article that was submitted for publication to one 10:16 2 of your -- one of the publications on which you sit as a 10:16 3 reviewer, and you had a question about the data or 10:16 4 methodology or analysis, would it be possible for you to 10:16 5 obtain the code that the programmer or analyst used to 10:16 6 perform the analysis? 10:16 7 MR. BRISSENDEN: Objection, vague, assumes 10:16 8 facts not in evidence. 10:16 9 MR. SELLS: I'm sorry. Say again? 10:16 10 MR. BRISSSENDEN: It was objection, vague. 10:16 11 MR. SELLS: Okay. 10:16 12 THE WITNESS: So should I answer? 10:16 13 MR. BRISSSENDEN: Yes. 10:16 14 A. Okay. If that arose, I would -- I could 10:17 15 request the code, for instance, and that might go to the 10:17 16 editor of the journal, who might convey that to the 10:17 17 referee. But typically, I do not. In fact, I do not 10:17 18 recall any instance in my career in which I have done 10:17 19 so. If I note errors that I think are errors or have 10:17 20 questions raised, then I might point them out. 10:17 21 For instance, in one article that I 10:17 22 recently reviewed, it looked to me like some of the 10:17 23 figures were off by orders of magnitude. I put that 10:17 24 note into my review, and indeed, the authors of the 10:17 25 article responded that I was correct in that, and they

34 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 14 of :17 1 corrected the figures, but there was no exchange of 10:17 2 code. 10:17 3 Q. (By Mr. Sells) How about on a dissertation; 10:17 4 would there be any exchange of code on a dissertation? 10:17 5 A. Well, there could be in a dissertation. 10:17 6 Particularly if I were the supervisor of the student, I 10:17 7 might ask to see a student's code and look it over. 10:18 8 Q. What would that enable you to do, if you were 10:18 9 reviewing that code? 10:18 10 A. Well, it would enable me to verify, if there 10:18 11 seemed to be any errors in logic in setting it up. 10:18 12 Typically, I would not raise a question if the results 10:18 13 seemed reasonable. If the results did not seem 10:18 14 reasonable, I might ask, well, why is this the case 10:18 15 without specifying that it might be something in the 10:18 16 code, and then the student might independently find an 10:18 17 error in the code and report it to me. 10:18 18 Q. Does your current position as a professor of 10:18 19 statistics encompass the field of database matching? Is 10:18 20 that something you regularly do as a professor of 10:18 21 statistics? 10:18 22 A. It is not something that I regularly do. 10:19 23 Q. Has your position as professor of statistics 10:19 24 given you any experience in database matching? 10:19 25 A. Yes.

35 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 15 of :19 1 Q. How so? 10:19 2 A. Well, there was one consulting job that I had, 10:19 3 which was a contract that I had with the Texas 10:19 4 Department of Health. I think that was about 15 years 10:19 5 ago or so. It's noted on my resume under the Contract 10:19 6 section. You know, it was quite a while ago. 10:19 7 Q. Any other experience with database matching? 10:19 8 A. That was the main example that rises to the 10:19 9 level of really substantive work. There are time -- 10:19 10 from time to time, I might be interested in whether two 10:19 11 data sets correspond, and so there is a program, a 10:20 12 procedure in SAS, for instance, called PROC compare, 10:20 13 which can compare two databases and find their 10:20 14 similarities and their differences, which I have used 10:20 15 from time to time. But that does not do matching except 10:20 16 for aligning the rows in one database with the rows in 10:20 17 another and report things like the number of rows that 10:20 18 are same with the number of rows that are different. 10:20 19 Q. Would you spell PROC compare for the court 10:20 20 reporter, please. 10:20 21 A. I'm sorry. PROC, P-R-O-C. I'm sorry, I lapsed 10:20 22 into SAS speak. It's short for procedure. And then 10:20 23 compare, C-O-M-P-A-R-E. 10:20 24 Q. Thanks. How long ago was the Texas Department 10:20 25 of Health engagement?

36 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 16 of :20 1 A. I think it was about 15 years ago. I think the 10:20 2 date is probably on my resume, so we could check, if you 10:21 3 want an exact year. 10:21 4 Q. But you said that database matching with the 10:21 5 PROC COMPARE function is something that you might do 10:21 6 from time to time just as a matter of course in 10:21 7 conducting some kind of analysis? 10:21 8 A. On occasion. 10:21 9 Q. And is database matching a technique, in 10:21 10 general, used in the field of statistics? 10:21 11 A. If you mean do most statisticians do database 10:21 12 matching from time to time, I would guess, although I 10:21 13 don't really know, that the answer is probably no, they 10:21 14 probably do not, as a regular matter of course. 10:21 15 Q. But the idea of database matching isn't novel 10:21 16 or untested, is it? 10:21 17 A. No, it is not. 10:21 18 Q. Okay. Tell me a little bit more, if you can, 10:22 19 about your database matching with the Texas Department 10:22 20 of Health. What were you asked to do? 10:22 21 A. Boy, that was a long time ago. As I recall, 10:22 22 the Texas Department of Health was concerned with the 10:22 23 issue of matching immunization records of children, 10:22 24 where reports of immunization would be filed with the 10:22 25 department, which would then have to match that

37 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 17 of :22 1 immunization record with a master record that they had, 10:22 2 and they were concerned about doing -- doing a good job 10:22 3 on that. 10:22 4 Q. Can you tell me a little bit more what you mean 10:22 5 by concerned about doing a good job? 10:23 6 A. Well, they would note things, like when doing a 10:23 7 match, matching by name, first name, last name, 10:23 8 sometimes didn't give a good match, even matching by 10:23 9 social security number, in cases in which that was 10:23 10 applied. And they would -- they were concerned about 10:23 11 the mismatches because they would think that these might 10:23 12 be two individuals who are the same and maybe they are 10:23 13 not matching, or perhaps, on the other hand, two 10:23 14 individuals would be matched together who are not, in 10:23 15 fact, the same children, and so they were concerned 10:23 16 about those error rates. 10:23 17 Q. And why is that important in the field of 10:23 18 health? 10:23 19 A. Well, if you want to have up-to-date records 10:23 20 and accurate records, then you want to know, did this 10:23 21 child get the immunizations that are recorded for them 10:23 22 and that they not belong to somebody else. 10:24 23 Q. If I heard you correctly, and I hesitate to 10:24 24 throw out some jargon, but what I think you just said is 10:24 25 that the Texas Department of Health was concerned about

38 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 18 of :24 1 false positives and false negatives? 10:24 2 A. Correct. 10:24 3 Q. Okay. Tell me what a false positive would be 10:24 4 in that instance. 10:24 5 A. In the context of matching records, false 10:24 6 positive is generally considered to be two records that 10:24 7 are matched together, putatively, for the same 10:24 8 individual when, in fact, they are not. So you might 10:24 9 call that a false match. The matching is done, but it 10:24 10 is really not for the same individuals. And if your 10:24 11 follow-up question is, what is a false negative. 10:24 12 Q. Yes. 10:24 13 A. Then it's where you have two records that are 10:24 14 really for the same individual, but they are not matched 10:24 15 together, so the match fails when it, in fact, should 10:25 16 have happened. 10:25 17 Q. So, can you put it in real world terms for me 10:25 18 what would be the consequences of a false positive in 10:25 19 the example of child vaccinations? 10:25 20 A. Okay. False positive, where you have a false 10:25 21 match, would mean that one child, say, in the master 10:25 22 file would have an immunization that was not, in fact, 10:25 23 that child's immunization. 10:25 24 Q. And because of the false match, the child would 10:25 25 then be -- would then not be given that immunization

39 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 19 of :25 1 down the road; it would be -- 10:25 2 MR. BRISSENDEN: Objection, calls for 10:25 3 speculation. 10:25 4 A. Well, as I understand it, the immunizations 10:25 5 were done, but then a record was sent to the Department 10:25 6 of Health to try to match to that child's master 10:25 7 record. And so what that would indicate would be that 10:25 8 the child is recorded as having an immunization, when, 10:25 9 in fact, the child did not have that immunization. 10:25 10 Q. (By Mr. Sells) Okay. And what would be the 10:26 11 real world consequences of a false negative in the 10:26 12 vaccination context? 10:26 13 A. In the immunization context, a false negative 10:26 14 would mean that the child would not be recorded as 10:26 15 having a vaccination that, in fact, the child did have. 10:26 16 Q. Both of which could be serious public health 10:26 17 issues? 10:26 18 A. I suppose. 10:26 19 Q. Tell me a little bit more of a role in that 10:26 20 project. Were you the one constructing the match 10:26 21 algorithm? 10:26 22 A. It's too long for me to say. I don't recall 10:26 23 the details of exactly what I did. 10:27 24 Q. Well, I can understand not recalling the 10:27 25 details. Can you recall any generalities about what

40 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 20 of :27 1 your role was? 10:27 2 A. I'm afraid for that, I would need to go back 10:27 3 and see if I could find reports that I wrote from that 10:27 4 time. I wouldn't want to speculate on what I did or did 10:27 5 not do. 10:27 6 Q. Where would those reports be? 10:27 7 A. If they still exist, they would probably be in 10:27 8 files that I would have in my office at home. 10:27 9 Q. Were those reports that you submitted to the 10:27 10 Texas Department of Public Health? 10:27 11 A. Oh, yes. They might -- they might have them, 10:27 12 although I do not know what their records retention 10:27 13 policy is, so I don't know if they, in fact, still exist 10:28 14 there. 10:28 15 Q. Did you publish anything arising out of that 10:28 16 engagement? 10:28 17 A. I did not. 10:28 18 Q. Have we covered the full spectrum of your 10:28 19 experience with database matching prior to your 10:28 20 involvement in this case? 10:28 21 A. It's difficult for me to say categorically that 10:28 22 that's the case, but let me say that as we sit here, 10:28 23 that's what I recall. 10:28 24 Q. Getting back to your background after that 10:28 25 tangent, aside from being a professor of statistics, do

41 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 21 of :28 1 you have any other outside employment activities? 10:29 2 A. Well, I do consulting, and sometimes that's 10:29 3 considered part of the job, because it gives one 10:29 4 experience in dealing with, presumably, real world 10:29 5 applications of a subject matter. 10:29 6 Q. Anything else? Outside employment? 10:29 7 A. No, I don't think so. 10:29 8 Q. No other business activity? 10:29 9 A. No, not currently. 10:29 10 Q. Okay. You qualified that by not currently. 10:29 11 Has there been in the past been some other outside 10:29 12 business activity? 10:29 13 A. Yes. 10:29 14 Q. What was it? 10:29 15 A. Well, I was involved with -- I have been 10:29 16 involved with a couple of startup companies. That was 10:30 17 back in the 1980s. 10:30 18 Q. What were those startups? 10:30 19 A. Well, they didn't come to fruition, so it's one 10:30 20 of those things where an investor comes in and doesn't 10:30 21 realize the fruits of their involvement. 10:30 22 Q. Gotcha. Do you have any significant volunteer 10:30 23 activities in the Austin area? 10:30 24 A. No. 10:30 25 Q. Past or currently?

42 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 22 of :30 1 A. Well, at one point, I guess I was involved as 10: in an investors group called Investors Education 10:31 3 Association, and I, for a while, arranged speakers for 10:31 4 that. 10:31 5 Q. What did that entail? That sounds interesting. 10:31 6 A. Well, it was a group of investors primarily 10:31 7 motivated by real estate. And again, was this back in 10:31 8 the 1980s. We had guest speakers who would come in and 10:31 9 talk about their investing experiences or different 10:31 10 types of investments that were available. 10:31 11 Q. Hmm. Do you have any experience in politics or 10:31 12 elections? 10:31 13 A. I vote. 10:31 14 Q. You vote. Okay. Are you a member of any 10:31 15 political party? 10:31 16 A. Well, in the last election, the primary that we 10:31 17 had here, I voted as a Republican. 10:31 18 Q. Have you volunteered for the Republican Party 10:32 19 in Texas or the county or the city? 10:32 20 A. Oh, no. 10:32 21 Q. Any volunteering on campaigns? 10:32 22 A. No. 10:32 23 Q. Do you have any other experience with voting, 10:32 24 elections, and politics? 10:32 25 A. Can you clarify what you mean by any other

43 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 23 of :32 1 experience? 10:32 2 Q. Other than the act of voting. You know, we've 10:32 3 covered volunteering for campaigns, volunteering for a 10:32 4 political party. Anything else? 10:32 5 A. Yes. Something just came to mind. From time 10:32 6 to time, I've been a consultant for the City of Austin 10:32 7 dealing with qualifying or not voter-submitted 10:32 8 petitions. This -- just to give a little bit of 10:33 9 background in case you want to know more about -- do you 10:33 10 want to know more about this? 10:33 11 Q. Certainly. 10:33 12 A. Okay. State law, as I understand it, permits 10:33 13 cities to qualify voter petitions by doing a random 10:33 14 sample of 25 percent of submitted signatures to check to 10:33 15 see if they are, in fact, valid signatures. And there 10:33 16 are issues about making an estimate of the number of 10:33 17 valid signatures in the whole population based upon 25 10:33 18 percent. Some rather technical issues arise which make 10:33 19 it less simple than it actually sounds, because of the 10:33 20 possibility of duplicate signatures in one part of the 10:33 21 sample as compared to the sample -- the part of the 10:33 22 population that was not sampled. And so I have done a 10:33 23 number of these over the last four or five years for the 10:33 24 city. 10:34 25 Q. Interesting. Have you ever worked as a poll

44 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 24 of :34 1 worker? 10:34 2 A. I have not. 10:34 3 Q. How about a poll watcher? 10:34 4 A. No. 10:34 5 Q. All right. Would it be fair to say that prior 10:34 6 to your involvement in this case, your experience with 10:34 7 voting, politics, and elections has been fairly minimal? 10:34 8 A. That is probably a fair characterization, 10:34 9 although I should point out that I vote in nearly every 10:34 10 election in which I'm eligible. 10:34 11 Q. That's great. So do I. I think everyone 10:34 12 should. 10:34 13 Let me ask you: What is it about your 10:34 14 background or experience that you think qualifies you as 10:34 15 an expert in this case? 10:34 16 A. For the role that I was asked to assume by the 10:34 17 Office the Attorney General, you have my general 10:35 18 statistics background and you have my data management 10:35 19 background, and you have my experience that we have just 10:35 20 gone over here with data matching in the past. 10:35 21 Q. Well, we didn't go over data management, so I'm 10:35 22 not sure what you mean by that. 10:35 23 A. Data processing. We were talking about -- you 10:35 24 were asking me questions about matching databases, and 10:35 25 also dealing with very large data sets, I have

45 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 25 of :35 1 experience with that. 10:35 2 Q. And tell me about that experience. 10:35 3 A. Well, for example, without going into details 10:35 4 that are covered by confidentiality agreements, there 10:35 5 was one case that I did for a consulting firm that had a 10:35 6 client, which is a bank here in Texas, that was 10:36 7 concerned with generation of fee income, and so I looked 10:36 8 at databases that the bank had, including 20 million or 10:36 9 more records, to process them and try to find out hidden 10:36 10 structure in the data that might lead to more fees for 10:36 11 the bank. 10:36 12 Q. Okay. So when you say data management, you 10:36 13 mean you are experienced in managing large data files of 10:36 14 the type that are involved in this case? 10:36 15 A. Yes, uh-huh. You were asking me if I had any 10:36 16 certifications in data management. My qualifications 10:36 17 are on the job. 10:36 18 Q. Okay. Well, let's turn to your involvement in 10:36 19 this case. That's probably a good segue. 10:36 20 When did you first become involved in this 10:36 21 case? 10:37 22 A. I was first contacted, I think it was, on May 10: nd. I remember that date because that is my 10:37 24 anniversary. 10:37 25 Q. Which year?

46 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 26 of :37 1 A. Oh, this year Sorry. 10:37 2 Q. That's a good way of remembering dates. All 10:37 3 right. 10:37 4 A. It also happened during the time that I was 10:37 5 teaching a short course for the statistics division at 10:37 6 the University of Texas, which meant that my wife and I 10:37 7 could not celebrate our anniversary then, to her 10:37 8 concern. 10:37 9 Q. What is a short course? I'm not sure what the 10:37 10 is. 10:37 11 A. Yeah. In this case, I'm talking about a course 10:37 12 that lasted four days. 10:37 13 Q. Oh, wow. 10:37 14 A. And it was three hours each day. It's an 10:37 15 opportunity for faculty, students, and indeed, members 10:37 16 of the community in general, which might involve state 10:38 17 government employees or, you know, people who are just 10:38 18 interested, to come to the university and take courses 10:38 19 of a limited duration based upon statistics. And so I 10:38 20 taught a course of at least four days dealing with time 10:38 21 series analysis. 10:38 22 Q. Who contacted you on May the 22nd? 10:38 23 A. That was Mr. Sweeten. 10:38 24 Q. Was he taking your course? 10:38 25 A. No.

47 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 27 of :38 1 Q. Do you know who else the State was talking to 10:39 2 to be an expert in this case, or who else the State was 10:39 3 considering to be an expert in this case, or were you 10:39 4 the only person? 10:39 5 MR. BRISSENDEN: Objection. Let me 10:39 6 instruct you that to the extent that requires you to 10:39 7 divulge communications that you have had with 10:39 8 Mr. Sweeten and other counsel, I instruct you not to 10:39 9 disclose those communications. 10:39 10 THE WITNESS: Okay. 10:39 11 Q. (By Mr. Sells) Do you know, from people other 10:39 12 than Mr. Sweeten, whether the State was considering 10:39 13 other experts? 10:39 14 A. I do not know. 10:39 15 Q. Did you execute a retainer agreement with the 10:40 16 State? 10:40 17 A. A contract? 10:40 18 Q. Contract, yes. 10:40 19 A. Yes. 10:40 20 Q. And when was that? 10:40 21 A. I think it was the next day, referring to May 10: rd. It was then or the day after. I think it was the 10:40 23 next day. 10:40 24 Q. So it didn't take you long to decide? 10:40 25 A. No.

48 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 28 of :40 1 Q. Is that contract in writing, I assume? 10:40 2 A. Yes. 10:40 3 Q. Do you have a copy? 10:40 4 A. I don't have a copy with me, no. 10:40 5 Q. Does the contract set forth your scope of work? 10:41 6 A. In general terms, it does. 10:41 7 Q. Is there another document that sets forth your 10:41 8 scope of work more specifically? 10:41 9 A. There is no other document. 10:41 10 Q. Now, do you understand that this portion of the 10:42 11 deposition, or this deposition, I should say, since 10:42 12 we're going to do two depositions -- let me start again. 10:42 13 Do you understand that this is a 10:42 14 deposition of the State of Texas, and that you have been 10:42 15 designated to testify on behalf of the State of Texas 10:42 16 and not in your individual capacity? 10:42 17 A. For this part of the deposition, I understand 10:42 18 that, yes. 10:42 19 Q. Okay. And do you, in fact, consent to testify 10:42 20 on behalf of the State of Texas? 10:42 21 A. Yes. 10:42 22 Q. When did you first learn that you would be 10:42 23 testifying on behalf of the State of Texas? 10:42 24 A. Yesterday. 10:42 25 MR. SELLS: All right. I'm going to

49 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 29 of :42 1 provide the witness with what I'll mark as Government 10:43 2 Exhibit 880. I'm told that's next the exhibit in the 10:43 3 series I have copies for your guys. 10:43 4 (Exhibit 880 marked for identification.) 10:43 5 Q. (By Mr. Sells) Professor Sager, have you seen 10:43 6 this document before? 10:43 7 A. I have seen Exhibit A. I'm not sure I have 10:43 8 seen the preceding body of the document. 10:43 9 Q. Okay. Would you read the title of the document 10:44 10 on the first page? 10:44 11 A. In boldface underline? 10:44 12 Q. Yes. 10:44 13 A. "Attorney General's Notice of Rule 30(b)(6) 10:44 14 deposition of Plaintiff, State of Texas and request for 10:44 15 documents relied upon by 30(b)(6) designees." 10:44 16 Q. And what is Exhibit A to this document? 10:44 17 A. The title of Exhibit A is to Notice of Rule 10: (b)(6) Deposition of Plaintiff, State of Texas, Topics 10:44 19 For Examination. 10:44 20 Q. Okay. So Exhibit A are the topics of 10:44 21 examination for the Rule 30(b)(6) deposition? 10:44 22 A. Yes. 10:44 23 Q. And it is Exhibit A that you have seen before? 10:44 24 A. Yes. 10:44 25 Q. Can you tell me which of the topics that are

50 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 30 of :44 1 listed in Exhibit A you're prepared to testify about? 10:44 2 A. Numbers 10 and :45 3 Q. I should have preceded that question by 10:45 4 asking: When did you see Exhibit A for the first time? 10:45 5 A. Yesterday. 10:45 6 Q. Okay. And you are prepared to talk about 10:45 7 topics 10 and 12? 10:45 8 A. Yes. 10:45 9 Q. Who asked you to talk about those topics? 10:45 10 A. Attorneys for the State of Texas. 10:45 11 Q. And what did you do to prepare yourself to be 10:45 12 able to testify on behalf of the State of Texas about 10:45 13 Topics 10 and 12? 10:45 14 A. I talked with two people who were concerned 10:45 15 with the matters referenced in Topics 10 and 12. They 10:45 16 were -- one is an employee of Legal and Technical 10:45 17 Services at the Office of the Attorney General. His 10:45 18 name the Oscar Valles, V-a-l-l-e-s. The other is Scott 10:46 19 Brandt -- I think that's B-r-a-n-d-t -- of the Secretary 10:46 20 of State. 10:46 21 Q. Did you talk with anyone else? 10:46 22 A. No. 10:46 23 Q. Did you review any documents when preparing for 10:46 24 this deposition? 10:46 25 A. You mean in reference to Topics 10 and 12?

51 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 31 of :46 1 Q. Topics 10 and 12, yes. 10:46 2 A. No, except that there is one document that we 10:46 3 brought with us today, which are notes upon conversation 10:46 4 that I had with Mr. Oscar Valles. I don't suppose that 10:47 5 would actually be a document that I reviewed in 10:47 6 preparation, but it's a document that was generated as a 10:47 7 result of that preparation. 10:47 8 Q. Who generated that document? 10:47 9 A. This is a document that was prepared by 10:47 10 Mr. Valles. 10:47 11 Q. Did he prepare that prior to your conversation? 10:47 12 A. No. 10:47 13 Q. It was subsequent to your conversation? 10:47 14 A. It was during the conversation. 10:47 15 Q. During the conversation. When you spoke with 10:47 16 Mr. Valles and Mr. Brandt, were you speaking with them 10:47 17 in person? 10:47 18 A. Yes. 10:47 19 Q. Did you speak with them at the same time? 10:47 20 A. No. 10:47 21 Q. One at a time? 10:47 22 A. One at a time. 10:47 23 Q. Okay. Are there any other notes of your 10:47 24 conversations with Mr. Valles or Mr. Brandt? 10:47 25 A. No.

52 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 32 of :47 1 Q. Okay. Did you bring me a copy of these notes? 10:48 2 A. I think there are multiple copies in here. 10:48 3 Q. Okay. Hand it over. 10:48 4 MR. SELLS: The witness is handing me a 10:48 5 copy of his notes. And why don't we go ahead and mark 10:48 6 these as an exhibit for the record. 10:48 7 Q. Do you have another copy for the record? 10:48 8 A. Let's see if there are three copies here. 10:48 9 Q. Okay. The next exhibit number would be 881, so 10:48 10 let me mark it. 10:48 11 (Exhibit 881 marked for identification.) 10:49 12 Q. (By Mr. Sells) Okay. And now, for the record, 10:49 13 would you please describe Exhibit 881? 10:49 14 A. 881 is a description of the process that the 10:49 15 Legal and Technical Services of the Office of the 10:49 16 Attorney General used in creating their so-called May no 10:49 17 match data set. 10:49 18 Q. And these are notes of a conversation between 10:49 19 you and Mr. Valles? 10:49 20 A. That is correct. 10:49 21 Q. Okay. Did you bring any other documents with 10:49 22 you? 10:49 23 A. Yes. 10:49 24 Q. What other documents did you bring with you 10:49 25 today?

53 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 33 of :50 1 MR. BRISSENDEN: Are you asking in 10:50 2 connection the 30(b)(6) topics or his expert? 10:50 3 MR. SELLS: Well, right now, I'm just 10:50 4 asking generally what other documents he brought with 10:50 5 him. If they are related to the expert report, we can 10:50 6 put those aside, but I'd like to know the full 10:50 7 inventory. 10:50 8 A. Okay. I have copies here of my first 10:50 9 declaration, which I think was dated June 1st. Let me 10:50 10 check. Yes. June 1st declaration. I also have copies 10:50 11 of my supplemental declaration, which I think is dated 10:50 12 June 8. No, sorry, June 11. I have a copy of 10:51 13 Dr. Stephen Ansolabehere's. I think this is his first 10:51 14 declaration dated June 1, And a copy of 10:51 15 Dr. Stephen Ansolabehere -- 10:51 16 Do I need to spell that for you? 10:51 17 THE REPORTER: Yeah, let me just look at 10:51 18 it real quick. 10:51 19 Q. (By Mr. Sells) Okay. 10:51 20 A. Okay. So we can proceed then? 10:51 21 Q. We can proceed. Let's put those to the side. 10:51 22 Those are all the documents -- 10:51 23 A. There are more. 10:51 24 Q. Oh, there are more. All right. 10:51 25 A. So we have the two declarations of

54 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 34 of :51 1 Dr. Ansolabehere. And let's see, I have also one page 10:51 2 of errata to make in my supplemental declaration. And I 10:52 3 brought with me also some copies of additional analysis 10:52 4 that was generated yesterday. 10:52 5 Q. Okay. And what does the additional analysis 10:52 6 relate to? 10:52 7 A. It relates to the second part of the deposition 10:52 8 for today. 10:52 9 Q. Okay. So let's put those documents to the 10:52 10 side. I appreciate you giving me that inventory. 10:52 11 Okay. Now I'd just like to ask you to 10:52 12 reread Topics 10 and 12 so that they're in the front of 10:52 13 mind. 10:52 14 A. Read them aloud? 10:52 15 Q. No. You can read them to yourself. 10:53 16 A. Okay. (Witness reviewing document.) Okay. 10:53 17 Q. Okay. I want to begin by asking you about the 10:53 18 production part of Topics 10 and 12, the production of 10:53 19 the data to the experts. 10:53 20 First of all, can you tell me what data 10:53 21 has been produced to Professor Shaw? 10:53 22 A. I do not know of data produced to Professor 10:54 23 Shaw. What data was provided, I do not know. 10:54 24 Q. Did you do anything to find out what data was 10:54 25 produced to Professor Shaw?

55 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 35 of :54 1 A. I did not. 10:54 2 Q. What would you have needed to do in order to be 10:54 3 able to answer that question? 10:54 4 A. I presume that I would need to ask Mr. Valles 10:54 5 or Mr. Brandt if they had knowledge of data that was 10:55 6 provided to Professor Shaw. 10:55 7 Q. But you didn't do that? 10:55 8 A. I did not. 10:55 9 Q. Well, if I ask a series of follow-up questions, 10:55 10 I suspect your answer is going to be "I don't know," so, 10:55 11 but let's just try it for the record. 10:55 12 Who gave the data to Professor Shaw? 10:55 13 A. I do not know. 10:55 14 Q. Do you know when data was given to Professor 10:55 15 Shaw? 10:55 16 A. I do not know. 10:55 17 Q. Do you know in what format data was given to 10:55 18 Professor Shaw? 10:55 19 A. I do not know. 10:56 20 Q. Do you know what fields were included in the 10:56 21 data that were given to Professor Shaw? 10:56 22 A. No. 10:56 23 Q. Do you know who specified the fields that were 10:56 24 given to Professor Shaw? 10:56 25 A. I do not know.

56 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 36 of :56 1 Q. And do you know whether the data that was 10:56 2 produced to Professor Shaw included the code used to 10:56 3 generate that data? 10:56 4 A. I do not know. 10:56 5 Q. Well, now I'm going to ask you some easier 10:56 6 questions. What data has been produced to you? 10:56 7 A. For me, I received what we could call the 10:56 8 December no match data set and the May no match data 10:56 9 set. I have also received copies of the driver's 10:56 10 license data, and the license to carry data, and the 10:57 11 voter registration data. 10:57 12 Q. Any other data that has been produced to you? 10:57 13 A. Yes. I'm not sure that they are related to 10:57 14 this part of the deposition, but I received some random 10:57 15 samples that are described in the appendices of my -- or 10:57 16 the exhibits in my report, the second -- the 10:57 17 supplemental, and I have received Dr. Ansolabehere's BRN 10:57 18 ID data. 10:57 19 Q. Who gave you all this data? 10:57 20 A. They were delivered to me by police escort, 10:58 21 because of the sensitive nature of the social security 10:58 22 information that is included within -- on diskettes, and 10:58 23 they came from the attorneys, I understand, ultimately 10:58 24 from the Secretary of State and the DPS. 10:58 25 Q. I were given December and May no match data

57 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 37 of :58 1 sets at the same time, or did one come before the other? 10:58 2 A. I think one came before the other. 10:58 3 Q. Do you remember when you got those data sets? 10:58 4 A. It was very soon after the contract was 10:58 5 signed. I don't remember what date. 10:58 6 Q. And would that have been the December no match 10:59 7 data set? 10:59 8 A. Let's see. You mean the one that came first? 10:59 9 Q. The first one, yes. 10:59 10 A. I, frankly, don't remember which one was first. 10:59 11 Q. When did you receive the driver's license 10:59 12 database? 10:59 13 A. I don't remember exactly which batch it came 10:59 14 in. Police officers came to my office and to my home, 10:59 15 came to my office once and to my home, oh, about three 10:59 16 times, I think. 11:00 17 Q. Well, I'm going to ask the same question with 11:00 18 regard to the license to carry and voter registration 11:00 19 databases. When did you get those? 11:00 20 A. Oh, the same answer. 11:00 21 Q. Who would know? 11:00 22 A. Since I had to sign for them each time I 11:00 23 received them, I presume that whoever has that signed 11:00 24 receipt would know. 11:00 25 Q. Did the driver's license, license to carry, and

58 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 38 of :00 1 voter registration data come at the same time as the no 11:00 2 match data sets, or did they come before or after? 11:00 3 A. Some of them came together, but I'm not sure of 11:01 4 the order. 11:01 5 Q. When did you get the random samples? 11:01 6 A. Oh, let me amend my answer, because I'm 11:01 7 recalling. I first analyzed the data set that had the 11: ,000 odd ones, which I think is the December no match 11:01 9 data set. 11:01 10 Q. Okay. Did you have the driver's license, 11:01 11 license to carry, and voter registration databases at 11:02 12 that time? 11:02 13 A. I believe that I did. I'm not 100 percent 11:02 14 certain, but that is what I recall at the moment. 11:02 15 Q. In what format did you get the data? 11:02 16 A. I think that all of the files were encrypted in 11:02 17 some manner, again, because of the sensitive social 11:02 18 security information that's included, as well as other 11:02 19 personal information, and the ultimate files that were 11:02 20 encrypted I believe were in text format. 11:03 21 Q. What fields were included in the data that were 11:03 22 produced to you? 11:03 23 A. Oh, I don't remember all of them. There were 11:03 24 quite a few? 11:03 25 Q. Who specified the fields to be included in the

59 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 39 of :03 1 data produced to you? 11:03 2 A. I do not know. 11:03 3 Q. Did any of the data produced to you include the 11:03 4 code used to generate it? 11:03 5 A. No. 11:03 6 Q. Do you remember getting the code along with 11:03 7 Professor Ansolabehere's data? 11:03 8 A. No. You are talking about Professor 11:03 9 Ansolabehere's code? 11:03 10 Q. Yes. 11:03 11 A. No. 11:04 12 Q. Did the State produce any data to any other 11:04 13 experts besides you and Professor Shaw? 11:04 14 A. I do not know. 11:04 15 Q. Did the drivers license database come together 11:05 16 with the license to carry database from the Department 11:05 17 of Public Safety, or were those two separate databases? 11:05 18 A. They were not in the same files. They were in 11:05 19 different files. 11:05 20 MR. SELLS: We've been going for about an 11:05 21 hour. Why don't we take a little bathroom break. 11:05 22 MR. BRISSENDEN: Sure. 11:05 23 (Recess from at 11:05 a.m. to 11:18 a.m.) 11:18 24 MR. SELLS: Back on the record. 11:18 25 Q. (By Mr. Sells) Professor Sager, I'd like to

60 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 40 of :18 1 turn now to another portion of Topics 10 and 12. This 11:18 2 one is actually part of Topic 10, and that is the 11:18 3 process, manner, and method of compiling data from the 11:18 4 Department of Public Safety. What were the basic steps 11:18 5 involved in compiling the data from the Department of 11:18 6 Public Safety? 11:18 7 A. This applies to both the May and December match 11:19 8 cases. As Mr. Valles and Mr. Brandt indicated, they had 11:19 9 had the data from the Department of Public Safety, and 11:19 10 then they performed matches from the voter -- voter 11:19 11 registration data, which they also had from the 11:19 12 Department of Public Safety. 11:19 13 Q. Let me make sure I understand you correctly. 11:19 14 Were Mr. Valles and Mr. Brandt working together? 11:19 15 A. No. 11:19 16 Q. They were working independently? 11:19 17 A. That's my understanding. 11:20 18 Q. What do you mean when you say they had the data 11:20 19 from the Department of Public Safety? 11:20 20 A. If you're asking me about the chain of custody 11:20 21 by which the data was transferred from the Department of 11:20 22 Public Safety to the Secretary of State to Legal 11:20 23 Technical Services, I do not know that. 11:20 24 Q. I'm not asking about the chain of custody, 11:20 25 Professor Sager, but I am asking about Topic 10. Let's

61 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 41 of :20 1 focus in on it, specifically the first clause. Do you 11:20 2 have it in front of you? 11:20 3 A. Yes, I do. 11:20 4 Q. Okay. It says the process, manner, and method 11:20 5 of compiling, cleaning, and filtering data from the 11:20 6 Department of Public Safety's driver license and license 11:20 7 to carry databases. 11:21 8 A. Yes. 11:21 9 Q. So are you prepared today to tell me about the 11:21 10 process, manner, and method of compiling, cleaning and 11:21 11 filtering data from the Department of Public Safety's 11:21 12 driver license and license to carry databases? 11:21 13 A. Yes. 11:21 14 MR. BRISSENDEN: In addition, let me just 11:21 15 be clear on the record. Two days ago, the State of 11:21 16 Texas presented two employees from the Department of 11:21 17 Public Safety who specifically addressed that topic in 11:21 18 terms as reflected as well in Topic 9, which describes 11:21 19 the process, manner, and method of compiling, cleaning 11:21 20 and filtering data from the Department of Public 11:21 21 Safety's driver license and license to carry 11:21 22 databases. So that may, in terms of your questioning 11:21 23 and helping things move along, that may help you here 11:21 24 today. 11:21 25 MR. SELLS: Who were those people that

62 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 42 of :22 1 you're referring to? 11:22 2 MR. BRISSENDEN: Dennis Leopold addressed 11:22 3 the license to carry database. Mark Whelan discussed 11:22 4 the data from the driver license database. 11:22 5 MR. SELLS: Is that the W-H-A-L-E-N? 11:22 6 MR. SWEETEN: W-H-E-L-A-N. 11:22 7 Q. (By Mr. Sells) Well, Professor Sager, you 11:22 8 answered in the affirmative my question whether you're 11:22 9 prepared to talk about that as well. 11:23 10 A. Yes. 11:23 11 Q. Okay. Tell me what you are prepared to talk 11:23 12 about. 11:23 13 A. Okay. In my conversations with Mr. Valles and 11:23 14 Mr. Brandt, both indicated that after their respective 11:23 15 agencies received the data from the Department of Public 11:23 16 Safety, that they engaged in no further processing of 11:23 17 the data prior to the matching that they did between it 11:23 18 and the -- well, between the voter registration database 11:23 19 and the Department of Public Safety database. So there 11:23 20 was no further compiling, no further cleaning, and no 11:23 21 further filtering of that data. 11:23 22 Q. Okay. But you're not prepared to talk about 11:23 23 any compiling, cleaning, or filtering that was done 11:23 24 prior to the production of the database from the 11:23 25 Department of Public Safety to the Office of the

63 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 43 of :24 1 Attorney General and/or Secretary of State? 11:24 2 A. That's correct. 11:24 3 Q. Do you have any knowledge, as you sit here 11:24 4 today, about the compiling, cleaning, or filtering of 11:24 5 data that occurred before it got to the Office of the 11:24 6 Attorney General or Secretary of State? 11:24 7 A. I do not. 11:24 8 Q. So is it fair then to say the extent of your 11:25 9 preparation and ability to testify here today begins at 11:25 10 when it was -- what data was received by the Office of 11:25 11 the Attorney General and/or Secretary of State? 11:25 12 A. That is my reading of the intent of the first 11:25 13 clause in Number 10. Because it refers to the 11:25 14 compiling, cleaning, and filtering data from the 11:25 15 Department of Public Safety, which implies to me that 11:25 16 the scope of number 10 starts after it has been received 11:25 17 by the Secretary of State and the Legal Technical 11:25 18 Services. 11:25 19 Q. Okay. Who was the person or persons who 11:25 20 compiled the data once it was received by the Office of 11:26 21 Attorney General and the Secretary of State? 11:26 22 A. Well my -- 11:26 23 MR. BRISSENDEN: Objection, compound. 11:26 24 You may answer. 11:26 25 A. My understanding is that there -- for

64 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 44 of :26 1 Mr. Valles and Mr. Brandt, is there was no further 11:26 2 compiling, cleaning, and filtering after it had been 11:26 3 received from the Department of Public Safety. 11:26 4 Q. (By Mr. Sells) Well, let me ask it this way: 11:26 5 Who received the data at the Department of Public 11:26 6 Safety? 11:26 7 A. I don't know the name of the person that 11:26 8 received it. I presumed it was transferred by -- well, 11:26 9 it would be speculation whether it was transferred under 11:26 10 police escort or not. I don't know. 11:26 11 Q. What database -- 11:26 12 A. That goes to my response -- I thought you were 11:26 13 asking about the chain of custody from between the 11:26 14 respective agencies. 11:26 15 Q. What database was the starting point? Let me 11:27 16 ask that a different way. What database from the 11:27 17 Department of Public Safety was the starting point for 11:27 18 whatever occurred in the match process? 11:27 19 A. It was a set of driver license records. My 11:27 20 understanding is that the DPS also is custodian of the 11:27 21 license to carry database. 11:27 22 Q. What was the date of the data extraction? 11:27 23 A. I don't know. There are file dates that appear 11:27 24 on the files that I received, but they could be the 11:27 25 dates that they were put on the media. I'm not sure

65 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 45 of :27 1 about that. 11:27 2 Q. Are there any other databases from the 11:27 3 Department of Public Safety that form the basis of this 11:28 4 match that was ultimately produced to you and to 11:28 5 Professor Shaw? 11:28 6 A. The driver license and the license to carry are 11:28 7 the ones that I know about. 11:28 8 Q. Have you ever heard the term, the annual jury 11:28 9 wheel process? 11:28 10 A. Say again, please? 11:28 11 Q. The annual jury wheel process. 11:28 12 A. Wheel, w-h-e-e-l? 11:28 13 Q. Correct. 11:28 14 A. No, I have not. 11:28 15 Q. So you wouldn't know whether the data from the 11:29 16 Department of Public Safety that went into this match 11:29 17 was from the annual jury wheel process? 11:29 18 A. I do not know. 11:29 19 Q. Do you know what tables were part of the 11:29 20 production from the Department of Public Safety to the 11:29 21 Office of the Attorney General or Secretary of State? 11:29 22 MR. BRISSENDEN: Objection, vague and 11:29 23 compound. 11:29 24 A. Is the question: Do I know the names of the 11:29 25 tables?

66 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 46 of :29 1 Q. (By Mr. Sells) Well, if you know them by some 11:29 2 other designation, I'd be satisfied with that. 11:29 3 A. May I refer to Exhibit 881? 11:30 4 Q. You may. These are your notes? 11:30 5 A. Yes. In point number 2, Mr. Valles indicates 11:30 6 that the set of matching criteria listed there were 11:30 7 executed between the DPS table "Person" and the SOS 11:30 8 table "Voters". 11:30 9 Q. What fields within each table were involved in 11:30 10 the match process? 11:30 11 A. There were three: First name, last name, and 11:30 12 date of birth. 11:30 13 Q. And you're reading that from Exhibit 881? 11:30 14 A. That is correct. 11:31 15 Could I ask you to restate your previous 11:31 16 question or perhaps have the court reporter reread it? 11:31 17 Because I think I need to amplify the answer in order to 11:31 18 give the whole truth. 11:31 19 Q. I don't have the question in mind; it's already 11:31 20 escaped me. 11:31 21 So can you help me out, Court Reporter? 11:31 22 (Requested portion read back by the 11:31 23 reporter.) 11:31 24 A. To amplify a little bit, I call your attention 11:32 25 to point 3 on 881 there. Within the voter registration

67 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 47 of :32 1 data field called Official ID Number, which was used in 11:32 2 subsequently winnowing the data from the 1,296,805 11:32 3 records that were not matched down to the lower 588,095 11:32 4 is for the May no match, and a similar process was 11:32 5 applied by the Secretary of State to get to its December 11:32 6 no match data. 11:32 7 Q. Well, let's -- 11:32 8 A. But that field is from the voter registration 11:32 9 data and you asked the question about the DPS data. I 11:32 10 just wanted to amplify that to make sure that we -- I 11:32 11 give a complete and full answer. 11:32 12 Q. Okay. Thank you. We'll come back to that when 11:32 13 we talk about the voter registration side, the Secretary 11:33 14 of State side. 11:33 15 Was there any searching for duplicates? 11:33 16 MR. BRISSENDEN: Objection, vague. Are 11:33 17 you -- is there a time frame? Or what are you asking 11:33 18 him about? 11:33 19 MR. SELLS: We're talking about Topic :33 20 We're talking about the Department of Public Safety's 11:33 21 database. We're talking about, since Professor Sager is 11:33 22 unable to talk about what happened before it got to the 11:33 23 Secretary of State's Office or the Office of the 11:33 24 Attorney General, we're talking about there. Keeping in 11:33 25 mind Topic 10, was there any searching of duplicates at

68 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 48 of :33 1 that point? 11:33 2 MR. BRISSENDEN: And which matching are 11:33 3 you referring to when ask you that question? 11:33 4 MR. SELLS: I guess if I don't specify, I 11:33 5 mean either one, the December or May. 11:33 6 MR. BRISSENDEN: Well, those are two 11:33 7 different matches, so it may help to clarify whenever 11:33 8 you refer to -- 11:34 9 MR. SELLS: Okay. 11:34 10 MR. BRISSENDEN: -- the process of which 11:34 11 match you're -- match process you're asking about. 11:34 12 MR. SELLS: Okay. 11:34 13 Q. (By Mr. Sells) Well, let's start with 11:34 14 December. Was there any searching for duplicates in the 11:34 15 December match process? 11:34 16 A. Filtering for duplicates in doing the matching 11:34 17 was not done in either the December or the May matches 11:34 18 at that stage. 11:34 19 Q. What was different about the December and May 11:34 20 matches that produced the different results? 11:34 21 A. The same procedure was applied in that matches 11:34 22 were conducted on first name, last name, date of birth 11:34 23 and a subsequent winnowing of the data set by official 11:34 24 ID number in the voter registration list. Those 11:34 25 procedures were the same in the May and the December no

69 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 49 of :34 1 match generation. 11:34 2 Q. Okay. You answered my -- you answered the 11:34 3 opposite of my question really. I wanted to know what's 11:35 4 different. You told me what's the same. 11:35 5 A. Oh. 11:35 6 Q. What was different? 11:35 7 A. The procedure was the same. There was no 11:35 8 difference in the procedure. Of course they're dealing 11:35 9 with different data sets, and so the records that 11:35 10 survive that winnowing process will be different. 11:35 11 Q. Is presumably the data sets were extracted at 11:35 12 different times? 11:35 13 A. I presume. 11:35 14 Q. Or some processing was undertaken prior to the 11:35 15 data's delivery to the Secretary of State or the 11:35 16 Attorney General's Office? 11:35 17 A. I have no knowledge of the latter. 11:35 18 Q. Do you have any knowledge about there being a 11:35 19 different extraction date? 11:35 20 A. I was informed that the December no match data 11:35 21 was generated from data that I believe were produced in 11:36 22 November. And the May no match was generated from data 11:36 23 that were generated either, let's see, the end of April, 11:36 24 I believe, or early May. 11:36 25 Q. But you don't know whether there was any

70 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 50 of :36 1 different processing prior the delivery of that data? 11:36 2 A. Prior to delivery, I do not know. 11:36 3 Q. So there may have been, there may not have 11:36 4 been, you have no idea? 11:36 5 A. I have no knowledge. 11:36 6 Q. So in the December and May matches, there was 11:38 7 no filtering for expired licenses? 11:38 8 A. No, that's not part of the matching process. 11:38 9 Q. But fields were available that could have 11:38 10 allowed that filtering to take place, correct? 11:38 11 A. Yes. 11:38 12 Q. There were was no filtering for suspended 11:38 13 licenses, was there? 11:38 14 A. Correct. 11:38 15 Q. And fields were available such that that 11:38 16 filtering could have taken place, right? 11:38 17 A. Yes. 11:38 18 Q. There was no filtering for deceased records or 11:38 19 records marked as deceased in the driver license 11:38 20 database, was there? 11:38 21 A. That is correct. 11:38 22 Q. But fields were available so that deceased 11:39 23 records could have been filtered out? 11:39 24 A. Yes. 11:39 25 Q. There was no filtering for noncitizens in the

71 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 51 of :39 1 license to carry database, was there? 11:39 2 A. Correct. 11:39 3 Q. But in fact there was a field available -- or 11:39 4 fields available so that that filtering could have taken 11:39 5 place? 11:39 6 A. Yes. 11:39 7 Q. Were there any checks or filtering run for data 11:39 8 entry errors in social security numbers prior to the 11:39 9 match? 11:39 10 A. No. 11:39 11 Q. Any other filtering at all performed prior to 11:39 12 the match? 11:39 13 A. There was none. 11:40 14 Q. Are you aware of the existence of any code of 11:40 15 any kind that was used to generate the Department of 11:40 16 Public Safety databases used in either of the December 11:40 17 or May matches? 11:41 18 A. That would have been prior to the delivery of 11:41 19 those databases to the respective agencies, and I'm not 11:41 20 aware of any code. 11:41 21 Q. Okay. I want to have a similar conversation 11:41 22 but shifting now to Topic 12, which is about the 11:41 23 Secretary of State topics, the voter registration 11:41 24 database, and I'll give you a moment to review Topic 12 11:41 25 just so you have that in mind.

72 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 52 of :41 1 A. Thank you. 11:41 2 MR. BRISSENDEN: Bryan, while he's 11:42 3 reviewing that just because I think we may run into a 11:42 4 similar matter with Topic 12 that we did with Topic 10, 11:42 5 and that is that in terms of the process, manner, and 11:42 6 method of compiling, cleaning and filtering data from 11:42 7 the T.E.A.M. voter registration database, that is also 11:42 8 reflected in Topic 11, and two days ago the State of 11:42 9 Texas presented Scott Brandt from the Secretary of 11:42 10 State's office and he testified as to that process on 11:42 11 that topic. 11:42 12 MR. SELLS: Thank you. 11:42 13 Q. (By Mr. Sells) Professor Sager, do you have 11:42 14 Topic 12 in mind? 11:42 15 A. Yes, I do. 11:42 16 Q. Okay. Is it in fact the case that you have no 11:43 17 knowledge of anything that happened with the T.E.A.M 11:43 18 database prior to its delivery to the Office of the 11:43 19 Attorney General for the match process? 11:43 20 A. For the May match process, that is correct. 11:43 21 Q. Okay. But do you have information regarding 11:43 22 the December match process? 11:43 23 A. Well, no. But the Secretary of State was the 11:43 24 one that did the December no match process, and so I 11:43 25 would presume that, that the information about what

73 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 53 of :43 1 processing they may or may not have done prior to their 11:43 2 actual matching would be with them. I do not have 11:43 3 knowledge of that. 11:43 4 Q. So the Secretary of State's office did the 11:43 5 December match and the office the Attorney General did 11:43 6 the May match? 11:43 7 A. That's correct. 11:44 8 Q. And you spoke with Scott Brandt in preparation 11:44 9 for this deposition here today, correct? 11:44 10 A. Correct. 11:44 11 Q. And he was involved with the December match in 11:44 12 the Secretary of State's office? 11:44 13 A. It my understanding that he did it or his 11:44 14 supervisor was responsible for it. 11:44 15 Q. And can you tell me, based on your 11:44 16 conversations with him, whether there was any filtering, 11:44 17 processing, compiling done prior to the match process at 11:44 18 the Secretary of State's office? 11:44 19 A. He indicated there was no compiling, cleaning, 11:44 20 or filtering these data. And by these data, I mean the 11:44 21 voter registration data. 11:45 22 Q. Based on your conversations with Brandt and 11:45 23 Valles and your knowledge of the differences of the no 11:45 24 match data sets, do you have any theories or guesses as 11:45 25 what went on between the two no match data sets?

74 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 54 of :45 1 MR. BRISSENDEN: Objection, vague, calls 11:45 2 for speculation. To the extent you know -- 11:45 3 A. I do not speculate about that. 11:45 4 Q. (By Mr. Sells) Well, it had to have been 11:45 5 something, right? 11:45 6 A. Yes. 11:45 7 Q. And you don't have, as you sit here today, any 11:45 8 idea what that is? 11:45 9 A. Well, it could be a difference in the databases 11:45 10 because of the lapse of time between their generation. 11:45 11 I think one was in November and one was in late April or 11:45 12 early May. So that might account for part of it. But 11:45 13 you asked me to speculate, so that is one speculation. 11:45 14 Q. What is some other speculation? 11:46 15 MR. BRISSENDEN: As to the extent you know 11: objection, calls for speculation. But the extent you 11:46 17 know. 11:46 18 A. I really don't know. 11:46 19 Q. (By Mr. Sells) Do you think that the lapse of 11:46 20 time is large enough to create a difference of the 11:46 21 magnitude observed here in the no match data set? 11:46 22 MR. BRISSENDEN: Same objection. 11:46 23 A. I do not know. Well, since you asked me to 11:46 24 speculate and purely speculation: Could be that there 11:46 25 was some quality control program with the agencies that

75 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 55 of :46 1 led to programs with more accurate records. But that is 11:47 2 purely speculation. I don't know. 11:47 3 Q. (By Mr. Sells) What do you mean by quality 11:47 4 control? 11:47 5 A. Well, most organizations that maintain big 11:47 6 databases, you would expect that they would be concerned 11:47 7 with the accuracy of the data, so they may have some 11:47 8 internal methods for correcting problems that might 11:47 9 arise. I don't really know what they are, but you just 11:47 10 asked me to speculate, so that's one other hypothetical 11:47 11 reason why they might differ. 11:47 12 Q. What's an example of what quality control 11:47 13 measure might be in place here or might have been 11:47 14 applied to the May database? 11:47 15 MR. BRISSENDEN: Objection, calls for 11:47 16 speculation. 11:48 17 A. Well, you're asking me to continue in a 11:48 18 hypothetical mode, so it could be that before updates in 11:48 19 records are processed against a master database, that 11:48 20 they're held in some repository prior to getting enough 11:48 21 of them to apply to doing the update. That's a 11:48 22 possibility. 11:48 23 Q. (By Mr. Sells) Could it also be, for example, 11:48 24 that there was some effort to eliminate duplicate 11:48 25 entries from the May database?

76 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 56 of :48 1 MR. BRISSENDEN: Objection, calls for 11:48 2 speculation. The witness testified he's not aware. 11:48 3 A. No, I really don't know. 11:48 4 Q. (By Mr. Sells) Would deduping be an example of 11:48 5 a quality control? 11:48 6 MR. SELLS: Same objection. 11:48 7 A. Same speculation. 11:48 8 Q. (By Mr. Sells) I'm not asking you to 11:48 9 speculate. In general database management, you having 11:48 10 that experience, would the elimination of duplicate 11:49 11 entries be something you would characterize as quality 11:49 12 control? 11:49 13 A. Perhaps. 11:49 14 MR. SELLS: Off the record. 11:49 15 (Recess from 11:49 to 11:58 a.m.) 11:57 16 MR. SELLS: Can you read back the 11:58 17 question, please? 11:58 18 (Requested portion was read back by the 11:58 19 court reporter.) 11:58 20 Q. (By Mr. Sells) Back on the record after a break 11:59 21 for technical reasons. 11:59 22 So Professor Sager, I want to make sure I 11:59 23 have all of this straight. As far as the extent of your 11:59 24 knowledge and who you talked with and what processing 11:59 25 may have been done, you spoke with Scott Brandt in the

77 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 57 of :59 1 Secretary of State's Office in preparation for this 11:59 2 deposition here today? 11:59 3 A. Correct. 11:59 4 Q. Okay. Scott Brandt participated in or 11:59 5 supervised the December match, correct? 11:59 6 A. Yes. 11:59 7 Q. And Scott Brandt told you that there was no 11:59 8 cleaning of the database done prior to the match in 12:00 9 December? 12:00 10 A. For the purpose of the match, yes. 12:00 11 Q. But in the May time frame, the Office of the 12:00 12 Attorney General did the match? 12:00 13 A. Correct. 12:00 14 Q. And that would have been participated in or 12:00 15 supervised by Mr. Valles? 12:00 16 A. Correct. 12:00 17 Q. And so Mr. Valles doesn't know whether there 12:00 18 was any processing done prior to the data's delivery to 12:00 19 him? 12:00 20 A. I do not know what Mr. Valles knows in that 12:00 21 regard. 12:00 22 Q. Well, he certainly didn't share any of that 12:00 23 with you, did he? 12:00 24 A. No, he did not. 12:00 25 Q. So as a result, you have no knowledge of any

78 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 58 of :00 1 processing that might have taken place before delivery 12:00 2 of the TEAM database to the Office of the Attorney 12:00 3 General for the May match? 12:01 4 A. Correct. 12:01 5 Q. But you could have asked Scott Brandt about any 12:01 6 processing that was done to the TEAM database for the 12:01 7 May match, correct? 12:01 8 A. Indeed, I did ask him. 12:01 9 Q. And what was his answer? 12:01 10 A. He said that there was no processing of the 12:01 11 data prior to the match for the purpose of the match. 12:01 12 Q. Is that true also of May? 12:01 13 A. Well, Mr. Brandt did not supervise the May 12:01 14 match. 12:01 15 Q. Okay. Did Mr. Brandt participate in the 12:01 16 production of that data to the Office of the Attorney 12:01 17 General and Mr. Valles? 12:01 18 A. I do not know. 12:02 19 Q. So you didn't ask Mr. Brandt about that? 12:02 20 A. About that? 12:02 21 Q. About any processing that was done on the TEAM 12:02 22 database file prior to the match in May? 12:02 23 A. I did not. 12:02 24 Q. And he didn't volunteer any information on 12:02 25 that?

79 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 59 of :02 1 A. No. As I say, the way that I read both Number 12: and Number 12, was that it was referring to what 12:02 3 happened after delivery of the data, from as both 12:02 4 statements read, from the respective agencies. 12:03 5 Q. I asked you a series of questions about 12:03 6 possible filtering that could have been on the 12:03 7 Department of Public Safety databases, either for the 12:03 8 May match or the December match, and I want to ask 12:04 9 similar questions. Could the TEAM database have been 12:04 10 filtered for deceased records prior to the match? 12:04 11 A. I'm not sure of the word "could" there. Could 12:04 12 it have been? You mean could those records have been 12:04 13 extracted? 12:04 14 Q. Are there fields available to do that kind of 12:04 15 filtering? 12:04 16 A. Oh, I see. Yes. Excuse me. You're asking a 12:04 17 question about the Secretary of State's data? 12:04 18 Q. Correct. 12:04 19 A. Does it have a field for deceased voters? Is 12:04 20 that your question? 12:04 21 Q. Yes. 12:04 22 A. I do not know that it does. 12:04 23 Q. Okay. I want to move forward and talk now -- 12:05 24 see if we can get this topic in before lunch -- about 12:05 25 the actual match process. And I realize there are two

80 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 60 of :05 1 matches, so I will do my best to specify. 12:05 2 Let's talk first about the December 12:05 3 match. Who prepared the no match list? 12:05 4 A. The December no match list was prepared by the 12:05 5 Secretary of State. 12:05 6 Q. Who within the Secretary of State's Office 12:05 7 prepared it? 12:05 8 A. My understanding is that Mr. Brandt either did 12:05 9 that himself or supervised or was responsible for its 12:05 10 production. 12:05 11 Q. Who prepared the May no match list? 12:05 12 A. The May no match list was prepared by the legal 12:05 13 and technical services of the Office of the Attorney 12:05 14 General with Mr. Oscar Valles doing that process. 12:06 15 Q. It may seem like a silly question, but when was 12:06 16 the December no match done? 12:06 17 A. Well, I presume in that sense, it's called 12:06 18 December, it was done in December. 12:06 19 Q. Do you know when in December it was done? 12:06 20 A. The date, I do not. 12:06 21 Q. The same question with regard to the May no 12:06 22 match. Do you know when the May no match list was 12:06 23 prepared? 12:06 24 A. I think that may have been after my engagement 12:06 25 with the case, but I don't recall -- I don't know what

81 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 61 of :06 1 day it was. 12:06 2 Q. Did Mr. Valles consult with you during that 12:07 3 process? 12:07 4 A. I had a conversation with Mr. Valles at the 12:07 5 same time that one of the State's attorneys was present. 12:07 6 MR. BRISSENDEN: And in terms of the 12:07 7 communications that you had with counsel present, then I 12:07 8 would instruct you not to disclose communications -- 12:07 9 THE WITNESS: Okay. 12:07 10 MR. BRISSENDEN: -- that you had with 12:07 11 counsel. 12:07 12 Q. (By Mr. Sells) And that discussion was during 12:07 13 the no match process -- or I should say prior to the May 12:07 14 no match list production? 12:07 15 A. I do not recall. 12:07 16 Q. What was the subject of that consultation with 12:07 17 Mr. Valles? 12:07 18 A. I have been instructed not to answer. 12:08 19 Q. Your conversation with Mr. Valles occurred in 12:08 20 May, though? 12:08 21 A. Yes. 12:08 22 Q. How do you know that it was in May? 12:08 23 A. Because my engagement in the case began on or 12:08 24 about May 22nd. 12:08 25 Q. And you remember that it was not in June?

82 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 62 of :08 1 A. There was a subsequent conversation with 12:08 2 Mr. Valles in June, but again, counsel was present. 12:08 3 Q. And during that conversation with Mr. Valles -- 12:08 4 first of all, is Mr. Valles an attorney? 12:08 5 A. Not to my knowledge. 12:08 6 Q. Okay. And did Mr. Valles discuss with you how 12:09 7 to set up the no match list in May? 12:09 8 A. I have been instructed not to answer. 12:09 9 Q. So walk me through the matching process in 12:10 10 December. 12:10 11 A. The matching process in December is actually 12:10 12 very simple, and it's outlined on Exhibit 881, and also 12:10 13 is described in Paragraph 5 of any expert declaration. 12:10 14 So turning now to Exhibit 881, there are 12:10 15 three steps that are outlined there. The first step is 12:10 16 inconsequential. It is simply adding a record number to 12:10 17 each record in the SOS voter's database, which would be 12:10 18 to number the records one, two, three, four, five, six 12:10 19 to however many there are, to give a unique record 12:10 20 number. But the substance of the match is described in 12:10 21 Points 2 and 3. 12:10 22 The voter registration list was matched to 12:11 23 the DPS data on three criteria: First name, last name 12:11 24 and date of birth with exact matches being required to 12:11 25 declare that the records had been matched. This

83 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 63 of :11 1 resulted in reducing the number of SOS voter 12:11 2 registration records to 1,296,805, from the over 13 12:11 3 million that began in the voter registration list. 12:11 4 So then a second step was applied, and 12:11 5 that's described in Point 3 right here, was to further 12:11 6 reduce this nearly 1.3 million by eliminating those 12:11 7 records with an official ID number. That's a field in 12:11 8 the voter registration database that is populated with 12:11 9 some value other than null. And that means the nearly 12: million that remained after pass, based on first 12:12 11 name plus last name plus date of birth was reduced 12: , :12 13 Q. And what does the official ID number field 12:12 14 represent? 12:12 15 A. Excuse me. You started asking about the 12:12 16 December match? 12:12 17 Q. Yes. 12:12 18 A. And beg your pardon. I was describing the May 12:12 19 match, the same procedures were involved in the December 12:12 20 no match, which, just to be complete on the record since 12:12 21 I'm supposed to, and you had asked me also subsequently 12:12 22 about the May match, to return, then, doing them out of 12:12 23 order, to the December no match, I do not know if step 12:12 24 number 1 was applied, but step number 2 and 3 were 12:12 25 applied by the processing done by the Secretary of State

84 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 64 of :12 1 in December. That is, the voter registration list was 12:12 2 matched to the driver license list on first name, last 12:13 3 name, and date of birth, with a match requiring exact 12:13 4 matches on those three criteria together. Then a second 12:13 5 step was involved, which was to reduce that by 12:13 6 eliminating those records that had an official ID number 12:13 7 that is not null, and that reduced the December no match 12:13 8 data set to 795,955, if my memory serves me, about 12: , :13 10 Q. Okay. So my question is what -- what does the 12:13 11 official ID number field represent? 12:13 12 A. My understanding is that field represents an 12:13 13 indicator of whether the voter, the voter registrant, 12:13 14 provided a statement to the Secretary of State that he 12:13 15 or she had ID at the time of registration. 12:14 16 Q. Is that a check box on the form? 12:14 17 A. I do not know what the form is, but it is an 12:14 18 indicator variable or an indicator field in the data 12:14 19 set. 12:14 20 Q. And so this step in the match process 12:14 21 eliminated anyone who had any value in that field in 12:14 22 their record in the voter registration database? 12:14 23 A. Yes. 12:14 24 Q. Was there any check done to assess the accuracy 12:14 25 of the official ID number provided in that field?

85 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 65 of :14 1 A. My understanding is that no processing or 12:14 2 filtering was done at that time. 12:15 3 Q. Looking at Exhibit 881, from which you were 12:15 4 reading a moment ago, or summarizing, there are two 12:15 5 portions in italicized text there that look like file 12:15 6 names to me because they end in.sql. 12:15 7 A. Yes. 12:15 8 Q. First of all, do you know what.sql indicates? 12:15 9 A. I presume that it refers to SQL. 12:15 10 Q. Okay. And what is SQL? 12:15 11 A. SQL is a database management or database 12:15 12 program. 12:15 13 Q. Is it like the SAS programming language? 12:15 14 A. Well, I wouldn't characterize its language as 12:15 15 being similar to SAS, but it has similar capabilities, 12:15 16 but it's more particularized to data management. 12:15 17 Q. Okay. 12:15 18 A. Whereas SAS's purpose is to provide data 12:15 19 management, report writing, statistical analysis, and so 12:15 20 forth, as I indicated earlier. 12:16 21 Q. Okay. And so these SQL files would be the code 12:16 22 or the program that was used in these two steps, 12:16 23 correct? 12:16 24 A. Well, I don't know if it is the code or if it 12:16 25 simply refers to the database itself.

86 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 66 of :16 1 Q. So you don't know what that file name 12:16 2 represents? 12:16 3 A. Whether it is the code or whether it's the data 12:16 4 itself, I do not know. 12:16 5 Q. It could be either? 12:16 6 A. I do not know. 12:16 7 Q. Is SQL -- are databases -- scratch that. 12:16 8 How familiar are you with SQL? 12:16 9 A. I am not very familiar with SQL. 12:16 10 Q. What did Mr. Valles tell you about these two 12:17 11 items? 12:17 12 A. Which two items? 12:17 13 Q. The ones that end in.sql on Exhibit :17 14 A. He did not. 12:17 15 Q. Did he show you any SQL code? 12:17 16 A. Yes. 12:17 17 Q. What did he show you? 12:17 18 A. He showed me a computer screen which he was -- 12:17 19 he had some programming there, which I presume was in 12:17 20 SQL, but I am not very familiar with SQL. 12:17 21 Q. And you reviewed that in preparation for your 12:17 22 deposition here today? 12:17 23 A. He showed it to me, but, you know, I did not do 12:17 24 any more than simply look at it. 12:17 25 Q. What did Mr. Valles represent to you he was

87 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 67 of :18 1 showing you? 12:18 2 A. Processing code. 12:18 3 Q. The code used in the match? 12:18 4 A. I presume. 12:18 5 Q. Well, what did Mr. Valles tell you he was 12:18 6 showing you? 12:18 7 A. That, yes. 12:18 8 Q. That being the code used in the match? 12:18 9 A. The code. 12:18 10 Q. Did the code require unique matches? 12:18 11 A. As I indicated, first name, last name, and date 12:18 12 of birth required exact matches between the SOS voter 12:18 13 registration date set and the driver's license data set. 12:19 14 MR. SELLS: And that is nonresponsive. 12:19 15 Q. (By Mr. Sells) Did the code require unique 12:19 16 matches? 12:19 17 MR. BRISSENDEN: Objection, vague, asked 12:19 18 and answered. 12:19 19 A. If you are asking me if I can judge whether the 12:19 20 code did that, I have to rely upon Mr. Valles's 12:19 21 representation. 12:19 22 Q. (By Mr. Sells) And what was his representation? 12:19 23 A. That it did. 12:19 24 Q. It required a unique match? 12:19 25 A. It required a unique match on all three items

88 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 68 of :19 1 simultaneously. 12:19 2 Q. First of all, let's clear up any confusion that 12:19 3 might exist about the term. What is a unique match? 12:19 4 A. A unique match occurs when there is a 12:19 5 character-by-character match in the sequence given 12:19 6 between a value in one field in one database and the 12:19 7 corresponding value in the same field in another 12:19 8 database. 12:19 9 Q. If there are two John Smiths with a birth date 12:20 10 of January 1st, 1950, in both of the databases, 12:20 11 according to this code, would that be a match or a not 12:20 12 match? 12:20 13 A. You would have to compare them letter by 12:20 14 letter, because when you say John Smith, I do not know 12:20 15 if you're spelling it, for instance, S-m-i-t-h or 12:20 16 S-m-y-t-h-e, or if there may be an extra space between 12:20 17 the names or maybe not. 12:20 18 Q. For purposes of this hypothetical, let's assume 12:20 19 they are either identical; the key factor is that there 12:20 20 are two entries with identical first name, last name, 12:20 21 date of birth in each database. 12:20 22 A. Yes. 12:20 23 Q. Does produce a match or a no match? 12:20 24 A. It produces a match. 12:20 25 Q. Does it produce one match or a two match?

89 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 69 of :20 1 A. It produces a match of the record was in the 12:20 2 voter registration file with the DPS file. 12:21 3 Q. Okay. But does -- I'm going to use the 12:21 4 shorthand John Smith 1 and John Smith 2. Does John 12:21 5 Smith 1 match with John -- in the voter registration 12:21 6 database, does John Smith 1 match with John Smith 1 in 12:21 7 the DPS database, or does it match with John Smith 2 in 12:21 8 the database? 12:21 9 MR. BRISSENDEN: Objection, vague. Do you 12:21 10 understand? Do you understand the hypothetical? 12:21 11 THE WITNESS: I understand the 12:21 12 hypothetical, yes. 12:21 13 A. It doesn't really matter for the purpose of 12:21 14 generating the no match data set. The voter 12:21 15 registration record that matches some record in the DPS 12:21 16 data set is a match of that voter registration, and so 12:21 17 that voter registration record would be excluded from 12:21 18 the no match data set. 12:21 19 Q. (By Mr. Sells) Was the matching case sensitive? 12:22 20 A. Yes, it was. 12:22 21 Q. For both December and May? 12:22 22 A. Let me backtrack. I presume that that's the 12:22 23 case, but since I am not a -- I'm not that familiar with 12:22 24 SQL, I can't say for sure. But I presume that is the 12:22 25 case, that it is not case sensitive.

90 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 70 of :22 1 Q. Okay. 12:22 2 A. That it is case sensitive. I'm sorry. I 12:22 3 believe that it would look at the character-by-character 12:23 4 comparison. And there are different as ASCII codes and 12:23 5 different computer codes for a capital letter versus a 12:23 6 lower case letter, and it's actually those computer 12:23 7 codes that are compared in comparison. And so yes, it 12:23 8 would be case sensitive. 12:23 9 Q. But you don't have any knowledge of whether 12:23 10 indeed it was or was not? 12:23 11 A. I think so, yeah. But in my review of the 12:23 12 files, I looked at a number of describers, and I think 12:23 13 that all of them are upper case in both data sets. 12:23 14 Q. Okay. And would the code return a match if 12:23 15 there were null entry in either of the three -- in any 12:23 16 of the three fields; first name, last name, or date of 12:23 17 birth? 12:23 18 A. Excuse me. Let me understand that. Are you 12:23 19 saying that taking the case John Smith, that you gave, 12:24 20 suppose John Smith is in the voter registration 12:24 21 database, and you have null Smith in the driver's 12:24 22 license file, then that would not be a match. 12:24 23 Q. Okay. Let me give you a different 12:24 24 hypothetical. In the voter registration database, you 12:24 25 have null Smith, January 1st, 1950, and in the driver's

91 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 71 of :24 1 license database, you have null Smith, January 1st, 12: Would that be a match? 12:24 3 A. I'm not sure. 12:24 4 Q. Who would know? 12:24 5 A. Mr. Valles would know. 12:24 6 Q. What would you need to do in order to be sure? 12:24 7 A. To ask Mr. Valles. 12:25 8 Q. If Mr. Valles showed you the code, would that 12:25 9 enable you to find the answer? 12:25 10 A. Well, since I'm not familiar with SQL, maybe 12:25 11 not. I'm sure that he could point out where in the code 12:25 12 that is addressed. 12:25 13 Q. I understand -- I think it's from your report, 12:25 14 or perhaps looking at the data -- that there are a lot 12:25 15 of what appear to be, potentially fictitious dates of 12:26 16 birth, for example, January 1st of a certain year. Was 12:26 17 there any accounting for those fictitious birth dates? 12:26 18 A. Are you referring to the matches that generated 12:26 19 the no match data sets in December and May? 12:26 20 Q. Yes, I am. 12:26 21 A. They were not filtered. 12:26 22 Q. Going back to a hypothetical again, let's posit 12:26 23 that there are several John Smiths in the voter 12:26 24 registration database. All of them are born on January 12: st, 1950, but only one John Smith born January 1, 1950

92 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 72 of :26 1 in the driver's license and license to carry databases, 12:27 2 so only one on the other side, is that one match, zero 12:27 3 matches, or three matches? 12:27 4 A. I am not sure of the way the code worked for 12:27 5 that hypothetical. 12:27 6 Q. Do you understand what I -- the term many-to- 12:27 7 one matching? 12:27 8 A. Yes. You were talking about that in your 12:27 9 hypothetical. 12:27 10 Q. And you don't know whether the code allows for 12:27 11 a many-to-one match? 12:27 12 A. Well, I'm sure that SQL itself does, but the 12:27 13 code that Mr. Valles wrote, I do not know. 12:27 14 Q. Okay. Thank you for that clarification. Yes. 12:28 15 And have you heard the term one-to-many 12:28 16 matching? 12:28 17 A. Yes. 12:28 18 Q. And would be an example of one-to-many 12:28 19 matching? 12:28 20 A. Well, if you take your hypothetical and reverse 12:28 21 it, taking the one John Smith from the driver's license 12:28 22 and putting him in the voter registration, take the 12:28 23 three John Smiths that you had in voter registration, 12:28 24 and put them into the driver's license, that would be an 12:28 25 example of matching the one in the voter registration to

93 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 73 of :28 1 the many in the driver's license. 12:28 2 Q. And you don't know whether the matching code 12:28 3 that was used would return one match, zero matches, or 12:28 4 three matches? 12:28 5 A. I am not sure. 12:28 6 Q. But Mr. Valles would know? 12:28 7 A. Yes. 12:28 8 Q. And -- 12:28 9 A. I would presume that since the purpose was to 12:28 10 see how many of the voter registration records remained 12:29 11 after doing the process that in the case of one-to-many, 12:29 12 finding a match for the one would match out that 12:29 13 particular individual, and then the program would go on 12:29 14 to the next person on the list. 12:29 15 Q. But that's just a presumption? 12:29 16 A. That's a presumption. 12:29 17 Q. That's speculation? 12:29 18 A. Yes. 12:29 19 Q. And Mr. Valles didn't tell you that for 12:29 20 certain? 12:29 21 A. Correct. 12:29 22 Q. What software was used for the match? And 12:30 23 let's use May, since we have that on Exhibit :30 24 A. Yes. SQL. 12:30 25 Q. SQL? Was SQL used for the December match as

94 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 74 of :30 1 well? 12:30 2 A. I do not know that. 12:30 3 Q. Did Mr. Valles write the code himself, or did 12:30 4 he get the SQL code from whoever did the December match? 12:30 5 A. I do not know that. 12:30 6 Q. So, in fact, you don't know whether it was the 12:30 7 same code run in May versus December? 12:30 8 A. I do not know that. 12:31 9 MR. SELLS: We're at a pretty decent 12:31 10 stopping point, so why don't we break for lunch. I may 12:31 11 be done with the 30(b)(6) portion. I need to check with 12:31 12 home base and see if they have any other questions to 12:31 13 ask. But how long shall we take for lunch? Do you want 12:31 14 to take an hour? 12:31 15 MR. SWEETEN: Sure. 12:31 16 MR. BRISSENDEN: That will be fine. 12:31 17 MR. SELLS: Okay. 12:31 18 MR. BRISSENDEN: Plan on being back here 12:31 19 about 1:30? 12:31 20 MR. SELLS: About 1:30? 12:31 21 Folks on the phone, why don't I call back 12:31 22 in? Does that work for you guys? 11:56 23 (Recess from 12:31 a.m. to 1:43 p.m.) 13:44 24 MR. SELLS: Folks on the phone, are you 13:44 25 there? Back on the record after lunch.

95 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 75 of :44 1 Q. (By Mr. Sells) Professor Sager, I understand 13:44 2 over the lunch break, you spoke with Mr. Valles and may 13:44 3 have some additional answers to the questions I posed 13:44 4 before lunch. 13:44 5 A. Yes, that's correct. 13:44 6 Q. What can you answer now that you couldn't 13:45 7 answer before? 13:45 8 A. During lunch, I talked telephonically with 13:45 9 Oscar Valles and asked him about the questions that you 13:45 10 posed earlier regarding the one-to-many and many-to-one 13:45 11 matching. 13:45 12 Q. Okay. 13:45 13 A. Okay. If you take the case of one-to-many, 13:45 14 meaning one record in the voter registration file that 13:45 15 matches by first name, last name, and date of birth to 13:45 16 any driver license file, that one record was indicated 13:45 17 as a match, and no record was made which of the records 13:45 18 that matched to in the driver's license file. 13:45 19 Then with regard to the many-to-one 13:45 20 matching where there might be, say, three records that 13:45 21 are the same on first name, last name, and date of 13:45 22 birth, in the voter registration file, matching to one 13:45 23 identical and first name, last name, date of birth in 13:46 24 the driver's license file, all three of the voter 13:46 25 registration records would be marked as a match.

96 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 76 of :46 1 Then one other question I asked him had to 13:46 2 do with the SQL coding. Mr. Valles indicated to me that 13:46 3 he wrote the SQL code; that it was, in fact, SQL that 13:46 4 was used for the matching; that he did not receive any 13:46 5 SQL code from the Secretary of State, referring to the 13:46 6 December matching that SOS did. 13:46 7 Q. Given that Mr. Valles wrote the SQL code from 13:46 8 scratch, then can you be sure that the May match was, in 13:47 9 fact, done the same way as the December match? 13:47 10 A. Mr. Valles indicated to me that they attempted 13:47 11 to do the same thing that the SOS did in December, which 13:47 12 is to match on identical first name, last name, and date 13:47 13 of birth, and then apply the official ID number as we 13:47 14 discussed before the break. 13:47 15 Q. Did you discuss with Mr. Valles whether a null 13:47 16 entry could be part of a match? 13:47 17 A. I did not. 13:47 18 Q. So you still don't know the answer to that 13:47 19 question? 13:47 20 A. That is correct. 13:47 21 Q. Okay. And did you talk with Mr. Valles about 13:47 22 whether the match was required to be case sensitive? 13:47 23 A. I did not. 13:47 24 Q. So you still don't know the answer to that 13:48 25 question?

97 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 77 of :48 1 A. That was a question? 13:48 2 Q. Yes. 13:48 3 A. Yes, that's correct. 13:48 4 Q. Did Mr. Valles tell you anything else related 13:48 5 to Topics 10 or 12 on Exhibit 880? 13:48 6 A. You're referring to the conversation at lunch? 13:48 7 Q. Yes. 13:48 8 A. No. It was a brief conversation. 13:48 9 Q. All right. Well, I appreciate those updated 13:49 10 answers. Is there anything else you would like to 13:49 11 correct from the morning session or update? 13:49 12 A. No. 13:49 13 Q. No? Okay. 13:49 14 MR. SELLS: Well, in that case, I just 13:49 15 want to put on the record that it is our position that 13:49 16 the State has failed to produce a witness who could, 13:49 17 with knowledge and ability, testify as to the entirety 13:49 18 of Topics 10 and :49 19 The information from Valles over the lunch 13:49 20 hour filled in some of the gaps, but not all of them. 13:49 21 There's, obviously, still matters that Professor Sager 13:49 22 doesn't know and isn't able to testify about. And that 13:49 23 includes both the matching process, but also the first 13:49 24 clause of Topics 10 and 12, involving the production of 13:49 25 the data from the Department of Public Safety and the

98 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 78 of :50 1 Secretary of State's Office that was used in the match 13:50 2 that ultimately came to Professor Sager. 13:50 3 MR. BRISSENDEN: The State's position is 13:50 4 that Dr. Sager has been reasonably informed of topics 13:50 5 and of information required to be reasonably informed 13:50 6 about Topics 10 and 12, as they are described in the 13:50 7 request. We have -- the State has made an effort to 13:50 8 have Dr. Sager here and to answer your questions here 13:50 9 today. And over the lunch hour, we made the effort to 13:50 10 try to fill in those gaps. But, you know, it is our 13:50 11 position that he is reasonably informed, and the State 13:50 12 is presenting him here today. 13:50 13 Other than two questions about null match 13:50 14 and case sensitive, are there any other questions that, 13:51 15 in your position, he has not been able to answer? 13:51 16 MR. SELLS: Well, there were certainly 13:51 17 more than that, but the large areas of knowledge, 13:51 18 missing knowledge includes everything that happened 13:51 19 prior to the production of -- I mean, prior to the 13:51 20 match. 13:51 21 MR. BRISSENDEN: And it's our position, 13:51 22 the State's position that prior to the match, that area 13:51 23 of topic, as described as the process, manner, and 13:51 24 method of compiling and cleaning and filtering data from 13:51 25 either the TEAM database or from the Department of

99 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 79 of :51 1 Public Safety driver's license and license to carry 13:51 2 databases, that same language and same description 13:51 3 appears in Topics 9 and :51 4 Two days ago, the State of Texas presented 13: designated and presented specific witnesses from 13:51 6 those departments, the Department of Public Safety and 13:52 7 from the Secretary of State's Office. Three witnesses 13:52 8 were presented. Mr. Freeman from the Department of 13:52 9 Justice spent a good portion of the day asking those 13:52 10 questions under those topics of those specific three 13:52 11 witnesses. 13:52 12 So with that, I believe we -- the State 13:52 13 has fulfilled its obligation in designating witnesses 13:52 14 responsive to all four topics. 13:52 15 MR. SELLS: Okay. Well, in that regard, I 13:52 16 would note that the three witnesses you identified were 13:52 17 not designated for any portions of Topics 10 and :52 18 Professor Sager is the only witness 13:52 19 designated for 10 and 12, which includes the first 13:52 20 clause that I mentioned. But we don't need to argue on 13:52 21 the record here. 13:52 22 Just a couple of loose-end questions from 13:53 23 me, perhaps, and then we can put the 30(b)(6) portion on 13:53 24 hold and open up the expert portion. 13:53 25 Q. (By Mr. Sells) Professor Sager, do you know

100 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 80 of 80 13:53 1 whether the December match process is similar to or 13:53 2 different from earlier match processes that the State 80 13:53 3 undertook, for example, in September? 13:53 4 A. I do not know of a September process. 13:53 5 Q. You're not even aware that a match occurred in 13:53 6 September? 13:53 7 A. That's correct. 13:53 8 Q. Okay. 13:54 9 MR. SELLS: I think that's all I have on 13:54 10 the 30(b)(6) portion. And pursuant to our discussion 13:54 11 earlier, the United States would like to leave it open 13:54 12 pending any potential motions that may arise from the 13:54 13 dispute over the lack of knowledge. 13:54 14 MR. BRISSENDEN: Well, we're not going to 13:54 15 consent or agree on that today, but we can work through 13:54 16 that. 13:54 17 MR. SELLS: Okay. 18 (Signature reserved.) 19 (Deposition concluded at 1:54 p.m.)

101 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 1 of 12 THE UNITED STATES DISTRICT COURT FOR THE DISTRICT OF COLUMBIA STATE OF TEXAS, Plaintiff vs. ERIC H. HOLDER, JR., in His Official Capacity as Attorney General of the United States, Case No. 1:12-CV (RMC, DST, RLW) Three-Judge Court Defendant. EXPERT DECLARATION OF THOMAS SAGER 1. I have been retained by the Office of the Attorney General ( OAG ) to provide expert services and assistance to the OAG in connection with its legal representation of the State of Texas. 2. I am professor of statistics in the Department of Information, Risk, and Operations Management in the McCombs School of Business at the University of Texas at Austin. I have a PhD in statistics from the University of Iowa in I have 38 years of experience teaching a wide variety of statistics courses, research in the theory and applications of statistics, and extensive consulting. My resume is attached as Exhibit A. 3. My hourly rate is $200, and my list of prior testimony is attached as Exhibit B.

102 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 2 of I understand that one of the issues in this litigation involves an attempt to determine the number of voters who may lack drivers licenses or other state identification by matching voter registration data provided by the Texas Secretary of State (SOS) to drivers license, license-to-carry (LTC) and state identification holders provided by the Texas Department of Public Safety (DPS). 5. I received DPS license and identification and SOS voter registration databases compiled in December I also received a separate database of 795,955 entries (the December No Match database) that I understand represents those records from the December SOS voter registration database that fail to match entries in the December DPS database based upon the strict requirement that a match exists only if all three of First Name, Last Name, and date of birth (DOB) are exactly identical between the SOS and DPS databases and further excludes those who self-reported to SOS that they had such an identification. 6. Second, I received SOS and DPS databases from May 2012 as well as a separate database of 588,095 entries (the May No Match database) that I understand represents those records from the May SOS voter registration database that fail to match entries in the May DPS database using the same criteria of paragraph I was asked to characterize these data sets and to perform specific analyses as detailed below in the tables that follow in this declaration. 8. Before discussing the results of my analysis, it is useful to explain how I used the databases to determine whether additional matches could be made of

103 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 3 of 12 registered voters to those holding qualifying identification (drivers license, state identification, or LTC). I understand that SOS created the initial set of 795,955 registrants by attempting to match December voter registrants to the December DPS database using identical First Name, identical Last Name, and identical date of birth (DOB). I understand that the same matching procedure was later followed to attempt to match voter registrants in the May SOS database to persons holding qualifying identification in the May DPS database. However, data entry errors and alternative names may prevent discovery of true matches on the basis of these strict criteria. For example, if an individual appears on the voter registration list as Marcus Johnson, but the same individual appears in the DPS database as Mark Johnson, the strict SOS criteria would not record a match and he would be listed in the No Match database. As another example, if Gladys Smith obtained a drivers license, but then married, changed her name, and registered to vote as Gladys Gonzalez, then she would not match and would be listed in the No Match database. 9. In addition, data entry errors can cause a failure to match. For example, the December No Match database contains thousands of suspect dates of birth. Among its 795,955 records, there are over 22,000 entries for persons born in 1900 (who would be 111 or 112 today), and over 3,000 entries for persons born in 1901, as well as over 17,000 for persons born in Many of these DOBs are 1/1/1900 or 1/1/1901, or 11/11/11 further indicating that these are likely not

104 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 4 of 12 accurate DOBs. 1 Registrants who have inaccurate DOBs in the SOS database but accurate DOBs in the DPS database are included in the 795,955 record No Match database, given my understanding of how the No Match database was generated. 10. The construction of neither the December nor the May No Match database involved the use of social security numbers in the attempt to match SOS voter records to corresponding DPS identification records. However, many SOS voter records include social security numbers either in full 9-digit format (SSN9) or 4-digit format (SSN4). The December and May DPS databases also contain social security information, typically in SSN9 format. With this additional identifying information, more individuals can be matched between the December (or May) No Match data set and the corresponding DPS databases. Therefore, I undertook additional matches of the December No Match database to the December DPS databases and of the May No Match database to the May DPS database using (1) SSN-9 (two distinct individuals should not have the same complete SSN) 2 and (2) SSN-4 plus First Name plus Last Name (all three criteria required to match in order to declare a match for the individual). 3 I performed spot checks in the matching to ensure that there were secondary indicia that these were the same individuals (e.g. for SSN9 that the name was similar even if not identical). 1 Parallel suspect counts of persons born in 1900, 1901, and 1911 occur in the May No Match database. 2 Fraud, data entry errors, confusion, or other mix-ups could result in identical SSNs for distinct individuals. 3 The probability that two distinct individuals with identical first and last names also have the same SSN4 is quite low one in 10,000, assuming random distribution of digits in SSN4.

105 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 5 of 12 Analysis of December No Match Data Set 11. The results of the SSN matching process for the December No Match dataset are below: Table 1 Matches from December No Match to December Drivers License Using SSN Match Criteria (795,955 to DL) Number of Matching Records SSN9 222,795 SSN4 & First Name & Last Name 5, I performed the same two matches between the 795,955 December No Match data and the December LTC database and obtained the following results: Table 2 Matches from December No Match to December LTC Using SSN Match Criteria (Dec. No Match to LTC) Number of Matching Records SSN9 16,195 SSN4 & First Name & Last Name ,251 of the 795,955 December No Match records were matched to both the DL and LTC databases on the basis of the criteria. 14. I was also asked to make an accounting of those individuals in the 795,955 December data set that will be over the age of 65 at the date of the next general election. I found that 239,055 entries in the 795,955 December data set are for individuals who will be 65 or older with birth years between 1900 and 2000, inclusively. (I excluded birth years prior to 1900 as likely the result of data entry errors or a lack of accurate data, as well as the birth years of 1900, 1901, and 1911 on account of the aforementioned anomalous spikes in birth counts during these years [see paragraph 9].)

106 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 6 of The following table breaks down the 795,955 December No Match records into 8 non-overlapping parts, based upon whether or not a match was found in December DPS databases (by either SSN9 or SSN4 + first name + last name), whether or not the voter registration is in suspense, and whether or not the record shows an age of 65 or more. All combinations of interest may be found by adding corresponding rows from the table. For example, the total number of matches may be found by summing the first four rows to yield 234,989. As another example, the total number of No Match records that satisfy at least one of the three criteria shown may be found by summing the first seven rows to yield 439,921. The total number of registrations in suspense may be found by summing rows 1, 2, 5, 6 to yield 84,650 and so forth. The rightmost column shows the percentage of each row that has a Spanish surname. Table 3 Breakdown of Three Filters Applied to December No Match Match? Suspense? Over 65? Count % Spanish Surname Yes Yes Yes 4,271 20% Yes Yes No 12,345 27% Yes No Yes 85,031 23% Yes No No 133,342 32% No Yes Yes 12,855 23% No Yes No 55,179 28% No No Yes 136,898 37% No No No 356,034 47% 16. In all, 439,921 or 55% of the December No Match data set fit one or more of these three filters. 234,989 were matched using auxiliary SSN based filters. 239,055 were older than 65. And 84,650 were classified as suspense. The

107 Case 1:12-cv RMC-DST-RLW Document Filed 06/23/12 Page 7 of 12 below Venn diagram represents the first 7 rows of Table 3 in graphical form to show the overlap of these three filters: 17. Further analysis of the final row in Table 3 (those not meeting any of the three filters) shows that 142,760 of these 356,034 records lack any SSN data. Altogether, 220,268 of the 795,955 December No Match data lack any SSN. Thus, there are 575,687 (= 795, ,268) No Match records that have SSN data. From rows 1-4 of Table 3, 234,989 of these were successfully matched to December DPS databases a success rate of about 40% (= 234,989 / 575,687). If SSN data were available for the 220,268 No Match records that lack SSN and if the same success rate were to apply, then one could anticipate that about 40% of the 220,268 (or more than 88,000) additional voter records might be matched with no other improvements in the quality of the record keeping and reporting. Moreover, 440,032 of the 795,955 No Match records have full SSN9s. 228,844 of these were