Data for Defense Lawyers

Long before Big Data and discovery dumps, courts warned about the danger of “trial by charts.”

In 1954, the Supreme Court warned that “bare figures have a way of acquiring an existence of their own, independent of the evidence which gave rise to them.”  Holland v. United States, 348 U.S. 121, 128 (1954).The Fifth Circuit warned around the same time that district courts have “grave responsibilities” to ensure that a defendant is “not unjustly convicted in a ‘trial by charts,’ however impressive the array produced.”  Lloyd v. United States, 226 F.2d 9, 17 (5th Cir. 1955). 

Now, more than ever, defense lawyers should guard against the improper use of data and summary charts by the government. Moreover, they should also look for ways to turn this around on the government by looking for errors and sloppiness in the government’s analysis and by finding information in the data that they can use themselves.

I.  Applicable Rules

There are a few rules to keep in mind when it comes to data analysis and summary charts.

Federal Rule of Evidence 1006 allows a party to use a “summary, chart or calculation” to “prove the content of voluminous writings” that “cannot be conveniently examined in court.” Courts have warned that summary charts are to be used with caution due to their potential for abuse.  See, e.g., United States v. Norton, 867 F.2d 1354, 1362 (11th Cir. 1989) (“We recognize the caution with which these summaries are to be utilized, given the possibilities for abuse,” and finding that there was no abuse of discretion in admitting a chart when the chart’s assumptions were “amply supported by the evidence presented to the jury,” when the witness who prepared the chart testified and explained that chart, and when the defense conducted a “thorough cross-examination of the witness” and had “the opportunity to present its own version of those matters,” and when a jury instruction was given)). 

A Rule 1006 summary should be a “surrogate” for voluminous records and must be an “objective accurate summarization of the underlying documents, not a skewed selection of some of the documents” to further one side’s theory of the case.  United States v. Oloyede, 933 F.3d 302, 310 (4th Cir. 2019) (finding that government charts did not comport with Rule 1006 because of their selectivity). 

Also, starting December 2024, Federal Rule of Evidence 107 will authorize a court to allow the use of "an illustrative aid to help the trier of fact understand the evidence or argument." Think about flowcharts and diagrams that might summarize information that is not necessarily "voluminous" but might help visual learners. Such charts also could have been allowed previously under Rule 611(a), which allows a party to use demonstrative charts to facilitate the presentation and comprehension of evidence already in the record.

The distinction between the rules can affect how the evidence comes in and what the jury can take back with them into deliberations. Charts that are admitted under Rule 1006 can be admitted as evidence, charts that are admitted under Rule 107 are typically demonstratives but can be allowed as evidence if all parties consent or if a court orders for good cause, and charts that are used under Rule 611(a) are only demonstratives.

If the government plans to use summary charts at trial, defense lawyers should consider working out a schedule for receiving draft charts so that they will have adequate time to review the charts before trial. Also, defense counsel should consider asking the government to identify the person who prepared the chart and consider asking questions about methodology.

Here are some things to consider:

  • Are the charts based solely on mathematical computations? If so, the chart is more likely to be admissible.

  • What assumptions were used to prepare the chart? Charts need not be “free from reliance on any assumptions,” but such assumptions must be “supported by evidence in the record.”  United States v. Diez, 515 F.2d 892, 905 (5th Cir. 1975) (finding no error in government’s use of chart when the government’s assumptions were “amply supported by evidence already presented to the jury”). 

  • Are the underlying records accurate, reliable, and admissible? Summaries of records are only admissible if the underlying records are otherwise admissible. Consider hearsay objections and point out ways that the records may not be admissible. For example, using bank records rather than accounting records may be better, as accounting records incorporate hearsay that may not be admissible.  United States v. Batio, 2023 WL 8446388 (7th Cir. 2023) (Quickbooks records effectively put in the defendant’s characterizations of the financial transactions. Underlying records should have been used.). 

  • Have all the underlying records been made available? Charts should be clearly sourced to other exhibits or discovery. If something has not been sourced, this can be a huge red flag. This is especially important now when data may be summarized from a live database at a particular time, making it impossible to reconstruct later exactly what records were summarized.

  • Are the underlying records being admitted into evidence? Rule 1006 does not require that the underlying records be admitted, but requesting that the underlying records be admitted is generally better. This way, defense counsel can use the underlying records to point out errors in the government's analysis and to make points that the government may not have anticipated.

  • Do the charts contain “conclusory captions”? If so, the chart is less likely to be admissible under Rule 1006. A witness may be able to testify about such conclusions, but there is a “danger in permitting the unrestricted use of such phrases upon charts” given “a jury’s natural tendency to accept such unsworn, conclusionary verbiage as authentic, primary proof.” Lloyd, 226 F.2d at 17. 

Given all this, request summary charts well before trial, particularly in complex cases where the analysis may take a while. Analyzing these charts can give the lawyer insights into what the government is planning to do at trial, which can open up ways to attack or undercut the government’s case.

II.  Think of Data Like a Witness

The government regularly dumps spreadsheets and financial records in the course of discovery, and the government itself often has not done a good job analyzing and understanding this information. But all this data can be very useful for defense lawyers and should be considered as part of overall case development and strategy.

Think of the data as a witness who the defense team can interview at length, who cannot be impeached, who can help impeach the government’s witnesses, and who might remember things that counsel’s client has forgotten or never even knew. It may take a lot of time and work to debrief that “witness,” but the data can help the defense present the client’s story and corroborate the defense in ways that the government may not even anticipate.

A big part of analyzing data is thinking about what questions data can help answer. Use questions like the ones below to help figure out if there are errors, sloppiness, and assumptions in the government’s case, just like counsel would review the interview reports that are the traditional part of discovery.

Does the data corroborate what the government’s witnesses or what the defense witnesses say? If what the witnesses say matches the data, then the witnesses have more credibility. But if what the witnesses say does not match the data, then the witnesses do not really know what they are talking about or are lying. The government’s lawyers should be checking this, but they sometimes take witnesses at face value.

In one of my cases, the government had multiple witnesses testify about alleged “patterns” of tests that a doctor always ordered. But the government had not checked this against the data, which showed that most of those tests were never billed. Either the witnesses did not know what they were talking about, or the doctor was performing tests for free. In another case, the government had a witness testify about some significant conversations but failed to check that witness’s story against the phone records. Those records showed that there was at least one significant call within the key time frame that the witness had not explained and could not recall, and they showed that another significant call might not even have happened. 

Does the data show the proper context for the government’s claims? The government likes to use big numbers, but those numbers can be broken down and picked apart. How much money is really at stake? A large amount of money might support knowledge and criminal intent, but if the money at issue is a small percentage of total payments or income, it might be more likely to suggest errors and mistakes.

Does the data show improper cherry-picking? The government probably has chosen specific examples that it thinks are strong and powerful, but defense counsel should check if they are really representative of the alleged scheme. The data can help show that the government’s examples are more complicated than the government thinks, or that the government’s examples are one-off instances that are not indicative of the overall practice.

One of the best examples of countering improper cherry-picking in recent years happened in a case of constitutional interpretation involving the Emoluments Clause. During the Trump administration, the Department of Justice tried to argue for its interpretation of “emolument” in part by cherry-picking a few dictionaries from the 1600s and 1700s that were in its favor. Rather than just cherry-picking a few counter examples from other dictionaries, the lawyers arguing against the DOJ presented the judge with a full survey of every dictionary they could find. The overall data analysis showed that the Trump administration’s argument was in the minority, and the judge cited this in part in ruling against the Department of Justice.  District of Columbia v. Trump, 315 F. Supp. 3d 875, 889-92 (D. Md. 2018). 

Does the data show that there is more to the story than the government alleges? One big challenge from the defense perspective is overcoming a psychological bias that Daniel Kahneman has called assuming “what you see is all there is.” The government wants to present a simple story of fraud, but the data can show variation and complexity that the government and its witnesses may not be able to explain, and that may help sow doubt among jurors.

Did the government get something wrong in the indictment? Check the wires or the submission dates and see if there is anything that the defense can use to undercut the government’s credibility. In a healthcare fraud case, the defense discovered almost immediately that the government had misread its own data, confusing the submission date for each healthcare fraud count with the date of service. The government did not realize the error until halfway through trial, which was too late to fix the indictment and avoid some embarrassing testimony.

III.  How to Review Data

One of the big obstacles for using data in a defense case is simply that many lawyers lack experience with it. Lawyers often joke about how they chose the legal profession because they were not good at math, but the evidence for many white collar crimes is in the spreadsheets and data the government provides or lawyers develop themselves. This is the era of big data, and not looking at data is leaving potential evidence on the table.

At its core, data analysis can help defense lawyers see things that otherwise would be invisible or difficult to see if they were looking at pieces of evidence individually or in isolation. Three general approaches will help the defense team analyze data: (1) show the denominator, (2) compare and contrast something, and (3) track something over time.

  1. Show the Denominator

In white collar cases, the government will try to show that something happened multiple times, trying to convince members of the jury that something happened so many times that they should infer a pattern or bad intent. But what is the denominator? What is the government’s number a percentage of? Numbers should always be placed into context, and calculating the percentage is a key step in this regard.

In a healthcare fraud case, the government tried to show that the defendant had performed a few hundred operations that were improper according to an expert and that the operations were not randomly selected. Even if the government was right about the merits of the operations (it was not), these operations represented less than one percent of his total operations and were not meaningful when viewed in context.

2. Compare and Contrast Something

Another approach is to compare one data set to another to determine if similarities exist. In a case involving the Anti-Kickback Statute, the government cooperating witness testified that there was an agreement to be paid on a per-patient basis. During cross-examination, defense counsel showed the witness the government’s own data indicating the number of patients that were billed to Medicare on a monthly basis. Defense counsel did the math with him for each month, and the data showed that there never was a month even close to matching the details of his testimony. Even the government’s case agent conceded on cross-examination afterwards that defense counsel’s “Excel class” had shown that there was no agreement to be paid on a per-patient basis.

3. Track Something Over Time

The government might be focused on a particular event or transaction, but what happened before or after might provide useful context. In a recent case, the government used one set of records to show a high number of refunds that should have been made. But defense counsel focused on the process and showed that the government’s extensive analysis was based on records that the defendant did not have and had no reason to look for. He was given a different set of records, and those records showed a lower number of refunds that he thought was better than the industry average.

In another case, where the government focused on one type of medical operation, the defense showed that the defendant had done many precursor operations without doing the type of operation on which the government had focused. This showing undercut the government’s suggestion that the defendant did both types of operations with criminal intent.

Defense lawyers should think about how this can be done with anything that they get in discovery. It may be possible to pull data out of the discovery they have been given. If the government interviewed employees, track when those employees worked for the company and determine if there are key time periods that are not covered. If the government is giving counsel patient files all signed on the same day, counsel should check the time stamps to see if that helps provide context.

Even criminal records can be data. One famous use of data in a criminal case was cited by Yale professor Edward Tufte as a great example of data presentation. The defense attorneys in the case made a demonstrative chart tracking all the crimes that the government’s cooperating witnesses had been convicted of. The jury even requested a copy of this chart during deliberations and then acquitted the defendant.

If new lawyers and law students do not know how to use Microsoft Excel or similar spreadsheets, they should take a class or seek instruction from friends or colleagues. Knowing how to read and analyze a spreadsheet is a valuable skill for any complex litigation, but this is a skill many lawyers do not have.

Here are a few tools that can help counsel focus or identify the important data.

  • The “hide” function. Spreadsheets provided by the government often have more columns than can be seen on a screen or that can be printed out in a legible manner. Use the “hide” function to take out columns that are not significant to the relevant analysis.

  • The “filter” function. Spreadsheets often will contain hundreds or thousands of rows of data. How can counsel focus on particular aspects of the data? Use the filter function to show only specific rows, such as the rows associated with a particular date or person. For example, using filter can help counsel focus on the transactions or people associated with a particular count of the indictment.

  • The “sort” function. Every spreadsheet will come to counsel sorted in some way that may have made sense when the data was created or last saved, but counsel can sort it other ways. Use the sort function to put the data into chronological order, in order by amount of money, or in some other way.

  • The “page format” function. When printing spreadsheets, one should adjust the settings so that the printouts are more legible. First, go to “Page Setup” and change the “orientation” to “landscape” and change the “scaling” to “fit to” 1 page wide while leaving the “tall” pages blank. Second, change the “sheet” to repeat “$1:$1,” which will result in Excel printing the column headers on every page. Third, hide all columns that are not significant. These steps will result in a spreadsheet that can be printed out and reviewed.

Much can be done with Pivot tables and formulas, including having the computer analyze the data. Here are two things that can be done by tech-savvy defense counsel or someone else on the defense team:

  • Count how many times something appears in the data. How many times did a particular type of transaction occur? In a healthcare fraud case, how many times did the doctor bill for a particular billing code? The “countifs” function lets the user do that.

  • Add up something. How much money did a particular person receive? How much money was associated with a particular billing code in a healthcare fraud case? The “sumifs” function lets the user do that.

Lawyers can use these formulas to ask questions of the data, just like they would ask questions of a witness.

Start with the counts in the indictment. If the government has picked a particular victim or patient, look at all the transactions involving that victim or patient. If the government has picked a particular day, look at everything that occurred that day. If the government has picked a particular billing code, look at how that billing code compares with other billing codes.

With healthcare fraud cases, one approach is to start by making three lists using the “sumifs” and “countifs” functions, but the same thing can be done with pivot tables. First, one can create a patient list to help figure out how the government chose the patients cited in the indictment and if there are patients whom counsel can use to undercut the government’s case. Second, one can make a list of billing codes to help put the government’s allegations into perspective against the overall practice. Finally, create a list that tracks payments by month so counsel can see if there are any particular points in time that could be significant, such as a shift in billing that might reflect a conversation. These lists can help counsel get a better sense of the client’s practice and sometimes help counsel come up with leads to investigate further.

IV.  How to Present Data

The government’s presentation of data in trials is typically very simple. A witness looks at a disc or a USB drive, confirms that he or she reviewed it before trial and that there are big spreadsheets on it, and the government moves to admit the disc or USB drive into evidence. And then the government moves immediately to the summary charts that make a few points based on the data.

Whenever the government does this, witnesses leave themselves open to a cross-examination that goes into the nuances of the data and shows that there is more going on. Has the government glossed over things? Has the government made bad assumptions? This is defense counsel’s opportunity to show it.

First, find something in the data that the government has glossed over or has not explained, and show this to the jury on cross-examination. Open those spreadsheets that the government admitted into evidence but did not show the jury. If defense counsel is lucky or the government has been sloppy (or both), defense counsel will be able to point out information in the government’s exhibits that undercuts the government’s case.

Second, defense counsel should use the government’s own witnesses to make some of the points the defense identified from its data analysis. The government’s data custodians just testified that they reviewed the data before trial and that they created summary charts for the government. Start filtering and sorting and hiding, and make those witnesses read into the record some facts that are good for the defendant.

Third, defense counsel may want to have a witness create summary charts that can be admitted via Rule 1006 and that make points which require more than filtering and sorting. These summary charts can also be helpful for jurors, who often use visual aids to help understand complex points and see things for themselves.

Rule 1006 generally requires a witness to have personally done the summary before court, but the witness does not need to have been the first person to do so. Defense lawyers should consider doing multiple types of analysis themselves and figure out which few are really important for trial. After this step, counsel can work with a witness who recreates the specific analysis that matters and can authenticate a chart that sometimes the witness and defense counsel create together.

The advantage of a summary chart under Rule 1006 is that it gives the jury something concrete that they can take with them into jury deliberations. The disadvantage is that it gives the government something to pick apart and spin.

On defense, it often is good enough to make a general point without getting bogged down in specifics. The government’s job is to do and present a thorough investigation, and the defense lawyer can communicate problems with the government’s case in a comprehensible manner without providing the definitive answers that the government should have provided. In one case, a witness conducted a thorough analysis, but the high-level summary was good enough to inform defense counsel’s cross-examination of the government witnesses and raise weaknesses in the government’s case. Had defense counsel created a summary chart, he might have had to turn it over well before trial, and the government might have fixed some of the problems or dropped some of its arguments. Using the data to inform the cross-examination of the government’s witness and to set up a short direct examination of the defense witness was good enough.

V.  How to Get Data

The defense team should not rely only on the data that the government thinks is relevant. Prosecutors may have dumped massive amounts of data on the defense, but that is just the data they thought was useful or might have been useful to them. If they conducted a sloppy investigation, they probably did not collect or analyze data that might be useful to the defendant.

Catalog the data the government provided, and then consider what is missing.

Did prosecutors obtain all the relevant bank accounts? They may have missed accounts that fill in gaps in the government’s theory. Defense counsel can check with the client for missing bank account records, and counsel can issue subpoenas for those records.

Did they collect data for the right time frame? They may have collected data at an early stage in the investigation and did not think to update the data as the investigation continued.

Did they get data that puts the data into context? The government’s lawyers may have gotten data showing all of the client’s transactions with a particular person, but they may not have gotten data showing that person’s transactions with other people.

All this is particularly important in healthcare fraud cases, which involve a data-rich environment. The Department of Justice has real-time access to Medicare claims data, and it can get data from Medicare, Medicaid, and private insurers relatively easily. If a lawyer is representing a healthcare provider, the provider should have access to all the claims the provider submitted. And defense counsel has access to some publicly available databases that show aggregate information on what was billed to Medicare.

For years, Medicare has made available aggregate data for Medicare providers, such as data summing up what services a provider billed to Medicare. One link is at https://data.cms.gov/tools/medicare-physician-other-practitioner-look-up-tool. As of 2024, this data goes back to 2013 and is updated periodically. For a client in a healthcare fraud case, this data may help counsel see things about the client’s practice that he or she may not even know.

Defense lawyers can download the data themselves from the Medicare website, or they can take advantage of the work of some journalists who have made the data available in a more user-friendly format. The news organization Pro Publica has made the data for 2013 through 2015 available at https://projects.propublica.org/treatment/, and the Wall Street Journal did its own analysis at Medicare Unmasked (https://graphics.wsj.com/medicare-billing/).

Many doctors and medical professionals do not realize how much of their data is available and can be analyzed. There are doctors and medical professionals who have been charged and convicted of crimes, and it is possible they could have avoided prosecution if they knew what that data showed. Lawyers who have clients facing an investigation should check this data.

In all cases, defense counsel should consider requesting additional data to fill in the holes in the government’s data collection.

First, consider asking the government to produce materials under Brady or 3500. If the government has data, it may just give counsel the data itself.

Second, file a motion requesting such data.

Third, issue a Rule 17(c) subpoena. In a healthcare fraud case, check the government’s production to see which Medicare contractor provided the government with the data the defense wants, and then subpoena that contractor for the data.

Do not assume that what the government collected is all that matters or is all that exists. More data can help, just like more witnesses can help.

VI.  Peer Comparisons

One way the government uses data in healthcare fraud cases is to compare a doctor or Medicare provider to others. Peer comparisons are an important piece of healthcare fraud investigations, and they sometimes are used as evidence in trial.

Defense counsel in a healthcare fraud case can do his or her own peer comparisons, at least using the Medicare data that is publicly available and the billing codes that are at issue. Counsel can do comparisons by going to the Medicare website, filtering the data for the billing code at issue, and then downloading the data in order to conduct some basic analysis and cleanup to make the data more user-friendly.

One big question is to see how the client compares in terms of absolute number of services or total payments based on such services. This is something that the government probably is very well aware of, and it might be the main data point that drives the government’s investigation.

Counsel should do some simple math to get what could be a more significant data point. Divide the total number of services by the total number of beneficiaries, and get the ratio of services to beneficiary. This approach was crucial in a case involving a doctor who was accused of doing unnecessary cardiac stent procedures. He was one of the top providers in the country for such stents, but he also saw more Medicare patients than most other providers. The absolute numbers looked bad and made him look like an outlier, but the ratio was in line with his peers. Basic data analysis helped put his practice into context and neutralized one piece of evidence that the government assumed would be damaging.

Ultimately, seeing how a client compares to peers can provide context that can be useful to point out to a prosecutor or to a jury. If the client is a significant outlier by one measure, see if that measure's significance is diluted by other measures or context. If the client is not an outlier, point out that this shows any improper claims were more likely to be mistakes and less likely to be the result of fraudulent practices. And if the government has not charged people who are more significant outliers than the client, try to use that to argue for a noncriminal resolution, especially if it can be shown that the government itself has handled similar cases inconsistently.

Here are two more points about peer comparisons.

When the government does a peer comparison, make sure that the government is comparing the client to a valid “peer” group. Medicare data categorizes providers by type, but those types do not reflect some significant differences. For example, some doctors may have a subspecialty that is not captured in the Medicare data, and thus would look like outliers when compared to the overall group but normal when compared to others in the subspecialty. Who are the client’s proper peers? Discuss this question with the client.

And fight back against any attempt to compare the client to a cherry-picked group of peers. This happened in one fraud case when the government tried to compare a cardiologist in a rural part of the United States that had a high incidence rate of heart disease to a few famous hospitals in other parts of the country. This was an improper peer group because it compared very different practices. Famous hospitals that draw patients from across the country and the world do not see the same kind of patients that go to a rural hospital, and those city hospitals do a different mix of services than rural hospitals. Fortunately, a judge agreed that this peer comparison was improper and excluded it from evidence.

VII.  Statistical Sampling

Many white collar criminal cases involve an alleged scheme or conspiracy as well as specific transactions or incidents that are charged as specific counts of that scheme. Data analysis can help address this by focusing on the significance and meaning of those specific counts.

Those counts may or may not be fair examples of a client’s practice or the overall scheme that the government has alleged. The government may attempt to cherry-pick, but defense counsel should push back on improper extrapolations and should use data to put those cherry-picked instances into context. Pushing back is especially important in terms of loss calculations when the government tries to extrapolate based on the counts of the indictment or the evidence that was presented at trial.

One way to do this is to keep in mind how the government proves loss amounts when it wants to do it completely right. The government can use random sampling to determine loss calculations that are statistically valid and meaningful. It does not have to do such sampling in criminal cases, but this is the typical method in the healthcare context, particularly via audits and civil False Claims Act cases.

If prosecutors are not using statistical sampling, push back on this and cite the government’s own procedures back at them. Ask government witnesses about statistical sampling, and ask them to explain how it is done.

The Medicare Program Integrity Manual, chapter 4, discusses how sampling should be done in noncriminal cases involving improperly billed claims to Medicare. The government should define the set of claims that are meant to be extrapolated from, which is called the “target population” or “universe.” The government should then analyze the universe to determine the appropriate number of specific instances that should be evaluated in order to have a meaningful result, and it should randomly select those instances. The government should then have those instances evaluated and calculate a loss amount based on the review of those random selections.{10} 10  https://www.cms.gov/regulations-and-guidance/guidance/manuals/downloads/pim83c08.pdf

If the government is basing its loss calculation on the patients it happened to interview in a healthcare fraud case, this is not valid because the government did not randomly select the patients. The government chose to interview patients based on criteria the government thought was meaningful or based on complaints by the patients. At least one judge rejected the government’s attempt to prove a “loss” amount in a restitution case in part because the patients were not randomly selected and because the number of patients interviewed was low in context. 11  United States v. Sorensen, 19 CR 745-1, oral ruling on March 20, 2024. 

And if the government is basing its loss calculation on what it says is a large enough number of interviews or instances, this probably is not valid unless a statistician said it was enough.

Other ways may exist to use data analysis to push back on inflated government loss calculations. Look at the methodology that the government used and try to come up with alternative calculations that might be more reasonable.

Many healthcare fraud cases involve “upcoding,” which means a particular service was billed at a higher level of reimbursement than appropriate. The loss here should not be the total payments associated with the upcoded bills, but the difference between the upcoded bills and what should have been billed. Properly calculating loss can be especially important in cases where someone upcoded the bills without the client’s knowledge or involvement. This can significantly increase the government’s loss amount, but defense counsel can argue that this differential should not be attributed to the client.

VIII.  Conclusion

Defense lawyers are practicing in an era of big data, and the government probably uses more summary charts now than it did a few decades ago. But just as the government can rely too heavily on a cooperator whom it took at face value and did not sufficiently corroborate, the government can rely too heavily on data that can backfire.

Counsel used data analysis in one trial to show that the government had presented a misleading view of a doctor’s practice, only showing a relatively small number of times when a government expert disagreed with the doctor. The government tried to counter this by highlighting one piece of data that suggested that the doctor had done an unnecessary procedure on a 92-year-old patient who died the next day.

Jurors gasped. This was the first and only time in the trial that the government had indicated that someone had suffered real physical harm as a result of one of the procedures in the case.

The government had read the data correctly, but the data was incomplete and failed to tell the whole story. The defense team quickly found and presented the full patient file, which provided the full context for the data point. The full story showed that multiple medical professionals had been involved in the patient’s care, that another doctor had recommended the procedure that the defendant had done, and that the patient’s death was due to other conditions and factors.

Defense attorneys should be skeptical of the government’s use of data, just like they would test any government witness for weaknesses and inconsistencies.

NOTE: This is an online version of an article that I originally wrote for the National Association of Criminal Defense Lawyers’ Champion magazine. Please contact me if you’d like a PDF version which has more graphics and footnotes.

Next
Next

Truth-Default Theory and White-Collar Criminal Cases