Using daily retail gasoline prices at individual stations, we developed statistical measures of the movements of these prices across brands and locations. In this anti-trust action, we showed that day-to-day price changes demonstrated a lack of coordinated retail pricing across outlets. While prices moved together when there were common market-wide cost changes, station-specific price changes for brands were not associated with one another.

Using accident records for first-time buyers of automobile insurance we helped justify the pricing of these policies as an accurate reflection of expected claims. Newly implemented state mandated auto insurance required previously un-insured motorists to purchase insurance prior to registering their vehicles or obtaining a drivers license. Motorists without prior insurance were initially classified into high risk pools and were charged high rates for insurance. State agencies then required that the insurers place these drivers into standard risk pools and to track their driving and accident histories. Welch Consulting was retained by an insurance carrier to review data on insurance claims for these newly insured drivers. We compared their claim history to that of comparable drivers who had a prior history of insurance coverage. After controlling for a variety of factors relating to their claims risk we found that those with “no prior” insurance had significantly greater claims which warranted their inclusion in a higher risk pool.

In a settlement of a defective product matter, three producers initially split claims payments in proportion to their market share. Welch Consulting was asked to audit the actual claims and to determine if the initial division of responsibility was consistent with paid claims.   We found that the claims were not proportional to gross market share and that a redistribution of cost was warranted.

Medicare service providers can be subject to a government audit when their billing practices stand out in some way compared to an auditor’s expected norm. Once flagged, an auditor, working on behalf of Medicare, requests an inventory of files and from this list draws a sample for review. The sample is reviewed for “exceptions” that may lead to a required reimbursement to Medicare. More importantly, the findings from the sample are then extrapolated to all of the provider’s Medicare billings during the review period.  Welch Consulting was retained by attorneys for a provider to assess the statistical validity of the auditor’s sample. In our report to the court, we showed that the sample was not drawn randomly, was too small to render meaningful results and was not representative of the broader inventory of cases that were subject to extrapolation. The court subsequently disallowed the extrapolation of reimbursements to the broader universe of files.

In a case alleging racial profiling Welch Consulting analyzed stop and arrest data for a law enforcement agency. We broke down this data into demographic groups by race, age and gender. We showed that the arrest rates were approximately the same for those stopped and searched, regardless of their demographic group. This result was consistent with a demographically neutral practice that resulted in the same efficacy of stops for all groups.

Employees working after the store had closed to the public were prevented from exiting the store until the manager unlocked the exit and allowed them to leave. Class action claims included uncompensated work time between clocking out and actually exiting the store. Rather than rely exclusively on deposition testimony, Welch Consulting used security video to help determine the actual wait time from clock-out to exit. We cataloged the universe of security video and developed a random sample of time increments during which employee exits were observed.  We developed a protocol for tracking employee exits and a observation log for video reviewers to record standardized events. Exit events were monitored by two reviewers independently to account for potential reviewer error. Statistical results showed that the delay time was minimal and far less on average than reported in testimony.