July 18, 2022
This is part of a series of articles about the online auto auction platform, Bring a Trailer. For my other posts about Bring a Trailer, click here.
Beginning on March 14, 2021, Bring a Trailer began having auctions end seven days a week. Previously, auctions only closed on weekdays. Since this expansion, I’ve heard BaT sellers, bidders and watchers speculate about whether auctions ending on weekends were likely to be more or less successful for the seller than weekday auctions. In this article, I will explore whether weekend endings influence either hammer prices or sell-through rates. Like in my previous articles about BaT, I will examine these questions using actual auction data, statistics and some visualizations.
For a detailed discussion of my data-gathering methodology, see part 1. For this article, I gathered data from 500 randomly selected Bring a Trailer auctions that closed between 1/1/22 (lot 62623) and 7/13/22 (lot 78591). After collecting this data, I wrote additional scripts for data transformation, analysis and visualization. I used pandas for data transformation and exploratory analysis. I also used statsmodels for some statistical tests. The plots in this article were created using plotly for Python.
Within our sample of 500 auctions, 359 ended on a weekday and 141 ended on a weekend. The table below shows how these breakdown by day of the week.
Table 3.1: Count of auctions in sample by closing day of week
Hammer prices #
For background on the overall distribution of hammer prices on BaT, check out part 2. Let’s compare closing prices when grouped by weekend and weekday closing dates.
This plot doesn’t show any immediately obvious relationship between weekends and hammer prices. The shape of the two distributions are similar to each other. They both show a strong positive skew, as we also saw in the overall distribution from part 2.
In order to determine if there is a statistically significant relationship between hammer price and weekends, we can use a two sample t-test. Running this test on our data provides a p-value of 0.355. Since this value is larger than the standard significance value of 0.05, we cannot reject the null hypothesis that there is no difference between weekend and weekday hammer prices. In plain terms, our statistical test shows the data does not provide evidence for a difference in hammer prices between weekends and weekdays.
What about other days of the week? Does the specific day of the week have an influence on closing prices?
Again, there’s no obvious relationship apparent in this plot. Let’s do another statistical test to check if the specific day of week has a significant effect. Since we have more than two groups, we will use a one-way ANOVA. Similarly to the t-test above, this provides a p-value that indicates a statistically significant difference between any of the groups (days of the week) and the mean. After running the ANOVA, we get a p-value of 0.192. Since the p-value is greater than 0.05 we cannot reject the null hypothesis that there is no difference in hammer prices on specific day of the week.
These results fail to show evidence of a relationship between the day of the week and hammer prices.
Sell-through rate #
For background on BaT sell-through rates, check out part 1.
Of the 500 auctions in this sample, 352 had a reserve. BaT achieved a 75% sell through rate for reserve auctions in this sample.
Table 3.2: Count of sold lots and sell-through rate for reserve lots in sample.
|Reserve not met||87||25%|
|Total reserve lots in sample||352|
If we group these reserve lots into weekend and weekday closing, we can then compare the sell-through rates for each category.
Table 3.3: Count of sold lots and sell-through rate for reserve lots in sample, grouped by weekend/weekday
The sell-through rate for weekend lots is 11% higher in our sample. This seems like a substantial difference, but is it statistically significant? What statistical test can we use to determine whether weekend closing dates affect the chance of a reserve lot selling? I used a two-sample proportions test for this purpose. Like the two-sample T-test used above, this test compares two groups and provides a p-value. Unlike the T-test, the proportions test compares rates of occurrence, instead of group means. This makes it appropriate for testing whether the sell-through rate differs between the two groups. I ran a two sample proportions test based on normal approximation (z-test), which produced a p-value of 0.03 (Z: 2.11). Since the p-value is less than 0.05, this result is statistically significant. Therefore, we can reject the null hypothesis that there is no difference in sell-through rates between weekend and weekday auctions.
Our proportions test shows that the difference in sell-through rate between weekend and weekday lots is statistically significant. This data provides convincing evidence that auctions ending on the weekend have a higher sell-through rate.
- Our data does not show a statistically significant relationship between auctions ending on weekends/weekdays and hammer prices.
- We also do not see a statistically significant relationship between the specific day of the week and the hammer price.
- The sell-through rate for auctions ending on a weekend was 83%, compared to 72% for weekday auctions. A two sample proportions test demonstrates that this difference is statistically significant.
When selling with Bring a Trailer, the starting and closing dates of the auction are assigned by BaT. As far as I know, there is no way for the seller to choose the specific date or day of the week when an auction is closed. It is important for Bring a Trailer to maintain an appearance of fairness and good auction participation regardless of day. It’s easy to imagine seller discontent arising from the idea that their auction was scheduled on a “bad” day. Other auction platforms like eBay sidestep this issue by giving sellers control over their auction dates.
If you frequently read auction comments on BaT or other venues, it’s often speculated that the day of the week influences the result. Anecdotally, many of these commenters believe that a weekend ending will hurt the auction. This belief may derive from the fact that weekday endings used to be the only option. As weekend endings are newer, some believe that some proportion of the audience is unaware of this and therefore auction participation will be reduced. This argument has become less common over time, as BaT has been closing auctions on weekends for more than a year. Other arguments against weekend auctions also tend to focus on lower bidder participation due to weekend leisure activities or other factors drawing potential bidders away from their web browser.
While it would be very difficult to assess the mechanisms influencing bidder participation across a typical week, my analysis lets us look at the effect of the day on the two metrics most important to sellers: hammer price and sell-through rate. While there is no evidence for an effect on closing prices, we do see that weekends seem to have a positive effect on sell-through rate. Our data shows reserve auctions are 11% more likely to meet reserve on the weekend, compared to the weekday.
This analysis does not explain the mechanism causing sell-through rates to be higher on the weekend. Explanations involving bidder behavior may seem likely, such as increased leisure time during the weekend enabling more auction participation. But we should also remember that BaT can influence sell-through rates and closing prices. They set reserve prices and could potentially factor day of week into that decision. I have no reason to believe this, but our data does not rule it out.