Another Case Of Inaccurate Container Shipping Data

Descartes, the large Canadian technology provider and data company, is issuing year over year volume growth figures for inbound containers coming into the US that are notably different compared to my Top 10 reports. My reports are based on the actual TEU counts provided by the ports while Descartes numbers are likely based on incomplete customs data that is then adjusted. I stand by my reports as representing the correct actual growth rates of inbound containers into the US. Actual container volume data can provide a real time pulse of economic activity. However, it first needs to be accurate before it can provide meaningful insight.

For June 2024, my numbers for the Top 10 US ports show that inbound volume increased 17.2% compared to June 2023. Descartes said that inbound volume in June 2024 increased 10.4% versus the previous year, a difference of 6.8% or 40% below my number. The standard deviation between the respective measures in 2Q24 was 5.1%, an increase from the 3.2% standard deviation in 1Q24. Over the last 18 months the standard deviation has been 2.9%. The larger June difference is a material understatement and extends on a pattern I wrote about a few months ago in my March container volume report. That report noted two straight months in February and March of understatement, both not nearly as wide as the latest understatement in June. That was addressed in the paragraph below from the March report shown in italics. It was followed by the next paragraph in the same report that referenced my concerns about other actual data integrity issues related to the container shipping sector.

My monthly reports are only issued after I’m satisfied that they accurately reflect the volume trends at US ports. While there are 33 ports with regular container activity, the Top 10 represent 86% of total volume and my analysis shows they consistently reflect the same changes as the large ports. An information provider based in Canada issues monthly reports on US container volume before my report comes out that are generally reported in the press. However, I do not believe they are as accurate as my reports. For instance, for February they showed inbound volume was up 23.3% or 3.2 percentage points below my 26.5% figure. The absolute and relative difference expanded to 3.5 percentage points in March when they showed 15.7% versus my 19.2%. They also tend to focus on unadjusted sequential changes that are less meaningful given number of days and seasonal differences.

This current data issue is reminiscent of an article I wrote in February 2023 detailing my concern regarding the accuracy of container shipping profit figures issued by a UK based information provider along with the prevalence of using spot indices rather than better overall measures of actual sector pricing. That CMS article can be accessed using the following link: https://centerformaritimestrategy.org/publications/time-to-recalibrate-part-one/. There will be different views on what future sector numbers will look like but there should be no differences on what has actually already occurred in the broad container sector. It is too important a part of our economy for anyone, particularly policy makers, to be influenced by materially wrong numbers. The value of the cargo that went through the Top 10 US ports was $444 billion in the first quarter of 2024 and $1.733 trillion over the last twelve months. Changes related to that have a pronounced effect on the US economy. I believe that information related to actual volume and pricing trends within the container shipping sector is so important that consideration should be given to the federal government establishing a clearinghouse containing such data. Detailed actual information like that provided by the Energy Information Agency on energy would be useful to many constituencies. In September 2022 in response to a public request for comments on information gathering, I submitted comments to the Federal Maritime Commission outlining such an approach. A publicly available dataset such as that housed at the FMC would in my view be a constructive addition to its mission. My submission can be accessed using the following link: https://medium.com/@john-d-mccown/submission-in-response-to-fmc-requests-for-comments-a4c1867e49c9.

The Descartes figures purport to be all inbound TEU’s coming into the US. My figures are based on the actual TEU counts as released by the Top 10 ports, which consistently are around 86% of the total inbound volume at all 33 US ports handling containers. I routinely check that relationship by comparing my Top 10 volume to overall inbound customs data weeks after the end of each month so it is fully populated. The customs data, however, has both timing and credibility issues and is not as accurate as the loaded TEU counts disclosed monthly by the largest ports. My Top 10 report is an accumulation of the actual count data provided by the ports.

Below is a worksheet comparing my year over year growth rates to Descartes over the last 18 months. The standard deviation over that entire 18-month period was 2.9%. While that is a noteworthy difference over the entire year and one half period, the standard deviation increased to 4.3% over the last six months and increased further to 5.1% over the last three months. The smaller ports not included in my Top 10 cannot be responsible for the differences with the Descartes reports. Those smaller ports, with a larger mix of East/Gulf Coast that among the Top 10, will show some coastal alignment but my testing has consistently shown that the year over year growth figures of the Top 10 ports and all 33 ports are virtually the same. Attached at the end of this article are additional comparisons including aggregate inbound TEU’s, three-month rolling averages and sequential comparisons of the last two months of each report. (Note that the average measures in the table below are simple averages of the actual monthly figures.)

As shown in the worksheet, there has been a pronounced upward trend in the standard deviation over the last few quarters. A key comparison is the quarterly trend and the widening differences over the last three quarters. During 4Q23, the Descartes data showed a monthly average growth that was 0.5% higher than the monthly average of my Top 10 report. In 1Q24, that flipped to showing the Descartes data averaging 2.2% lower compared to my data. The average difference widened further to 2.4% in 2Q24. The even more relevant standard deviations had a sharper upward slope and went from 0.5% to 3.2% to 5.1% over those same three quarters. The latest actual quarterly standard deviation is half what the Descartes reports are claiming, making them extremely inaccurate representations of what has actually happened.

I’ll later outline the factors that I believe are in play that result in the striking differences in year over year growth of inbound containers coming into the US. I emphatically stand by the accuracy of my numbers.

I first became aware of the monthly Descartes reports sometime in 2023. Prior to coming across summaries of their reports in various publications, summaries of my Top 10 reports were regularly covered in gCaptain, Freightwaves, Seatrade, Splash and Lodestar among maritime publications. They were also occasionally referenced in the Wall Street Journal and Bloomberg. However, with the Descartes reports coming out much sooner, often a week ore earlier after the end of each month, they quickly eclipsed the availability of my reports that circle around coming out two weeks or more after the end of each month. When a publication runs a story on what Descartes says is going on with all inbound containers into the US, they will have no interest in running a story a weeks later on what my report says is happening with the Top 10 reports. As articles referencing Descartes reports increased and ones referencing my reports virtually ended, I concluded that they must have had a process allowing them to publish more complete numbers well ahead of me. I now recognize my assumption that they were accurate numbers was misplaced. Being displaced in publications by what I now believe are bad numbers was unfortunate.

In researching this situation on the internet, the earliest Descartes monthly container volume report I could find was from March 2022. Apparently the reports are a result of data that is available on Descartes Datamyne, which is a subscriber-based service. The beginning of Descartes publicly issuing these monthly reports broadly coincided with the peak congestion West Coast ports were experiencing and the increase in reporting on maritime supply chain issues. My own reports had been going out to an email distribution list since 2017. From May 2020 they were going out via Constant Contact email software, which provides data on recipients accessing the reports. That first Constant Contact email went to 308 addresses and had a 37% open rate and by the end of 2021 it went to 984 addresses with a 50% open rate. Driven by interest in supply chain issues resulting from the pandemic, the subscriber base continued to grow to 1,719 addresses at the end of 2022 and 2,281 at the end of 2023 with open rates always above 50% and as high as 64%. The June 2024 email went to 2,436 addresses and had a 54% overall open rate. People who opened the June email did so an average of 1.96 times, indicating the report was typically revisited.

The Descartes report must be based on bill of lading data filed with US Customs as their reports come out prior to any port releasing its actual inbound TEU count. The monthly data begins to be accessible right after the end of the month and continuously populates from some 60,000 bills of lading filed each day. I suspect that Descartes chooses the same number of days after each month to download the available TEU data by port. While too early to have a full data set, they then likely adjust that data to get what they purport to be the total inbound TEU’s to the US. That adjustment factor would be based on research of the historical relationship between the TEU’s when downloaded at the same time and the ultimate TEU’s when the database is typically fully populated some three weeks after the end of the month.

It’s a logical approach and I imagine they believe that by using the same framework and consistently downloading the same number of days after each month, they are measuring growth accurately. But as evidenced by the worksheet comparisons, that is actually not the case and what Descartes is showing as actual growth rates are materially wrong. From my operating experience, I believe both the filing of bills of lading and the related processing is subject to a myriad of variables that change month to month. Any adjustment factors used by Descartes are not accurate because they are being applied to an ever-shifting base number. There is also a less than subtle timing difference in the customs data as all the dates which drive what month the load is in are based on vessel arrival. For vessels being unloaded in consecutive days over month end, that will be a difference compared to actual TEU counts by the port that are based on the actual day when the container was unloaded.

Descartes must know from my reports that their overall growth figures are inaccurate. That inaccuracy is I believe driven by the desire to get their report out early to maximize press coverage. That would include getting it out a week or two before my report. Absent their timing advantage, they would have the same publishing displacement issue that I have experienced. The need for speed should never trump data integrity, but I believe that is exactly what is going on in with this situation.

The author of Descartes monthly reports is their EVP of Industry and Services. The reason I believe that they are aware of inaccuracies in their reports is that this gentleman has been getting my reports since November 2022. My email software indicates that of the 36 reports in emails sent to him from then until July 18, he opened 97% of those emails. Most of the emails showed multiple openings at different times. After opening three separate times my June report, the one with the largest difference ever, that gentleman unsubscribed from my reports on July 18. He was likely aware before me of the growing divergence in our respective year over year growth rates. He also must have been aware of the paragraph above in my March report that softly criticized his company without naming it, as he opened that report six times over three different days. While my contact information was readily available in the reports he followed, neither he nor anyone at Descartes ever contacted me to offer either criticism or praise of my reports.

The aggregate inbound TEU’s in the tables accompanying the Descartes reports are represented as actual figures when in fact I believe they are estimates derived from a flawed process. Rather than correcting their numbers when more reliable data is available, they maintain that historical data so those year over year growth figures are based on a currently flawed number compared to a prior flawed number. If there was consistency in the flaws, the outcome wouldn’t be as bad but the comparisons here demonstrate that is not the case. The differences as can be seen in the accompanying tables are a random walk. They bounce all over the place compared to the definitive numbers represented by the actual inbound TEU’s reported by the Top 10 ports.

While the Descartes report show aggregate inbound TEU’s, it does not show such information for individual ports. The only data shown on individual ports is a sequential comparison of the current month to the prior month. While sequential comparisons benefit from data recency that is roundly outweighed by factual issues related to varying number of days in the month and seasonality. The former injects 3–10% differences and the actual seasonality differences often make the sequential comparison even less meaningful. These differences make such unadjusted comparisons less meaningful even if they were based on accurate data, which we now know the Descartes reports don’t have.

Confirmation of the inaccuracy of Descartes sequential comparison can be seen in the sharp double digit percent differences in all but two of the ten ports in the last table below. For example, note what Descartes claims was the change from May to June at Long Beach in inbound TEU’s. Their most recent report says the increase was 7,071 TEU’s. The actual data released by the port of Long Beach for those two months shows the actual increase was 74,427 TEU’s, a number that was more than ten times what Descartes claimed.

The Descartes reports neither shows the aggregate inbound TEU’s by major port nor do they disclose the year over year growth percentages by port. Either of those measures would make the inaccuracies in the Descartes data more detectable and I imagine that is precisely why that information is not disclosed. The ports focus on year over year growth and very few mention sequential difference because that is not viewed as particularly relevant for the earlier reasons given. The discussion of individual ports in the Descartes reports is limited to the less relevant sequential comparisons that have the effect of shielding the inaccuracies in their underlying data.

This situation is an issue of actual data integrity. As evident in the tables, Descartes numbers on inbound container growth are materially inaccurate. Unfortunately, they are not the only example in the container shipping sector of incorrect numbers that purport to be what has actually happened. The Center of Maritime Strategy article referenced earlier noted inaccurate numbers on the actual earnings of the container shipping industry and the mischaracterization of industry pricing levels and trends.

As the maritime supply chain issues during the pandemic have clearly demonstrated, what is occurring in the container shipping sector plays a significant and more recognized role in the US economy. As such, likely future directions in the sector are being modeled more by a variety of private and public parties. Developing those views comes with its own difficulties and complexities, but one should not be related to bad numbers on what has already actually happened. Navigating the future starts with the best factual information of where you actually are now.

Forecasts are one thing and rely on various algorithms and assumptions and are expected to be different. But actuals are a different thing and should be based on actual actuals.

John D. McCown is a shipping expert with over four decades of operating and investing experience, primarily in the container shipping sector. That experience includes CEO of a U.S. flag container shipping company he co-founded and managing worldwide transport investments at a $20 billion hedge fund. He currently advises lectures and writes about maritime issues that interest him. Mr. McCown worked on a daily basis for more than twenty years with Malcom McLean, the inventor of container shipping, and holds an MBA from Harvard Business School as well as two maritime related patents.