STB orders UP-NS to make merger employee-impact data public by 27 July

The Analysis

The order was issued on 22 July 2026 and it gave Union Pacific and Norfolk Southern five days. By 27 July 2026, the employee impact exhibits that had been filed as highly confidential had to be redesignated as public and filed with the Board. Five days is where I would start, because a compliance window that short tells you how the Board viewed the request before it even reaches the reasoning.

Underneath that sits a longer timeline, and the timeline is the part worth holding on to. On 28 May 2026 the Board held the merger proceeding in abeyance, in Docket No. FD 36873, pending further order of the Board, while it sought supplemental information from the applicants. On 2 July 2026 several labour unions filed a joint motion asking the Board to order the redesignation, on the argument that the applicants had designated the material improperly as highly confidential. The Board granted that motion on 22 July, finding that this type of information has been publicly disclosed in prior Board proceedings and does not appear to be commercially or competitively sensitive.

My interest here is not the labour story. It is the record. A document that states how many craft and management positions would be abolished, created or transferred, and where the transfers would land, is the closest thing to a public map of where service capacity is going to change. Whether I agree with the Board's reasoning is beside the point. The reasoning is what governs the parties, and the material is now public.

Strictly speaking, this is an interlocutory order inside an adjudicatory proceeding, not a rule of general application. It binds the applicants. It does not itself prescribe service levels, it does not approve or reject the merger, and it does not decide anything about the conditions the Board may eventually attach. I say that at the top because there is a habit of reading anything an agency issues as though it were policy, and this one is not.

What the order actually decides is narrow: the confidentiality designation of certain exhibits filed in the appendices to the revised application. Read narrowly, it is a ruling about who gets to see what. Read for what it reveals about the Board's posture, it is broader, and here I want to quote the shape rather than the words. The Board said it is committed to ensuring that its proceedings remain open to meaningful participation by all interested persons, and it noted that the applicants may take the opportunity to address, as part of their supplemental responses, any concerns about public participation. That sentence is the one a shipper should underline.

The premise of the ruling is worth isolating, because it is the thing that would be tested if anyone tried to extend it. The Board found two things: that information of this type has been publicly disclosed in prior Board proceedings, and that it does not appear to be commercially or competitively sensitive. Both findings are about this type of information in this proceeding. It does not follow that every employee-related exhibit in every rail case will be treated the same way, and I would not advise anyone to assume that the reasoning generalises. Here is the exception that makes the point concrete: an exhibit that combines headcount with customer-specific traffic forecasts, or with pricing, would raise a different question entirely. The ruling I am describing is about position counts and transfer locations.

The proceeding is captioned Union Pacific Corporation and Union Pacific Railroad Company, Control, Norfolk Southern Corporation and Norfolk Southern Railway Company, Docket No. FD 36873. I give the full caption because there is a practical reason behind it. Anything you file, and anything you keep for later use, should carry that docket number. A service log without a docket reference is a business record. The same log, with the docket number and a stated date range attached to it, becomes a record that can be placed into a proceeding, and the distance between those two things is worth more than most of the effort that goes into building them.

Abeyance is the word in this story most likely to be misread, so let me be exact about it. The proceeding has been in abeyance since 28 May 2026. Abeyance is not dismissal, and it is not a suspension of everything attached to the case. The Board said expressly in that 28 May decision that the abeyance did not affect discovery. The formal calendar is paused; the evidentiary process is not. Anyone who concludes from the word abeyance that nothing is happening has read the wrong half of the sentence, and the cost of that misreading is measured in quarters, because evidence built late is evidence that never gets into the record at the right moment.

Sequence matters here, and it is worth holding the three dates in order rather than as scattered facts. Twenty-eighth of May 2026, the abeyance and the request for supplemental information. Second of July 2026, the joint motion from the unions. Twenty-second of July 2026, the grant, with compliance due by the twenty-seventh. Read in that order, the picture is of a Board that paused its calendar without pausing the case, and that resolved a confidentiality challenge within a month of it being filed. A body that moves that quickly on an interlocutory matter is unlikely to be patient later with parties who ask for more time to assemble their evidence.

The five-day compliance window tells you something practical. The exhibits already existed. They were sitting in the appendices to the revised application, written, numbered and complete. Redesignation is an act of relabelling and refiling, not an act of analysis. That is why five days was enough, and it also means the material will appear quickly rather than after a fresh round of drafting. For anyone whose traffic depends on the two networks involved, the reading window opens this week, and there is no reason to expect a delay.

What the exhibits contain, according to the Board: detailed information on how the proposed transaction would affect craft and management employees of both carriers, including the numbers of positions that would be abolished, created, or designated as transferred because of the proposed merger, and the job transfer locations for certain positions. Note the qualifier. Transfer locations are given for certain positions, not for all of them. Do not read the exhibits as a complete census of every facility, because on the Board's own description they are not that.

Now the burden point, and this is the paragraph I would want a shipper to take away. In a proceeding of this kind, an argument gets weight roughly in proportion to the evidence standing behind it. A shipper who writes to the Board saying that service at a particular terminal will degrade, and attaches twelve months of measured transit times for that terminal, is making a different argument from one who writes the same sentence unsupported. I am not in a position to state the formal evidentiary standard that governs a given submission here, and I am not going to invent one. What I am confident about is the practical asymmetry: dated, measurable records get heard, and general apprehension does not.

It is equally worth saying what these exhibits are unlikely to tell you, and saying it plainly. Position counts and transfer locations describe people. They do not describe throughput, they do not describe how many trains a terminal will switch in a day after the transaction closes, and they do not describe transit times on your lanes. There is a strong temptation to read a reduction in craft positions at a named yard as a direct forecast of dwell on your own traffic. Strictly speaking that is an inference, not a fact in the record. It is a reasonable inference, and it is exactly the kind of inference that turns into a decent argument once it is paired with your own measured data. Standing alone it is a guess with a footnote attached.

There is also a distinction between public and usable that tends to get lost the moment a document is redesignated. The exhibits will sit in the docket and anyone can retrieve them. Whether you can do anything with them depends on whether you hold the traffic data to put beside them. A table of transfer locations, read without a per-lane volume map, produces a list of place names. The same table read next to a lane file showing how many of your containers pass through each of those places produces an exposure ranking, and an exposure ranking is the only form in which this information changes what anybody does on a Monday morning.

There is an exception to the value of the exhibits too, and it matters. These are the applicants' projections. They are not Board findings, and the Board has not endorsed the numbers by ordering them into the public record. A projection of how many positions would be abolished is a statement of intent filed by a party with an interest in how it is read. Treat it as evidence about the applicant's own plan, which is genuinely useful, and do not treat it as a determination of what will happen.

So what should be assembled, given that the material is public and the proceeding is in abeyance but discovery is running? A per-lane transit time series with dates, maintained by lane rather than by quarter, because a quarterly average hides exactly the variance that becomes the argument. Terminal dwell measurements, taken the same way every time so that the series is comparable to itself. Every car supply notice and embargo notice received, with the date received and the traffic affected. A service failure log tied to specific shipments rather than to impressions. Contracts with service commitments and the remedy language in them, so that a degradation can be expressed as a number rather than as a complaint. And dated copies of every Board filing relied on, with the docket number written on them, because a filing without its docket reference is difficult to cite later.

The horizon deserves to be stated plainly, because this is where shippers consistently misjudge the clock. A major rail merger, once applied for, runs on a calendar measured in quarters, and the service consequences of crew and facility consolidation show up over years after the decision, not in the quarter it is announced. Building the baseline now is cheap. Trying to reconstruct a baseline in three years, when the argument has become concrete and somebody needs numbers, is expensive and often impossible.

Let me do the arithmetic, with every assumption written out, because abstract worry is not a planning tool. Assume the exhibits, once published, show that a classification yard through which forty of your containers pass each month loses fifteen of its sixty craft positions, which is a twenty-five per cent reduction. Assume the remaining crew absorbs the same volume and that the measurable effect on your traffic is an additional twelve hours of dwell at that node. Assume each container carries eighteen thousand dollars of goods, so the lane carries seven hundred and twenty thousand dollars of value in transit each month. Assume inventory carrying cost of twenty per cent a year.

Twelve hours is half a day. To protect service against a half-day of additional and, importantly, more variable dwell, assume you hold one additional day of buffer inventory in that lane. One day of lane value is seven hundred and twenty thousand divided by thirty, which is twenty-four thousand dollars, and carrying that at twenty per cent a year costs about four thousand eight hundred dollars a year. Against that, assume a service failure rate of two per cent of shipments missing a delivery window, which on forty containers a month is roughly point eight shipments, and assume each failure costs fifteen hundred dollars in expediting and contractual penalties. That is about fourteen thousand four hundred dollars a year. A buffer costing four thousand eight hundred against fourteen thousand four hundred of avoided failure cost is worth holding, on these assumptions.

Run the sensitivity, because the conclusion should not rest on one set of inputs. If the exhibits show five positions rather than fifteen, an eight per cent reduction rather than twenty-five, assume the dwell effect is four hours rather than twelve, and one third of a day of buffer is needed. That is eight thousand dollars of inventory and about one thousand six hundred dollars a year of carrying cost, still well below fourteen thousand four hundred. The buffer remains justified. Now take a lane carrying seventy-two thousand dollars a month instead of seven hundred and twenty thousand: the buffer is two thousand four hundred dollars and the carrying cost is about four hundred and eighty dollars a year. Still positive in principle, but at that size the administrative cost of maintaining and monitoring a separate buffer may exceed the benefit, and that is a judgement the numbers cannot make for you.

The pitfalls, collected in one place, because they are predictable. Assuming that abeyance means the case is dormant, when the Board said discovery was unaffected. Assuming that the published position numbers are Board findings, when they are the applicants' projections. Assuming that a listed transfer location implies continuity of service at that location, when a transfer is a statement about people and not about throughput. Assuming that you can wait for the final decision and gather evidence then. Assuming that a five-day filing deadline means the data will arrive late. Assuming that a confidentiality ruling in one docket settles the question in another. Assuming that a terminal named in the exhibits is the only terminal exposed, when the exhibits describe certain positions rather than all of them.

The Board leaned on prior proceedings, and that is the part of the reasoning with the longest shelf life. A finding that information of a particular type has been publicly disclosed before is a finding about the Board's own practice, and practice accumulates. What that means for a shipper is modest but real. The category of material that can be kept out of public view in a rail merger case is not settled by what the applicants assert about it; it is settled by what the Board has been willing to let into the open previously. So the useful habit is not to wait for the material you would like to see. It is to watch what gets redesignated and to read the docket on the assumption that more of it becomes public over time rather than less.

One more reading habit, because it costs nothing and it pays. When an order contains a sentence saying a party may take the opportunity to address something as part of a future filing, that sentence is an invitation, and invitations in agency decisions are rarely accidental. The Board did not order the applicants to do anything about public participation. It noted that they may address concerns about it in their supplemental responses. That is not a requirement and I am not going to read it as one. But it does tell you what the Board expects to see addressed, and a shipper who puts a concrete, numbered concern in front of the applicants while that window is open is using the procedure as it was described rather than as it was hoped for.

On participation, I will stay on the procedural side and not the strategic side. The Board has said it wants proceedings open to meaningful participation by interested persons, and it has said the applicants may address concerns about public participation in their supplemental responses. What meaningful participation requires in practice, what form a submission takes, and what weight it carries are questions governed by the Board's regulations and by the orders issued in this docket. I have not read the regulations governing this proceeding and I am not going to paraphrase them. Read them before you write anything, and read the procedural orders in FD 36873 alongside them.

What I am comfortable saying is about records, because records are the part entirely within your control and the part that survives a change in the calendar. Keep a dated copy of the 22 July 2026 decision and of the 28 May 2026 decision, with the docket number on each. Keep the redesigned exhibits as filed, with the date you retrieved them, and keep a short internal note naming every facility in them that your traffic touches. Keep the transit time series, the dwell series and the service failure log running from a fixed start date so that the series has a defensible beginning rather than one chosen after the fact.

Keep the reasoning too. If a lane is judged exposed, write down why, with the numbers used and the date. If a lane is judged not exposed, write that down as well, because a decision not to act is also a decision that may later need explaining. Twelve months from now the exhibits will have been superseded by supplemental material, the numbers will have moved, and the only thing that will show the judgement was reasonable when taken is the note written on the day it was taken.

On how long to keep all of this, I would reason from the horizon rather than from a rule I cannot cite. The service consequences of a major rail combination, if it is approved and if conditions are attached to it, play out over years. A transit time series that starts in August 2026 and stops in December 2026 tells you nothing about a change that only becomes visible in 2028. So the discipline is a fixed start date and no planned end date, with the series maintained monthly for as long as the lanes exist. The cheap part is continuing. The expensive part is restarting.

My closing position is narrow again, and I am content with narrow. The order decides a confidentiality question and nothing more. It does not tell you the merger will be approved, and it does not tell you the service picture. What it does is put a document into the public record that describes where people will move, and it does so on a five-day clock, while a proceeding sits in abeyance with discovery running. Read the exhibits when they appear. Match them against your own lanes with your own numbers. Build the baseline now, while the argument is still abstract, because the quarter in which it becomes concrete is the quarter in which nobody has time to build anything.

↑ Back to the key points

— By Dr. Ingrid Voss