Uncited Press › Star Wars Open the interactive journal →
Wars · Political Economy & Social Structures

Droids in the Freight Yards: Organic Employment, Paid Hours and Wages in 24 Coruscant Cargo-Handling Yards as Labour Droids Arrived, 30–24 BBY

Dr. Linnea Veyre1
1 Coruscant Academy of Letters and Sciences, Faculty of Social Sciences
Received 21 Sep 2026 · Revised 22 Oct 2026 · Accepted 1 Nov 2026

Abstract

Labour droids entered Coruscant's cargo-handling yards at different dates, and whether they displaced organic workers, moved them into other work or mainly cut their hours has not been measured. We assembled quarterly payroll ledgers and plant registers for 24 yards over 26 quarters, 30 to 24 BBY. Twenty-one yards adopted droids and three had not by the close of the record. Each adopting yard was compared with yards that had not yet adopted, using stacked adoption cohorts with yard and quarter effects, Poisson models for headcounts and linear models for log hours and log pay, with intervals on 23 degrees of freedom. After adoption, organic headcount was unchanged (rate ratio 0.996, 95% CI 0.951–1.042). Loading posts fell by 9.8% (5.2–14.2) and supervisory and repair posts rose by 70% (40–106). Paid hours per worker fell by 9.9% (6.8–12.8), hourly pay rose by 2.9% (0.6–5.3), though not detectably within either grade, and quarterly earnings fell by 7.2% (3.7–10.6). Hours carry 96% (log scale) of the fall in total paid hours. Hours losses did not differ detectably by species, but loading posts held by Wookiees fell by 22.9% (12.7–31.9) against 4.9% (a fall of 10.6% to a rise of 1.2%) for Humans, a contrast with p = .047 after adjustment for ten comparisons, conditional on the choice of outcome, and Humans took 72% of the added supervisory and repair posts. In these yards droid adoption worked through hours and through a change of grade, and the loss of posts fell unevenly across species.

1. Introduction

We find that the arrival of labour droids in Coruscant's cargo-handling yards left the number of organic workers where it was and took about a tenth of their paid hours instead. In 24 yards observed quarter by quarter from 30 to 24 BBY, the loading posts that droids made unnecessary were replaced almost one for one by supervisory and repair posts, and the remaining loss of work fell on the hours of those who stayed. The loss of loading posts was not shared evenly. Posts held by Wookiees fell by 22.9%, those held by Humans by 4.9%, and Humans filled most of the new posts.

Goods reach Coruscant's surface and its many levels through yards that tranship cargo between landing platforms, lifts and surface haulers. The work is physical and organised in gangs, and the yards employ a mixture of species, engaged by the day. Labour droids are owned property, they draw no wage, and a yard operator may add them without consulting anyone. Official attention has been thin, and the Senate's research office has summarised hours and conditions in the Core cargo trades from yard returns (Galactic Senate Legislative Research Office, 25 BBY), but returns report totals for a yard and say nothing of who works, in which grade, or for how long. Whether droids displaced workers, moved them into supervision and repair, or mainly shortened their hours therefore remains unsettled, and so does the question whether the answer depends on species.

Earlier work supplies three expectations. Port and cargo trades are held to adjust paid hours before they adjust headcount, because a gang can be sent home early at no cost to the yard's register (Maldrin, 29 BBY). Mechanical handling is held to reallocate tasks upward into supervision and maintenance (Mallowen, 27 BBY). Muster rolls show that species are unevenly spread across cargo grades, with some concentrated in the heaviest lifting (Mervane, 28 BBY). None of these studies observed droid adoption yard by yard against a payroll, and the three expectations have not been tested against one another.

This paper takes the vantage of 24 BBY. It uses the payroll ledgers and plant registers of 24 yards, closed at the end of the second quarter of that year, and compares each yard that adopted droids with yards that had not yet done so.

2. Methods

Sources. The sources are the quarterly payroll ledgers kept by yard paymasters. For each organic worker on a yard's rolls in a quarter, the ledger enters the grade, the species as the paymaster recorded it, the paid hours and the pay. Droids never appear on a payroll, so their presence is read from the plant register that each yard keeps of equipment in service, which lists a labour droid from the quarter it is first put to work (Neskarin, 25 BBY). We approached 31 yards and used the 24 whose ledgers and registers were complete for all 26 quarters, from the first quarter of 30 BBY to the second quarter of 24 BBY. The other seven yards had gaps, or opened or closed inside the period. The panel is balanced: 24 yards by 26 quarters gives 624 yard-quarters. Copies were consulted with the consent of the yard operators, and the yards are named here only by working labels, Y-01 to Y-24.

Grades and species. The payroll carries two grades in these yards, of which loading covers hand-loading, hauling, slinging and stowage and supervisory and repair covers gang leaders, berth supervisors and the servicing of droids and handling plant. Tally and gate work are paid from a separate office and lie outside the ledgers, so organic headcount in this paper means the sum of the two grades. Among the 21 adopting yards, the mean over the six quarters before adoption was 236.0 workers per yard-quarter, of whom 208.4 were in loading and 27.6 in supervision and repair. Species is entered as Human, Twi'lek, Rodian, Wookiee or other, the last pooling several species too few to estimate singly. Before adoption, loading posts were 46% Human, 18% Twi'lek, 14% Rodian, 12% Wookiee and 10% other, shares that are our own tabulation, and muster rolls kept by guilds (Mervane, 28 BBY) were not used to correct the paymasters' entries.

Adoption. A yard's adoption quarter is the first quarter in which its plant register lists a labour droid in service. Twenty-one yards adopted, between the first quarter of 29 BBY and the last quarter of 25 BBY, and three had not adopted when the record closes. By year of adoption the cohorts hold 3 yards (29 BBY), 4 (28 BBY), 6 (27 BBY), 5 (26 BBY) and 3 (25 BBY). The number of droids and their tasks differ between yards and are not used.

Design. Adoption is staggered, and comparing adopters with all other yards at any one date would set early adopters beside late ones and use already-adopted yards as controls. We therefore stacked the data by year of adoption (Nethlar, 27 BBY). For each cohort we took the treated yards from six quarters before adoption to six after, cut at the ends of the record, and as comparators every yard that had not adopted by the end of the cohort's window, the three never-adopting yards included. A comparator is always a yard that had not yet adopted. The five stacks hold 766 yard-quarter rows, 262 treated and 504 comparator, with between 3 and 14 comparator yards per stack. Each outcome was regressed on an indicator for a treated yard in the quarters from adoption on, with a separate yard effect and a separate quarter effect in every stack. The estimate is the average effect over the post-adoption quarters, and later cohorts contribute fewer of them because the record closes.

Outcomes and models. Headcounts are counts and were fitted by Poisson regression with a log link, so a coefficient is the log of a ratio of expected headcounts. Paid hours per worker-quarter and hourly pay, which is quarterly pay divided by paid hours, are continuous and positive. They were fitted by linear regression of their logarithms on worker-quarter lines, about 180,000 in the stacked data (766 yard-quarter rows of roughly 236 workers each), so a coefficient is close to a proportional change. Quarterly earnings, the product of the two, was fitted the same way. Species-specific estimates come from fitting each model separately within a species group.

Inference. Standard errors are clustered by yard, and with 24 yards intervals and p-values use the t distribution on 23 degrees of freedom (Morwenna, 25 BBY). Differences among the five species were assessed by a heterogeneity statistic: the sum over species of each estimate's squared deviation from the inverse-variance weighted mean, divided by its variance, referred to the chi-square distribution on four degrees of freedom. It treats the five estimates as independent. Because they share comparator yards and quarter effects, they are positively correlated, which makes the statistic, and the pairwise contrasts built the same way, conservative. An event-study version of the design, with an indicator for each of quarters −6 to −2 before adoption and quarter −1 as the reference, was fitted to test for movement before adoption.

3. Results

Table 1 gives the average effects of adoption. Total organic headcount did not change: the rate ratio was 0.996 (95% CI 0.951–1.042, p = .84), a change of −0.4%. Beneath that flat total, the grades moved in opposite directions. Loading headcount fell by 9.8% (rate ratio 0.902, 0.858–0.948), and supervisory and repair headcount rose by 70.1% (1.701, 1.403–2.061). Applied to the pre-adoption means as the reference level, the two ratios give 188.0 loading posts and 47.0 supervisory and repair posts per yard-quarter, a loss of 20.4 posts in one grade and a gain of 19.3 in the other.

Paid hours per worker fell, and across all workers the multiplier was 0.901 (0.872–0.932), or 9.9% fewer paid hours per quarter, and among loading workers it was 0.882 (0.849–0.915), or 11.8% fewer. Hours in supervision and repair did not move detectably (0.985, 0.907–1.070). Hourly pay across all workers rose by 2.9% (1.029, 1.006–1.053, p = .015), but within grade the rise was not detectable: 0.8% in loading (1.008, 0.985–1.031) and 1.2% in supervision and repair (1.012, 0.941–1.088). Weighted by the post-adoption shares of lines in each grade, the two within-grade estimates account for 0.009 of the pooled coefficient of 0.029, and the remaining 0.020, which is a residual and carries no interval of its own, is what the move of workers into the supervisory and repair grade would have to contribute. The within-grade intervals are wide enough (the loading interval reaches 0.031) that they do not exclude a within-grade rise as large as the pooled one. Quarterly earnings fell by 7.2% (0.928, 0.894–0.963).

Multiplying the headcount ratio by the hours multiplier gives a fall of 10.3% in total organic paid hours. Because headcount did not change, the hours margin carries 96% of that fall on the log scale. Within loading alone the two margins are closer: the log fall in loading posts was 0.103 and the log fall in hours per loading worker was 0.126, so hours account for 55% of the fall in loading labour-hours and the loss of posts for 45%.

Table 2 sets the species groups side by side. Paid hours per loading worker fell in every group, from 9.9% among Humans (0.901, 0.863–0.941) to 18.8% among Wookiees (0.812, 0.751–0.879). The heterogeneity statistic for hours was 6.25 on four degrees of freedom (p = .18), so the ledgers do not separate the groups on loading hours lost. The unadjusted Wookiee–Human contrast in hours has p = .025 and, multiplied by ten, .25, so it does not survive adjustment. They do not separate them on hourly pay across all workers either (statistic 2.60, p = .63), though the point estimate is highest for Humans (+4.2%) and lowest for Wookiees (+0.4%).

Loading headcount did differ by species (statistic 10.52, p = .033), and because it was the one outcome of the three to show heterogeneity, the pairwise contrasts were examined for it alone. Wookiee loading posts fell by 22.9% (0.771, 0.681–0.873). Human loading posts fell by 4.9% (0.951, 0.894–1.012), an interval that includes no change. The Twi'lek, Rodian and other groups lay between, at falls of 10.0%, 12.0% and 13.0%, each with an upper limit just under 1 (0.988, 0.980 and 0.991). The ratio of the Wookiee ratio to the Human ratio was 0.811 (0.706–0.931, p = .005). It is one of ten pairwise contrasts among five groups, and multiplied by ten it stands at p = .047. The heterogeneity statistic is one of three, and its p of .033 is not adjusted for that (multiplied by three it is .098), so the tenfold adjustment is conditional on the choice of outcome. The new posts went mostly to Humans. Of the 19.3 supervisory and repair posts added per yard-quarter, Humans took 13.9, or 72%, and their share of such posts rose from 62% to 66%. The Wookiee share fell from 5.8% to 4.2%: a gain of 0.4 posts set against a loss of 5.7 loading posts. Combining the two grades, the net change in organic employment per yard-quarter was +9.2 for Humans (+8.2%), −0.7 for Twi'leks, −2.4 for Rodians, −1.7 for the other group and −5.4 for Wookiees (−20.1%). These figures are descriptive, built from the ratios and the unrounded grade shares, and carry no interval of their own.

Nothing moved before adoption. The five lead coefficients of the event-study version were jointly indistinguishable from zero (F = 0.88 on 5 and 23 degrees of freedom, p = .51).

When the comparators were restricted to the three yards that never adopted, hours per loading worker fell by 11.1% (0.889, 0.827–0.955), close to the main estimate of 11.8%. Loading headcount fell by less in that restricted fit (ratio 0.922, 0.828–1.027) and its interval includes 1, as three comparator yards allow little precision.

4. Discussion

Each of the three questions has its own answer, and the answers differ in size. Taken together, the yards did not displace their organic workforce: headcount after adoption was the same as before, within a few percent either way. Workers did move between grades, with 20.4 loading posts lost and 19.3 supervisory and repair posts gained per yard-quarter. Hours fell for loading workers in every species group and for the workforce as a whole, while hours in supervision and repair did not move detectably. The cut in loading hours is the one change that reached every species group.

Hours carry the main result. Total organic paid hours fell by 10.3% while headcount stood still, so nearly all of that fall (96% on the log scale) passed through hours per worker. Hourly pay averaged 2.9% higher, but within grade the rise is not detectable, and by the point estimates about two-thirds of the pooled rise comes from the larger share of the workforce in supervision and repair, though the intervals do not rule out a within-grade rise. Retained workers received fewer paid hours without a within-grade rise in hourly pay large enough to offset them, and quarterly earnings fell by 7.2%. The expectation that cargo trades adjust hours before heads (Maldrin, 29 BBY) holds in these yards for the labour force as a whole.

A shift between grades is real and large, and it is nearly one for one. A rise of 70% in supervisory and repair posts is a small count in absolute terms because that grade began at 27.6 posts per yard-quarter, but it matched in number the loading posts that ended. Handling plant and droids need continuing servicing by trained staff (Norvik, 26 BBY), and the yards met that need with posts on their own payrolls. The upward reallocation expected from mechanical handling (Mallowen, 27 BBY) is therefore present, though it did not leave the loading gangs whole.

Unevenness by species is the part of the result easiest to miss. A count of heads for the yards as a whole would have recorded no effect at all. Wookiee loading posts fell by 22.9% and Human loading posts by 4.9%, and Humans took 72% of the new supervisory and repair posts. Muster rolls had already placed Wookiees among the heaviest lifting in the Core trades (Mervane, 28 BBY). The ledgers do not record which tasks within loading the droids took over, so this paper does not say why Wookiee posts fell furthest. The ledgers show that they did, by a margin that survives adjustment across the ten pairwise contrasts, conditional on having examined this outcome because it alone showed heterogeneity, and that the new grade went mostly to one species.

Yard returns of the kind the Senate's research office summarises report totals and cannot show either the hours result or the grade result. Records of hours, grade and species by worker are the minimum that registers the effects of droid adoption on a yard's people.

5. Limitations

Adoption quarters were chosen by yard operators. The stacked design removes differences between yards that stay constant and shocks that hit all yards in a quarter, and the lead test finds no movement before adoption. A shock peculiar to one yard that coincides with its adoption, such as a rise in the cargo passing through it, loads onto the estimates, and the ledgers record no throughput to test it. Each estimate is an average over the quarters after adoption, up to six, and later cohorts contribute fewer of them, so nothing is estimated for longer horizons.

Twenty-four yards of the 31 approached form the sample, chosen because their records were complete. The seven excluded yards are unobserved, and the design cannot say whether they differ. With 24 clusters, intervals rest on the t distribution with 23 degrees of freedom and are wide for the species groups. The Human loading estimate includes no change, and the Twi'lek, Rodian and other estimates have upper limits just under 1. Species is as entered by paymasters, and the group labelled other pools several. The heterogeneity statistic treats the five groups as independent, though they share comparators and quarter effects; the correlation this induces makes it, and the pairwise contrasts, conservative. The pairwise contrasts were examined for loading headcount only, the one outcome with heterogeneity, and the adjustment does not cover that choice.

Ledger coverage stops at the two grades named above. Tally and gate work, and casual labour hired by gang leaders outside the payroll, are not seen. A worker who leaves a yard's rolls cannot be followed, so the ledgers do not show whether a lost post ended in unemployment, in a move to another yard or in departure from the trade.

These results do not show that droids cost less than the workers whose hours they took, or that the yards handled more cargo for it. They do not show that the workers who lost loading posts are the workers who gained supervisory and repair posts, since the aggregates by grade and species do not identify individuals. They do not say where the Wookiee loaders whose posts ended went next. That question needs the hiring rolls of other yards and of the trades beyond the yards, which this study did not reach.

labour droidscargo-handling yardspayroll ledgersstaggered adoptionpaid hoursoccupational shiftspecies and employment

References

  1. Galactic Senate Legislative Research Office (25 BBY). Hours and conditions in the cargo-handling trades of the Core, a summary of yard returns. Galactic Senate Legislative Research Office Reports, Report LRO-25/62.
  2. Maldrin, P. (29 BBY). Hours before heads, the adjustment of port and cargo labour to changes in handling capacity. Muunilinst Economic Analysis Bulletin, 14(3), 150–174.
  3. Mallowen, T. (27 BBY). Task reallocation when mechanical handling enters a trade. Muunilinst Economic Analysis Bulletin, 16(3), 77–103.
  4. Mervane, K. (28 BBY). Guild muster rolls and the distribution of species across cargo grades in the Core trades. Muunilinst Economic Analysis Bulletin, 15(3), 119–143.
  5. Morwenna, D. (25 BBY). Cluster-robust intervals for panels with few clusters, a table of degrees of freedom. Proceedings of Applied Speculative Statistics, Series B, 9(1), 90–111.
  6. Neskarin, A. (25 BBY). The plant register as an instrument of record for owned labour droids in cargo yards. Coruscant Journal of Droid Engineering, 17(1), 31–54.
  7. Nethlar, B. (27 BBY). Stacked cohorts and not-yet-treated comparators in staggered rollouts. Proceedings of Applied Speculative Statistics, 8(3), 140–163.
  8. Norvik, E. (26 BBY). Fault intervals and servicing hours of yard-service droids and handling plant. Coruscant Journal of Droid Engineering, 16(3), 189–214.
Read this article inside the full journal experience — browse by faculty, search across universes, and filter by discipline.
Open in Uncited Press →