Federal Register - July 28, 2021
Version en texte Qu'est-ce que c'est?Dateas est un site Web indépendant, non affilié à un organisme gouvernemental. La source des documents PDF que nous publions est l'agence officielle indiquée dans chacun d'eux. Les versions en texte sont des transcriptions non officielles que nous faisons pour fournir de meilleurs outils d'accès et de recherche d'informations, mais peuvent contenir des erreurs ou peuvent ne pas être complètes.
Source: Federal Register
40752
Federal Register / Vol. 86, No. 142 / Wednesday, July 28, 2021 / Rules and Regulations TABLE 4GTL MATCHING SUMMARY STATISTICS
Number of contracts
Mean $
Median $
Std. dev.
$
Minimum $
Maximum $
Pre-Matching Larger Jails
Prisons
REDACTED
REDACTED
REDACTED
REDACTED
REDACTED
REDACTED
REDACTED
REDACTED
REDACTED
REDACTED
REDACTED
REDACTED
REDACTED
REDACTED
REDACTED
REDACTED
REDACTED
REDACTED
REDACTED
REDACTED
REDACTED
REDACTED
REDACTED
REDACTED
Post-Matching Weighted Larger Jails
Prisons
REDACTED
REDACTED
REDACTED
REDACTED
REDACTED
REDACTED
Post-Matching Maximum Larger Jails
Prisons
REDACTED
REDACTED
15. Prior to the adjustment, GTLs perminute costs are both high compared to other providers and essentially flat across facility types. There is no statistically significant difference in per-minute costs between GTLs larger jails and prisons. This is highly unusual, as the Commission would expect firms to exhibit economies of scale by spreading their fixed costs over more call
REDACTED
REDACTED
REDACTED
REDACTED
minutes, thereby reducing their per-minute costs on larger contracts. For comparison, the average larger jail contract has 9.3 million minutes of use while the average prison contract has 34.6 million minutes of use. For example, REDACTED After performing the k-nearest neighbor adjustment, GTL costs also exhibit economies of scale, and the difference in per-minute costs between GTL
prisons and larger jails is statistically significant at the 1% level.
16. The Commission can now estimate the effect that the GTL cost adjustment has on the overall distribution of per-minute costs in the contract-level data. Table 5 presents the average per-minute costs across all nonoutlier prison and larger jail contracts after adjusting GTL costs.
TABLE 5ALL CONTRACTS POST-MATCHING SUMMARY STATISTICS
Number of contracts
Mean $
Median $
Std. dev.
$
Minimum $
Maximum $
Post-Matching Weighted Larger Jails
Prisons
REDACTED
REDACTED
REDACTED
REDACTED
REDACTED
REDACTED
REDACTED
REDACTED
REDACTED
REDACTED
REDACTED
REDACTED
REDACTED
REDACTED
REDACTED
REDACTED
REDACTED
REDACTED
Post-Matching Maximum
khammond on DSKJM1Z7X2PROD with RULES2
Larger Jails
Prisons
REDACTED
REDACTED
17. Even when using the conservative approach of replacing GTLs per-minute costs with the highest costs of the three nearest neighbors, the overall per-minute cost of prisons and larger jails drops substantially.
This is unsurprising as not only are GTLs costs high, but GTL also operates REDACTED prison contracts and REDACTED larger jail contracts. With the adjusted GTL observations, the full contracts data now indicate a decreasing per-minute cost of operating larger facilities. The reason is twofold: first, because GTL has a larger market share in the provision of inmate calling services for prisons than for larger jails, even a uniform reduction in its costs per minute across facility types would exert greater downward pressure on the average costs of prisons compared to larger jails; and second, because other firms do exhibit returns to scale, the results of the nearest neighbor matching procedure highlight this important aspect of the data. Hence the procedure adjusts GTL per-minute costs for each facility type to reflect this market reality.
18. Finally, to better visualize the GTL data adjustment, the Commission presents
VerDate Sep<11>2014
19:16 Jul 27, 2021
Jkt 253001
REDACTED
REDACTED
REDACTED
REDACTED
overlaid histograms of GTL and non-GTL perminute costs for prison and larger jail contracts after performing the k-nearest neighbor matching procedure in Figures 2
and 3. These are overlaid histograms rather than stacked bar charts. Therefore, the dark red color represents the intersection of GTL
and non-GTL contracts, and the total number of contracts at any cost bin is the sum of the GTL and non-GTL bars. REDACTED
Figure 2CPM Distributions for Prisons with k-Nearest Neighbor Matching REDACTED
Notes: CPM is the cost per minute. Dark red areas are where the Non-GTL and GTL
bars overlap.
Figure 3CPM Distributions for Larger Jails with k-Nearest Neighbor Matching REDACTED
Notes: CPM is the cost per minute. Dark red areas are where the Non-GTL and GTL
bars overlap.
2. Analysis of GTL Neighborhoods 19. To further examine the nearest neighbor results, the Commission explores
PO 00000
Frm 00072
Fmt 4701
Sfmt 4700
the matches for each of GTLs REDACTED
non-outlier contracts. Aside from the choice of contract characteristics on which to perform the matching, the approach is nonparametric and relies only on the data to find the nearest neighbors of each observation.
Nevertheless, the Commission wants to understand whether a single firm is dominant in the matches or if there is variation in the neighbors found. Even if the matches are overwhelmingly to a single firm, the legitimacy of the procedure is not in doubt as it is only a reflection of the data.
However, the results would be less robust if an argument could be made for that firm also having unreliable cost data. In Table 6 below, the Commission presents the total number and percentage of time that each firm matches with a GTL contract, categorized by type of facility. The Commission notes that within the total dataset of 2,900 contract observations, GTLs smaller jail contracts only matched with other providers smaller jail contracts.
E:FRFM28JYR2.SGM
28JYR2
