Nerding out with the data some more, I also did a regression of days to delivery on car and order characteristics. I used only orders from May 1, 2020 (after delivery times level out) to September 30, 2020 (not so recent that many are still not delivered). Here are the results:
Code:
Source | SS df MS Number of obs = 932
-------------+---------------------------------- F(61, 870) = 3.17
Model | 58323.113 61 956.116607 Prob > F = 0.0000
Residual | 262611.604 870 301.852418 R-squared = 0.1817
-------------+---------------------------------- Adj R-squared = 0.1244
Total | 320934.717 931 344.720426 Root MSE = 17.374
-------------------------------------------------------------------------------------------
delivery_days | Coef. Std. Err. t P>|t| [95% Conf. Interval]
--------------------------+----------------------------------------------------------------
performance | 12.74784 4.490055 2.84 0.005 3.935233 21.56044
fsd | -2.531619 1.338799 -1.89 0.059 -5.159273 .0960352
white_int | 1.428309 1.322099 1.08 0.280 -1.166568 4.023185
tow_hitch | 4.300699 1.345878 3.20 0.001 1.659151 6.942247
|
color |
Deep Blue Metallic | 4.375689 1.631864 2.68 0.007 1.17284 7.578539
Midnight Silver Metallic | 1.496763 1.549327 0.97 0.334 -1.544092 4.537618
Red Multi-Coat | 1.732264 2.321584 0.75 0.456 -2.824295 6.288824
Solid Black | 1.083944 2.183176 0.50 0.620 -3.200963 5.368851
|
days_to_q2 | .4548672 .0786639 5.78 0.000 .300474 .6092604
days_to_q3 | .0485498 .022743 2.13 0.033 .0039123 .0931874
|
wheels |
20 inch | -1.119032 1.291963 -0.87 0.387 -3.654761 1.416696
21 inch | -16.30479 4.842879 -3.37 0.001 -25.80989 -6.799703
|
state | [omitted for brevity, see attachment if interested]
|
_cons | 32.19171 17.53327 1.84 0.067 -2.220739 66.60417
-------------------------------------------------------------------------------------------
So for the baseline LR Model Y in white with 19 inch wheels and black interior, you’ll expect to wait about 32 days (the “_cons” line). If you upgrade to performance, expect 12 more days (the “performance” line). Tow hitch adds about 4.3 days. FSD
reduces time to delivery by about 2.5 days (maybe there is some small priority given?).
I added controls for days until end of Q2 (longer delays then) and days until end of Q3. In both cases, it seems there is an end of quarter delivery push. Ordering earlier in the quarter is correlated with longer waits, but moreso in Q2 than Q3. (No data used from Q1 or Q4).
Using this to predict days until delivery for orders in the sample and after (the red dots), you can see that there is still a lot of unexplained variation. On the other hand, I think it validates many of our “feelings” about things that affect wait time.
View attachment 614494