Thanks. Interesting that you clump the 95-96 with the 97-01. My hypothesis was that the automatic "hit" is larger against the years that the 3.2 was the manual engine (97-04) vs the earlier cars... looks like it's not a valid hypothesis (or could be a low data thing). Frankly, seeing only 1% difference on 5 2001 automatic cars vs manuals is shocking. I assume these are all corrected for all the other factors that affect value.
That's the beauty of these investigations, sometimes there's more than one answer to the question! The question I answered was more "how do I handle the automatic value difference internally, and what does it look like for each NSX class?" Correct, I filtered out as much data as possible to get as close to nominal comparisons as possible, with the intent of evaluating the automatic value impact only.
In other words, what I showed in the previous post was like colors comparison (blue to blue dash, red to red dash), comparing "likes to likes" to show the automatic impact compared to the closest comparable NA1 manual NSX. So yes, I do value 1995-2001 Targa Automatics as 1995-1996 Targa Manuals with an value multiplier for the automatic transmission (and the results are generally favorable).
With your clarification, I think you're looking more at what is the raw difference when moving vertically on the graph for each year. Here's the data for that (again filtered such that only the automatic impact is considered):
<google-sheets-html-origin style="color: rgb(0, 0, 0); font-size: medium;">
| Avg % Value Difference Auto vs Manual | Auto Count | Manual Count |
1991 | -19.12% | 46 | 198 |
1992 | -27.91% | 9 | 64 |
1993 | -9.48% | 3 | 49 |
1994 | 0.43% | 10 | 31 |
1995 | -21.58% | 6 | 73 |
1996 | -18.23% | 9 | 44 |
1997 | 15.44% | 3 | 27 |
1998 | -29.24% | 1 | 27 |
1999 | No Autos | | 15 |
2000 | No Autos | | 21 |
2001 | -15.71% | 5 | 17 |
2002 | -28.91% | 1 | 18 |
2003 | No Autos | | 19 |
2004 | -46.30% | 1 | 19 |
2005 | No Autos | | 31 |
<tbody>
</tbody>
</google-sheets-html-origin>
Since the raw sale prices were used in this calculation, an uneven or irregular distribution of sale prices throughout time can lead to outliers. In other words, if all 3 1997 NSXs automatics sold recently, the average would be misleadingly higher than the corresponding 1997 manuals which have sold over a broader range of time since 2010 (one explanation for the abnormal 1997 result).
Again, little to no Automatic data impacts the strength of the conclusion drawn, but there does appear to be a general trend toward the automatic having more of a negative effect on value when compared to the equivalent manual NSX from the same year as the year increases (especially NA2s vs same year automatics), which I believe is your original hypothesis.