You would ideally only sample vehicles with the same engine and options (and the same make and model). That leaves only originality and condition as variables. You use those variables to determine which percentile to use when you price the car you are looking at. Regression analysis says that it works.

 

Tough to find a meaningful sample when you start matching options, engine, etc. Also, it is tough to find actual "sale" prices as opposed to asking prices which could be off by 10-30% or more.

 

You really need to get hooked up with a dealer some nite and go to a wholesale auction. These guys don't look up anything when cars are run thru every couple minutes in multiple lanes.

 

Involving the asking price versus only actual sales with their prices and ignoring options and condition make the NADA book more and more 'realistic'.

Just for fun, If I post a bunch of '72s for sale at 500K, 750K and 1 million, what would that do to your model?

 

IMO it should use actual sale numbers, not advertised prices. Maybe he can get info on completed sales from EBay, but even those wouldnt be accurate, because people bid up on the cars and then never pay and the get relisted and sell for less.

 

really hard to find selling prices. it is usually a data point that is hard to get. most folks won't say, or they tell some surveyor a much higher number. Checking EBAY for 70-72 completed listings.. of the 92 cars it lists, only 17 sold.. and how many of them truly sold? A couple of them were projects and most were not near 20K. in fact the top 35 priced cars, did not meet reserve or had zero bids. I just don't feel asking price is a good statistic to use. ( i wish it was, then i have a better argument for the price of my car perhaps)

only in an auction where the auction house publishes the selling price does one know the price. and then factor in commission and fees, and deadbeat bidders. where else are selling prices published with any degree of certainty of the data that they publish.

do you think proteam will tell us what their cars sold for? don't think so. then they would be advertising what % they discount from asking price..

if one has a highly optioned and desirable car, it is only a matter of time where a seller will get a high price ( within reason) for their car. Such cars are hard to guess. but the common car where there are many of the same to choose from, then it is a buyers market.

 

Those are all great observations, and are all things that can be taken into account for a statistical analysis. You are correct that I'm using list rather than sale price. In the first version of the calculator I was handing out I talked about how to adjust for that in the instructions. In the second version, I listed percentiles for both list and for typical sales prices. It's easy to adjust for something that can be accurately estimated based on thr law of averages.

But look - I've gone over and over on how this thing works. Really there are only two questions that need to be answered: 1) does it work, and 2) is it useful if it does.

1) Does the calculator work? So far while many people have said that they don't see how it could work, nobody has said that it doesn't. One person questioned a 90th percentile of $46,000 for a 72 Corvette, but further analysis shows that was an accurate percentile.

It's not hard to tell if it works. You run it backward. Take an actual sale price and see if it fits into the model. 90% should. Check more and see if they tend to make sense with the specific percentiles.

Or you can do what I did. You can use regression analysis to check your assumptions. I'm assuming that price is an accurate predictor of condition - which if true would also make condition an accurate predictor of price.

I guess really it comes down to my credibility. Do I have the knowledge and experience to create this thing? I've got an MBA in Lean Manufacturing from the University of Michigan, and I'm about to finish a Masters of Science in Manufacturing Operations from Kettering University. I also have 15 years of industry experience. You can decide for yourself if that is enough to know if to and how to do this kind of analysis.

The bottom line is that the calculator works, but nobody is going to twist your arm if you don't want to use it. I know full well that some people don't trust statistics. Some people hate them so much that they'll do whatever the statistics say NOT to do. Then they wonder why the business went bankrupt.

2) The next question is.. If the calculator works, does that mean that the information it gives you is helpful? The calculator will give you an accurate estimate of a vehicle's worth based on condition. That's all it will tell you. It does NOT tell you what the condition of any given car is. To get the value of a specific car, you'll need to be able to gauge it's condition and place the car into the correct percentile.

I used to collect coins when I was younger, and had a subscription to the coin value guide newspaper (CoinWorld was it?). This is the EXACT same situation - the newspaper could tell me what a particular coin of a given grade was worth - but it could not tell me what grade a given coin was in. This calculator operates like that magazine. It gives the same kind of information.

Now - I'll freely admit that it would be nice not to have to estimate the actual sale price based on the list price. It's always nice when you don't have to estimate anything. But just as with a time study, sometimes you have to make estimates and when you do you use statistics to come up with an accurate estimate. This may add a little variance to the output, but I've been posting results and nobody has said anything about them being out of whack.. Plus, I've done the analysis to check the results. I KNOW it works.

Is the calculator worth using? That's up to each individual person to decide for themselves.

If you want a copy, I'm happy to send it to you. If you don't want a copy, that's fine too.

 

Initially I was using eBay becuase it did give actual sale prices, but I gave up on it for this exact reason. Even if you can find enough that sold to use, it just takes too long looking for them.

 

You'd have to post quite a few. And it would impact the sample size more than the actual result.

 

I looked physically at around 10 cars and not physically but spoke to sellers probably an additional 40 cars when I bought mine a couple of months ago. I already told on here what I paid and have nothing to hide - $31K for a 69 L46 matching #s convertible (Fathom Green/Black). Car has PS, PB, PW, factory side pipes and a hard top. Of the total 50 cars I looked at the asking prices were anywhere from $15k-$40k around. The car I bought had an asking price from Corvette Mike of $39K or $39.9K (don't remember exactly). I ended up getting it for around 20% less than the ask price. I would argue that the car was probably worth closer to $25-$27K to "most people" on this forum given the issues it had. Having said that, I am sure someone out there looking would have paid what I paid or maybe even more. I just brought my car in and saw a 72, t-top, automatic, 350 that was in "fair shape" that the buyer just paid $37k for. The mechanic at the shop said that the price was high that the buyer paid but that "these cars sell at prices all over the place". Bottom line is you have to look at a lot of cars and think about what it is "worth" to you. Impossible to figure out what it is worth to someone else because some might be 30% higher and 30% lower than you. Not sure why most people are not comfortable disclosing what they paid for their car but I would imagine that a lot of people throw out a lower # than they really paid. The hard lesson that I learned and most important when buying one of these cars is having someone check it out who truly knows what they are looking at.


Like Mike Ward said, asking prices are all over the place and are meaningless, if you want a very rough barometer for value you should look at NADA or one of these:

http://www.buyclassiccars.com/offsit...nus/makes.html

http://www.hi-bid.com/ (NEED TO SUBSCRIBE)

 

I've got a sneaky suspicion that the ones that make money do a lot of research first..and that some use tools just like this one to help with their research. If I can develop something like this - others can develop it too.

 

I wouldnt give 46 grand for 20 dodge darts.

 

Too funny - was going to make my point by saying the same thing. Different strokes for different folks.

 

I wouldn't either. But Autotraderclassics has one for sale right now for $60,000.

I'm not selling anything here. If you don't want to use it, don't! I made it to help people and distributed it for free. You can do whatever works for you.

 

If only our cars were worth what we thought they were.

 

The only dart worth $46K would be a GTS with a big block. I would pay $46K for a nice one as rare mopars always hold value. They have probably lost the least value in the last several years as compared to C3's. I would only pay $46K for a 72 LT1, not even for a nice convertible. The model is an interesting idea if you could somehow capture actual sales to remove variability of dealer mark-up. Thanks for the intellectual stimulation

 

You can't capture actual sales unless someone published it, but you can gather the average difference between list and actual sales prices - that gets studied regularly. And since that's a known amount, you can use it to compensate when using list prices.

This is all very basic stats gentlemen. There is nothing ground-shattering in the calculator. It uses pretty basic techniques. And it is accurate. But if you don't trust it, you don't have to use it.