A statistical analysis of the value of third party grades

abitofthisabitofthat

Hi – My apologies in advance for the length of the post, but I have a few things that I want to say. I am a part time dealer and an avid collector of 19th century proof coinage. I have viewed thousands of these coins, and own quite a few. I have made observations about how PCGS and NGC grade coins, and the differences between these two services. I would like to present a (hopefully) unbiased analysis of grading based upon verifiable fact, and not opinion.

Of course I have my opinions, as do all of you, as does NGC, PCGS, CAC,etc. However, I believe that a factual enquiry may make for an interesting discussion of TPG grading.

Although my focus is on all 19th and early 20th century proof type coins, I have kept the sample in my statistical analysis small to make a few simple points without getting bogged down in too much data. The sample that I am using is Proof Seated Dimes from 1860 through 1891, graded by both PCGS and NGC, and auctioned through Heritage during 2008. This sample of over one hundred coins sold at auction gives a good data set to work with. It is certainly not exhaustive. There are many venues other than Heritage, there are many series other than proof Seated Dimes. However, I believe that my conclusions will be useful and can generally be applied to the wider numismatic marketplace.

Rather than focus on arbitrary and difficult-to-define standards of A, B, and C quality coins, I will focus solely on prices realized as a determinant of value. After presenting the data, I will make some observations and will certainly welcome your thoughts as well.

In grades PF60-PF62, the data set for NGC coins was 14 auctions records with an average price (all prices include fees) of $393, a max of $632 and a min of $276, PCGS had 9 records and an average price of $384, a max of $460 and a min of $322..

In PF63, NGC recorded 18 coins with an average price of $555, a max of $862 and a min of $431, PCGS recorded 11 coins with an average price of $760, a max of $1,380 and a min of $460.

In PF64, NGC recorded 23 coins with an average price of $835, a max of $1,035 and a min of $578, PCGS recorded 23 coins with an average price of $1,101, a max of $1,955 and a min of $575.

In PF65, NGC recorded 17 coins with an average price of $1,498, a max of $2,760 and a min of $1,006, PCGS recorded 11 coins with an average price of $2,185, a max of $3,737 and a min of $1,207.

In PF66, NGC recorded 20 coins with an average price of $2,048, a max of $2,760 and a min of $1,498, PCGS recorded 8 coins with an average price of $4,154, a max of $10,350 and a min of $1,840.

In PF67, NGC recorded 26 coins with an average price of $4,059, a max of $6,325 and a min of $2,530, PCGS recorded 2 coins with an average price of $3,953, a max of $4,312 and a min of $3,594.

In PF68, NGC recorded two coins and PCGS none.

Thoughts and conclusions. In this sample, on average, PCGS coins brought higher prices in all grades of 63 and above. Also, PCGS coins have higher Max and lower Min observations. The higher the grade, the greater the disparity between services, until you reach 67, when the sample breaks down due to too few observations.

Of note, an NGC coin graded PF 66 sell on average for $2,048. A PCGS coin graded one point lower, 65, sells for approximately the same price - $2,158. An NGC coin graded PF 67 sell on average for $4,059. A PCGS coin graded one point lower, 66, sells for approximately the same price - $4,154.

This is not to say that one service is better. Only that their standards are, apparently, different. Buying a PCGS 65 and cracking it out to get an NGC 66 is not an upgrade. It was a $4K coin, and it still is. You cannot expect that the average NGC 65, or 66 will cross at the same grade to PCGS. If you buy the average NGC 65 for $1,498 and cross it to PCGS you have a coin that, on average, sell for $2,158. It is not logical that this could exist long term. The standards are different.

I have both PCGS and NGC coins in my collection. I buy the coin and not the holder. In my preferred habitat – Proof 64 – there is not a huge difference, IMO, between the two services, either in prices or in my subjective opinion of grading. I believe that there are differences in the higher grades.

This little bit of statistics was not at all surprising to me. It validated what I have already observed. By the way, it was not hard to do and took about an hour. Why don’t you try the same with the series that you collect and see if the results are similar? Post your observations.

I welcome your thoughts and comments.

BTW – I believe that CAC is grading more in line with PCGS standards, and it will be interesting to see if things change going forward. It has been my recent experience that since NGC changed to the new holder, their standards have moved closer to what PCGS employs.

All of this is my opinion and observations based upon a set of facts derived from auction records. Again, I do not feel one service to be “better”. Just different. What I value is consistency in grading over the long term.

Thanks for reading.

FatMan

Let me know when the cliff notes version is available. It looks interesting but too long for me to read.

RWB

Thanks for posting your work. Might be more accessible if you added a graph of the results.

abitofthisabitofthat

I tried to add a table - but could not figure out how. Oh well.

RWB

If you make a graph in Excel, save it as an image, then post it just like you would a coin photo via Photobucket or other service.

tradedollarnut

No doubt that NGC grades proof seated coinage differently than PCGS - I once did a thread in the Set Registry Forum discussing this very fact. Of course, my conclusion was a bit different - that it was impossible to create a collection of the finest proof seated coins in PCGS holders without committing financial suicide.

LRCTom

Interesting post...thanks.

My "Cliffs Notes" summary, if I read you right, is that the same coin brings the same price at auction, regardless of whose slab it's in. The PCGS number tends to be one point less than the NGC number, overall. But the bottom line is...people are buying the coin, not the slab.

Is this a fair summary?

...Tom

abitofthisabitofthat

<< <i>Interesting post...thanks.

My "Cliffs Notes" summary, if I read you right, is that the same coin brings the same price at auction, regardless of whose slab it's in. The PCGS number tends to be one point less than the NGC number, overall. But the bottom line is...people are buying the coin, not the slab.

Is this a fair summary?

...Tom >>



Exactly - people are smart. The market is reasonably efficient. The grades on the holders, at least in certain instances, are not comparable.

Raybo

Looks like to much work to me.

I buy what I like for the price (Cliff note version)

Coinosaurus

Thanks for posting - most illuminating.

lathmach

Interesting.

Ray

DonWillis

A very astute post. Imagine what it was like in the old days when the only numbers we put on coins was the price! If we did that today a whole lot of BS would get cleared up very quickly. But that won't happen because too many people are making too much money telling the uneducated that the only numbers they need to be concerned with are the numbers printed on the insert.

Bear

<< <i>A very astute post. Imagine what it was like in the old days when the only numbers we put on coins was the price! If we did that today a whole lot of BS would get cleared up very quickly. But that won't happen because too many people are making too much money telling the uneducated that the only numbers they need to be concerned with are the numbers printed on the insert. >>



Dog gonnit Don, your starting to sound like a Forum Member more every day.

PaleElf

Interesting thread... Thanks for the info.

lkeigwin

I read the post and found it interesting. Thank you for spending the time. I love this place.

The two TPG's are fairly close at the lower gem grades, a point difference at the higher. Not very surprising. As long as everyone gets this then it isn't problematic.

I think it's a bit different with copper, though. I would guess the average NGC coin is graded at least a point above PCGS. My personal experience isn't huge...maybe 40 early Lincolns crossed in the last 12 months. But this is what I hear from well respected dealers too. FWIW...

Lance.

CertifiedGoldCoins

<< <i>A very astute post. Imagine what it was like in the old days when the only numbers we put on coins was the price! If we did that today a whole lot of BS would get cleared up very quickly. But that won't happen because too many people are making too much money telling the uneducated that the only numbers they need to be concerned with are the numbers printed on the insert. >>


I think I can see light at the end of the long, dark tunnel we're in.

Raybo

I think most of the fun of my hobby went south with all these numbers.
I know it's a business for a lot of people, but IMHO it has got out of control.

Ray

1960NYGiants

<< <i>
.... Also, PCGS coins have higher Max and lower Min observations. ....
>>



This statement does read as an accurate assesment to me. Your PCGS vs NGC values consistently show higher max and higher min prices within the the grade for PCGS, at PF63 and above.

Frankcoins

I don't think any analysis is possible mixing TWO variables...the fact that experts can grade the same coin from AU58 to MS63 plus the fact that
auction prices are already all over the place depending how badly someone wants the coin, or if they are even looking at the auction.

ressikanflute

Thank you for the information.

I was doing something similar. Try this one time... buy an 30 year old Coin magazine off ebay and compare todays advertised coin prices vs say... 1973.

I comparing advertised prices for 1891 CC Morgans. In Decemer 1973 Coinage they could be had RAW BU for $85. Today In Coins there was an advertisement for RAW Gem BU for $450. I paid $1000 for a PCGS slabbed MS64.

Pretty shocking when you venture outside of the hobby aspect and look at it as an investment. Not much there after you factor in inflation.

Other notables from 1973...

1972 Lincoln double die in Bu... $4 to $10
1937 Buffalo three legged BU... $85

DRUNNER

Superb work and very much worth the time it took to complete it . . .and read it.

Bravo . .

Drunner

TomB

I'll give a little criticism to your post. You state-

<< <i>The higher the grade, the greater the disparity between services, until you reach 67, when the sample breaks down due to too few observations. >>


I put the text in bold to point out that this statement is making an assumption on what you expect to find vs. what is actually noted in the data. In other words, you expect that if there were more PCGS data points that you would have a different conclusion. However, you have no statistical basis to make this statement. It is essentially a statement that illustrates you either have a bias toward the data or are extrapolating beyond reason.

Please note that for many years I have written that PCGS and NGC have their own proprietary grading standards that do not have to agree with one another or agree with any ANA published materials. Additionally, I believe that the PCGS standards for most series and in most grade ranges are at least as strict as the NGC standards for these series or ranges. Lastly, I also have written for many years that the overlap of the grade window between the two services becomes less as the window gets smaller for many series and ranges and this applies typically to higher grades.

lasvegasteddy

market swings as to day of week-month-season plays a role here

coinkat

What were the Dates involved in your analysis? What period of time is involved and when did the PCGS and NGC coins sell? What was the selling price at that time in relationship to either the greasheet or other aucyion estimate?

Nice work but it still seems incoplete to draw the types of conclusions that folks seem to be making here.

For the record, I have no axe to grind as I like PCGS and NGC

abitofthisabitofthat

<< <i>What were the Dates involved in your analysis? What period of time is involved and when did the PCGS and NGC coins sell? What was the selling price at that time in relationship to either the greasheet or other aucyion estimate?

Nice work but it still seems incoplete to draw the types of conclusions that folks seem to be making here.

For the record, I have no axe to grind as I like PCGS and NGC >>



Certainly good points. I offered this up for discussion and invite follow up or analysis of different dates/series. The results could certainly have been skewed by the relative mix of dates, the relative frequency of cameos in the group, the timing of the auctions, and a number of other factors.

I personally find the information useful because it conforms to my experience and observation. This is not news to anyone - that PCGS coins tend to do better at auction. I was just trying to take some of the subjectivity out of most conversations around grading standards - dealers and collectors have "voted" with their wallets on grading and, to make a very broad generalization, the market seems to believe that the standards of PCGS and NGC differ by one full point for this series in this grade range. Others series will likely produce different results.

As a collector, if a dealer offers me an NGC coin and quotes auction records for PCGS coins to support his pricing, I know to discount this information. It would be more useful to benchmark to other NGC coins, or PCGS auction records for coins graded one point lower.

Again, these are generalizations, and obviously each coin stands on its own merits. I own both NGC and PCGS coins in my collection and value both services.

DHeath

A very astute post. Imagine what it was like in the old days when the only numbers we put on coins was the price! If we did that today a whole lot of BS would get cleared up very quickly. But that won't happen because too many people are making too much money telling the uneducated that the only numbers they need to be concerned with are the numbers printed on the insert.

Mr.Willis, great post! You clearly favor collecting vs. investing.

coinkat

I appreciate your efforts in an attempt to:

" take out the subjectivity out of most conversations around grading standards - dealers and collectors have "voted" with their wallets on grading and, to make a very broad generalization, the market seems to believe that the standards of PCGS and NGC differ by one full point for this series in this grade range. Others series will likely produce different results".

Unfortunately, that is a very tall order considering that grading itself has such a large subjective component. The quality of the coin is as much as a driving factor as the plastic, and therefore, the coin itself needs to be factored into the analysis. Not all 63, 64 and 65 coins are created equal regardless of which holder the coin is currently housed. How do you factor the coin into the analysis? I think that becomes alittle troublesome because the list of factors could be surprising. If it is a coin that is rarely offered, there could be several collectors chasing that coin which could easily impact the analysis. I suppose one could even look at population reports as a means of comparing a specific date between the two services?

By the way, where is Longacre? He should be all over this

Aegis3

I'm not sure how valuable this is...yet.

First, the data. Prices compared over a year just introduces more variables we are really comparing. I think a better way would be to compare prices of the same date/grade/designation, when sold in the same auction. So only compare, say, 1875 PF-66 when there are at least two (one PCGS, one NGC) in the same auction. Perhaps you can get away with all the auctions held at or before a certain show. And compare the differences in the price. Probably percentage, rather than absolute.

Second the statistics. I am far from an expert in statistics, and other than one undergrad course, I have taught myself what I have needed when I have needed it. And I haven't needed the statistics to compare differences between populations. But I know at a minimum you'll need the average differences, the standard deviation of the difference, and the sample size. What statistical tests you need to do after tha I'm unsure of. From what you have done so far I'm not convinced there's a statistically valid difference in the prices.

Third, the explanations. Even if, after you compile the data and run through the statistics, you'll need to explain it. Perhaps it is possible that it can be explained by different corporate "grading standards." Perhaps part of it can be. But what about psychology? How will you probe this? This certainly is a valid hypothesis, and would need to be tested. The most obvious way (and don't think it will be easy) will be to find some coins sold at one time in auction A, and than sold later together in auction B, with the same grades, but one crossed over to a new holder. If there is no psychological effect, the difference in the price changes will, on average, be zero. Of course, you'll need multiple examples to probe this. My suspicion is that such an effect does exist, and is probably at least as important as any difference in "grading standards."

abitofthisabitofthat

<< <i>I'm not sure how valuable this is...yet.

First, the data. Prices compared over a year just introduces more variables we are really comparing. I think a better way would be to compare prices of the same date/grade/designation, when sold in the same auction. So only compare, say, 1875 PF-66 when there are at least two (one PCGS, one NGC) in the same auction. Perhaps you can get away with all the auctions held at or before a certain show. And compare the differences in the price. Probably percentage, rather than absolute.

Second the statistics. I am far from an expert in statistics, and other than one undergrad course, I have taught myself what I have needed when I have needed it. And I haven't needed the statistics to compare differences between populations. But I know at a minimum you'll need the average differences, the standard deviation of the difference, and the sample size. What statistical tests you need to do after tha I'm unsure of. From what you have done so far I'm not convinced there's a statistically valid difference in the prices.

Third, the explanations. Even if, after you compile the data and run through the statistics, you'll need to explain it. Perhaps it is possible that it can be explained by different corporate "grading standards." Perhaps part of it can be. But what about psychology? How will you probe this? This certainly is a valid hypothesis, and would need to be tested. The most obvious way (and don't think it will be easy) will be to find some coins sold at one time in auction A, and than sold later together in auction B, with the same grades, but one crossed over to a new holder. If there is no psychological effect, the difference in the price changes will, on average, be zero. Of course, you'll need multiple examples to probe this. My suspicion is that such an effect does exist, and is probably at least as important as any difference in "grading standards." >>



Thanks for the post. Good points all. As for the first two points, I got a good bit of razzing about the data to begin with, so I did not want to get involved in standard deviations. However, they would not really be relevant. For a standard deviation to be truly meaningful data needs to be normally distributed about the mean, and this data is not. A more specific analysis of same date / grade in the same auction but in different holders, as you suggest, would be interesting to perform.

The comment on psychology is much more interesting to me. I believe that it is true, and plays no small part. However, given the differences of, in some cases, a 100% premium for PCGS over NGC, I believe that many, if not most, market participants would attempt a cross to PCGS prior to auction. A $40 fee compared to an average $1,000 price realized differential is certainly worth the shot. Therefore, I believe that many (most?) of the NGC coins are in NGC holders because they would not cross at the same grade.

Of course, at this point I have left my attempt to be unbiased based upon data and am well out onto the limb of conjecture. Also, to stress, I am not saying that this is useful to purchasing or valuing specific coins. Each coins must stand on its own. I just thought this approach interesting and a good conversation starter.