I’m Todd L. Burns, and welcome to Music Journalism Insider, a newsletter about music journalism. Click here to subscribe!
Dr. Bill Carroll is Adjunct Professor of Chemistry at Indiana University, Bloomington. Bill has been a music lover all his life. “From the time I was about five, I was fascinated by the charts—the earliest being Dick Clark’s top tens in the late 1950s. Growing up in Chicago, I had WLS Silver Dollar Surveys taped to my bedroom wall. I worked in radio in college and grad school and read Billboard faithfully. Later, my company was in the business of making vinyl resin for records, and I got the company to spring for Billboard as a business expense.” Recently, Bill published the article “Did Billboard, Cash Box, and Record World Charts Tell the Same Story? Perception and Reality, 1960-1979.” Fascinated by the premise and methodology, I reached out to Bill to hear more.
Can you please briefly describe yourself and what you do in music?
My background is in science, and I spent my career in the chemical industry; however, from the time I was about 5, I was fascinated by the charts—the earliest being Dick Clark’s top tens in the late 1950s. Growing up in Chicago, I had WLS Silver Dollar Surveys taped to my bedroom wall. I worked in radio in college and grad school and read Billboard faithfully. Later, my company was in the business of making vinyl resin for records, and I got the company to spring for Billboard as a business expense.
From the mid-90s I had the idea to extend statistical and mathematical analytics to the charts. I started that work in earnest in 2013 by transcribing Billboard, Cash Box and Record World singles charts week to week between 1955 and 1991 to create my new “laboratory.”
So, I’m a quant. “Good” and “bad” don’t appear in my work; “higher scoring” and “lower scoring” do. There are three underlying principles to my approach. First, the entire week-to-week scoring history of a record counts toward its lifetime score. Second, higher ranks receive disproportionately higher scores—1000 for #1, 824 for #2, 658 for #3 and so on down to 2 for #99 and 1 for #100. Finally, scores must be normalized to reflect their era, since the charts changed markedly over forty years.
That final statement requires a little explanation. In the mid-‘60s, about 750 records entered the Billboard chart each year. By the mid-‘80s, it was about 400. In the mid-‘60s, the average record spent about 7 weeks on the chart; by the mid-‘80s, it was about 13 weeks. Since a record can’t score if it’s off the charts, records in the mid-‘60s are subject to a structural points disadvantage to those in the mid-‘80s. Point scores are not directly comparable.
My normalization procedure divides each record raw score by the scores of all records charting in the 52 weeks surrounding its entry. Thus, in every era an average record has a score of 1. The principle is that a record scoring 10x the average in its era is more comparable to a record scoring 10x the average in another era.
Can you briefly summarize your article?
I like to formulate questions that can be answered explicitly by these methods, and this article contained two of those questions. Despite the fact that each magazine’s “secret sauce” for creating its charts was unknown to those outside the magazine, people in the industry believed that the three magazines measured record performance differently. There is a quote from Tommy James in the paper that outlines his opinions of how the magazines were grossly and systematically different. He wasn’t the only one who felt that way. Were they correct?
Similarly, the godfather of chartology, Professor Peter Hesbacher of the University of Pennsylvania tried to relate the performance of number 1 records in the 1970s between Billboard and Record World. While he found concordance in the popularity of the very top records of the decade between the two magazines, just below the very top he found gross disparity. How could “How Deep Is Your Love?” by the Bee Gees rank 7th for the decade in Billboard but 99th in Record World?
The paper set out to answer those questions. I constructed a dataset of the lifecycles week-to-week of 4578 records, each of which peaked in the top 40 in all three magazines for the era 1960-1979.
First, I found statistical but not perceivable differences in the components of record lifecycles: rise, time at peak, fall, and total duration; that is, the differences among the magazines were fractions of a week; the discernable granularity for observing chart behavior is one week. No difference among the magazines.
On the other hand, there were marked differences between the decades 1960-69 and 1970-79 in those same chart components across all magazines. In all cases, the durations in the 1970s were discernibly longer than those of the 1960s.
While there was no systematic difference among the magazines, I found that the three magazines could treat individual records very differently. In about 15% of the records in the dataset one of the three scores was a significant outlier from the other two. Thus, judging the highest scoring records of an era will depend on the magazine you choose; if you choose to average the three, in many cases you are averaging in an anomalous outlier. This was exactly what Prof. Hesbacher found in his studies.
Why are there outliers? Because the details of methodologies are not known, it is impossible to say definitively. However, for at least a good part of its history, Cash Box was known not to use airplay as a popularity component. The other two magazines did. Curiously, Cash Box tended to have the fewest outliers, either high or low. In those days, singles markets had a strong local flavor. Given the number of radio stations in the US it seems challenging to generate a consistent representative sample of airplay nationally. I suspect airplay is the source of the variability.
What was the most surprising thing that you found in your research?
I’m easily surprised. To me, the entire thing was surprising, but particularly that there was the lack of systematic differences nationally among the magazines.
Can you point us to any further research on the topic that you think is relevant?
I think there’s little more that needs to be done here; however, I keep looking for interesting chart related questions.
What’s next for you?
I’m currently studying the evolution of the LP between 1962 and 1972. At the beginning of that period, albums were soundtracks, original casts, comedy, classical and collections by instrumentalists (Mantovani, Ferrante & Teicher…) or song stylists (Al Martino, Vic Dana…). There was precious little rock. Typically an album might be constructed from two singles, two B sides and 8 “filler” songs, in most cases covers of other hits. By the end of the period, rock pretty much owned the album charts and while there were still generally only two singles on an album, it was much more common for rock acts to be writing and recording their own songs rather than covers. And the instrumentalists and song stylists were going extinct.
I’m interested in generating statistics on how that transition occurred. I think 1967 was probably the fulcrum year, but I’m still organizing the data set which contains about 5800 albums consisting of about 60,000 cuts.
Did you enjoy this interview? You can support this newsletter by subscribing here. Among other things, you’ll get full access to all 400+ interviews I’ve done as part of the newsletter. I’ve talked with writers and editors from Pitchfork, Rolling Stone, the Guardian, and many more.