Contents

Sections

1. Introduction
2. About Me
3. Description of Methodology
4. Raw Data
5. Normalized Figures
6. Analysis
7. Conclusions (?)

Tables

1. Raw Sales Figures, 1999-2002 (millions of dollars)
2. Sales Figures Normalized to 2000 Sales (per-company)
3. Average Sales Per Company Per Year (1999-2002)
4. Average Effect of Economy Each Year (1999-2002)
5. Average Effect of Each Business/Group (1999-2002)
6. Expected Net Sales & Error (1999-2002)
7. Average Sales Per Company Per Year, Take 2 (1999-2002)
8. Average Effect of Economy Each Year, Take 2 (1999-2002)
9. Average Effect of Each Business/Group, Take 2 (1999-2002)
10. Expected Net Sales & Error, Take 2 (1999-2002)

Graphs

1. Non-CD Sales, 1990 - 2000
2. Categorized Percent Change in Sales, 2000 - 2002

Introduction

The news these days is plastered with stories of the untold billions of dollars the recording industry is losing to digital piracy. On a recent plane flight from coast-to-coast, after the main movie the airline showed a series of short films, including one about how digital piracy was wrecking the music industry. Hillary Rosen, the Recording Industry Association of America (RIAA) lobbyist frequently talks of the billions of dollars each year the industry is losing in the wake of the "plague" of digital piracy.

Yet at the same time, this country is in one of the worst economic situations in decades. The states are facing budget deficits not seen since the end of World War II. The Dow Jones Industrial Average (DJIA) has dropped over twenty percent in the last two-and-a-half years; the NASDAQ has lost over seventy percent of its value, from a peak of around 5,000 to its current level in the neighborhood of 1,500. Is it even remotely possible that this could account for some percentage of the sales decline the RIAA members see?

This page presents a brief examination of market trends and sales figures of the economy at large versus the sales figures of RIAA members. This is by no means a complete economic or statistical analysis -- I have a minimal background in those fields -- but rather an attempt to gauge trends in the economy at large and RIAA sales figures.

About Me

I am a graduate student in computer science at Duke University. I have finished three years of study, and am currently an intern at HP Research Labs.

NOTE: The views, figures, opinions, and conclusions expressed here are my own, and do not represent those of Duke University or Hewlett-Packard in any way whatsoever. This is an indepedent project, and is not funded or supported by any third parties whatsoever.

I have taken a class in copyright law at the Duke University Law School, and have taken classes in probability, statistics, and the analysis of experimental data. I do not profess to be an expert in these fields, but know enough to perform this cursory analysis.

Description of Methodology

In order to obtain a high-level view of the trends in sales figures for the economy at large, I decided to compare the sales data for RIAA members against those of the thirty members of the Dow Jones Industrial Average. These thirty companies include banks, manufacturers, technology companies, and medical research and production companies. They are




3M	Alcoa	American Express	AT & T	Boeing	Caterpillar
Citigroup	Coca-Cola	DuPont	Eastman-Kodak	Exxon-Mobil	General Electric
General Motors	Hewlett-Packard	Home Depot	Honeywell	IBM	Intel
International Paper	Johnson & Johnson	JP Morgan/Chase	McDonald's	Merck	Microsoft
Philip Morris	Proctor & Gamble	SBC Communications	United Technologies	Wal-Mart	Walt Disney




Sales data was obtained from the CBS Marketwatch website, and goes back to 1999. The study uses the Net Sales row of the Financial data for the company' stock ticker symbol. The list of companies and the sales data was downloaded and is accurate as of 2003/05/16.

The majority of the figures here will study on the differences in sales and other statistics between 2000 and 2002. 2000 is the year that RIAA sales peaked, and coincidentally it's also the year that the dot-com bubble hit its peak. For both the RIAA most of and the rest of the economy it's been a steady downhill march to the sales figures of 2002.

There are two measurements for the RIAA data. The first -- usually simply labeled 'riaa' -- are the net sales for RIAA-member media, including CDs, CD singles, cassettes, vinyl, and music videos. The second -- labeled 'riaa-cd' -- are the sales for CD albums only. There are two reasons for this split. The first is that CDs comprise over ninety-five percent of total RIAA-member sales. The second is that Hillary Rosen and other RIAA spokespeople have insisted that the loss in sales is in the CD market, that they're losing sales of $15 - $20 albums to digital pirates.

Raw Data

This section will present the raw sales data for the thirty DJIA members and the two categories of RIAA sales, both in tabular and graphical form. First, a look at the raw sales figures, sorted by 2002 net sales:

Table 1

Raw Sales Figures, 1999-2002 (in millions)

Index	Company	1999	2000	2001	2002	Change 2000-2002
1	wal-mart	165013	193116	219672	246525	53409
2	exxon-mobil	182529	227596	208715	200949	-26647
3	general-motors	176558	184632	177260	186763	2131
4	general-electric	111630	129853	125913	131698	1845
5	citigroup	94396	111826	111202	92556	-19270
6	ibm	87548	85089	83067	81186	-3903
7	philip-morris	78596	80356	80879	80408	52
8	home-depot	38434	45738	53553	58247	12509
9	hewlett-packard	42370	48870	45226	56588	7718
10	boeing	57993	51321	58198	54069	2748
11	merck	32714	40363	47715	51790	11427
12	jp-morgan-chase	51852	60317	50723	43372	-16945
13	sbc-communications	49531	51374	45908	43138	-8236
14	proctor-and-gamble	38125	39951	39244	40238	287
15	at-and-t	54973	46850	42197	37827	-9023
16	johnson-and-johnson	28007	29172	32317	36298	7126
17	microsoft	19747	22956	25296	28365	5409
18	united-technologies	24127	26583	27897	28212	1629
19	intel	29389	33726	26539	26764	-6962
20	walt-disney	23435	25325	25172	25329	4
21	international-paper	24573	28180	26363	24976	-3204
22	dupont	26918	28268	24726	24006	-4262
23	am-express	21278	23675	22582	23807	132
24	honeywell	23735	25023	23652	22274	-2749
25	alcoa	16323	22659	22497	20263	-2396
26	caterpillar	19702	20175	20450	20152	-23
27	coca-cola	14406	14659	15605	16889	2230
28	3m	15748	16699	16054	16332	-367
29	mcdonalds	13259	14243	14870	15405	1162
30	eastman-kodak	14089	13994	13229	12835	-1159
31	riaa	14584	14323	13740	12614	-1709
32	riaa-cd	12816	13214	12909	12044	-1170

A very cluttered graph of this data can be found here (87KB). A slightly less cluttered graph that only examines the bottom twenty-five companies in close-up can be found here (87KB).

What can we tell from the raw sales data? The first is that the RIAA comprises a very small percentage of the overall economy. The DJIA company with the largest sales -- Wal-Mart -- had sales in 2002 approximately twenty times that of all RIAA-member sales. The next is that the RIAA net sales have indeed dropped by one-and-three-quarter billion dollars per year since its peak in 2000, with slightly over two-thirds of that (68.5 percent) being a decline in CD sales.

This observation brings up two questions.

1. With claimed losses of four billion dollars annually, but only an observed change of less than two billion dollars since 2000, was the RIAA already suffering two-and-a-quarter billion dollars of losses per year to digital piracy in 2000?
2. While CD sales accounted for over 95.5 percent of RIAA sales in 2002, they only account for 68.5 percent of the decline in sales. The other 31.5 percent decline in sales -- comprising $539 million -- comes from the other 4.5 percent of sales -- originally $1.1 billion in 2000. Where are these losses coming from?

I would venture that the answers to those questions are

1. We don't really know. Of course the RIAA will gladly make that claim, blaming those losses on Napster and other P2P tools. But is that actually the case? I would venture to say "no", given that the sales figures for the RIAA exhibited a steady climb over the 1990's, peaking at 2000. For there to be another two billion dollars of lost sales on top of their record-breaking year is unlikely.
2. The answer to this one is much easier. Let's prune out CD sales from the RIAA data and look at those trends from 1990 - 2000. This includes cassettes, vinyl records, music videos, and DVD sales. We'll plot each one separately, and the sum total.

Non-CD Sales, 1990 - 2000



We can clearly see that sales of non-CD items -- DVDs excluded -- have either barely held steady or have plunged dramatically since 1990. Net sales from these items plunged from nearly four billion dollars in 1990 to slightly over one-and-three-quarters billion dollars in 1998, the year that Napster went online. That's an average of $250 million per year in declining sales before P2P even got off the ground. Thus, the losses of $530 million in non-CD sales over two years (2000 to 2002) should come as no surprise.

Normalized Figures

The RIAA press people routinely speak of the percentage decline in sales: four percent this year, ten percent that year. The section above references percentages of sales and percentages of lost sales. This section will flush out the concept of percent changes in sales over time for all thirty-two companies and organizations. We do this in an attempt to provide a common yardstick by which we can measure the DJIA companies -- representative of the economy as a whole -- against the change in sales for the RIAA. It would be somewhat unfair otherwise, as Wal-Mart's sales alone increased by over fifty-three billion dollars from 2000 - 2002, and Exxon-Mobil's sales decreased by over twenty-six billion during that period; a two billion dollar change in RIAA-member sales nearly insignificant.

For each company/category, we normalized net sales for each company to their net sales in 2000. Thus, if a companies sales decreased ten percent between 2000 and 2001, and another ten percent between 2001 and 2002, their normalized sales chart would look like this

Company	2000	2001	2002	Change
Jim's Gizmos	100	90	81	-19%

Showing an overall drop in net sales of 19 percent between 2000 and 2002.

Table 2

Sales Figures Normalized to 2000 Sales (per-company)

Index	Company	1999	2000	2001	2002	% Change
1	merck	81.04	100.00	118.21	128.31	28.31
2	wal-mart	85.44	100.00	113.75	127.65	27.65
3	home-depot	84.03	100.00	117.08	127.34	27.34
4	johnson-and-johnson	96.00	100.00	110.78	124.42	24.42
5	microsoft	86.02	100.00	110.19	123.56	23.56
6	hewlett-packard	86.69	100.00	92.54	115.79	15.79
7	coca-cola	98.27	100.00	106.45	115.21	15.21
8	mcdonalds	93.09	100.00	104.40	108.15	8.15
9	united-technologies	90.76	100.00	104.94	106.12	6.12
10	boeing	113.00	100.00	113.39	105.35	5.35
11	general-electric	85.96	100.00	96.96	101.42	1.42
12	general-motors	95.62	100.00	96.00	101.15	1.15
13	proctor-and-gamble	95.42	100.00	98.23	100.71	0.71
14	am-express	89.87	100.00	95.38	100.55	0.55
15	philip-morris	97.80	100.00	100.65	100.06	0.06
16	walt-disney	92.53	100.00	99.39	100.01	0.01
17	caterpillar	97.65	100.00	101.36	99.88	-0.12
18	3m	94.30	100.00	96.13	97.80	-2.20
19	ibm	102.88	100.00	97.62	95.41	-4.59
20	eastman-kodak	100.67	100.00	94.53	91.71	-8.29
21	riaa-cd	96.98	100.00	97.69	91.14	-8.86
22	alcoa	72.03	100.00	99.28	89.42	-10.58
23	honeywell	94.85	100.00	94.52	89.01	-10.99
24	international-paper	87.20	100.00	93.55	88.63	-11.37
25	exxon-mobil	80.19	100.00	91.70	88.29	-11.71
26	riaa	101.82	100.00	95.92	88.06	-11.94
27	dupont	95.22	100.00	87.46	84.92	-15.08
28	sbc-communications	96.41	100.00	89.36	83.96	-16.04
29	citigroup	84.41	100.00	99.44	82.76	-17.24
30	at-and-t	117.33	100.00	90.06	80.74	-19.26
31	intel	87.14	100.00	78.69	79.35	-20.65
32	jp-morgan-chase	85.96	100.00	84.09	71.90	-28.10

Another cluttered graph of this data is here (88KB).

Analysis

What does all this data mean? Are there trends we can spot? The number of companies whose sales increased since 2000 is slightly larger than the number of companies whose sales declined since 2000: sixteen to fifteen. Removing the non-CD-sales RIAA element, we got a distribution that looks like this

Categorized Percent Change in Sales, 2000 - 2002



However, if we claim that an increase in net sales of less than two to three percent is essentially "holding the course" in the face of inflation, cost-of-living increases, we get a slightly different picture.

• Ten companies posted an increase in sales.
• Six "held the course".
• Two exhibited a slight decline.
• Thirteen companies posted minor to major losses.

So how much of this is expected because of the economy, and how much is due to the RIAA in particular? If we perform a two-factor full factorial design without replication, we can see what kind of effect these two factors -- economy and RIAA-specific conditions -- have on CD sales of RIAA members. It is statistically shady to play with normalized values or averages, so instead we will work with the raw net sales figures for each company.

The first analysis of the raw data includes all thirty-two groups. For the original sales data, refer back to table 1. The average annual net sales over all thirty-two groups is $53.9 billion. In addition to this average, we calculate a column effect and row effect for each year and each group, respectively. The column effect indicates how values in that column tend to be different than the average -- in our case, how the economy effected net sales -- and the row effect indicates how values in that row tend to be different than the average -- in our case, whether or not that business tended to have better or worse net sales than the average over our sample group.

Thus, the average value plus the column effect (the economy that year) plus the row effect (how our business is doing) gives the expected value for that company for that year. The tables below list the average net sales, the row effects, and the column effects.

Average Sales Per Company Per Year (1999-2002)

Avg Annual Sales (millions of dollars)
53904

Average Effect of Economy Each Year (1999-2002)

Avg Effect of Economy (millions of dollars)
1999	2000	2001	2002
-3767	1412	888	1467

Average Effect of Each Business/Group (1999-2002)

Difference from Overall Average (millions of dollars)
Business/Group	Difference
wal-mart	152176.71
exxon-mobil	151042.46
general-motors	127398.46
general-electric	70868.71
citigroup	48590.21
ibm	30317.71
philip-morris	26154.96
home-depot	-4911.79
hewlett-packard	-5641.29
boeing	1490.46
merck	-10759.29
jp-morgan-chase	-2338.79
sbc-communications	-6417.04
proctor-and-gamble	-14515.29
at-and-t	-8443.04
johnson-and-johnson	-22456.29
microsoft	-29813.79
united-technologies	-27200.04
intel	-24800.29
walt-disney	-29089.54
international-paper	-27881.79
dupont	-27925.29
am-express	-31069.29
honeywell	-30233.79
alcoa	-33469.29
caterpillar	-33785.04
coca-cola	-38515.04
3m	-37696.54
mcdonalds	-39460.54
eastman-kodak	-40368.04
riaa	-40089.54
riaa-cd	-41159.04

Expected Net Sales & Error (1999-2002)

Business	1999		2000		2001		2002
	Expected	Error (abs)	Expected	Error (abs)	Expected	Error (abs)	Expected	Error (abs)
wal-mart	202314.15	37301.15	207493.15	14377.15	206969.52	12702.48	207549.18	38975.82
exxon-mobil	201179.90	18650.90	206358.90	21237.10	205835.27	2879.73	206414.93	5465.93
general-motors	177535.90	977.90	182714.90	1917.10	182191.27	4931.27	182770.93	3992.07
general-electric	121006.15	9376.15	126185.15	3667.85	125661.52	251.48	126241.18	5456.82
citigroup	98727.65	4331.65	103906.65	7919.35	103383.02	7818.98	103962.68	11406.68
ibm	80455.15	7092.85	85634.15	545.15	85110.52	2043.52	85690.18	4504.18
philip-morris	76292.40	2303.60	81471.40	1115.40	80947.77	68.77	81527.43	1119.43
home-depot	45225.65	6791.65	50404.65	4666.65	49881.02	3671.98	50460.68	7786.32
hewlett-packard	44496.15	2126.15	49675.15	805.15	49151.52	3925.52	49731.18	6856.82
boeing	51627.90	6365.10	56806.90	5485.90	56283.27	1914.73	56862.93	2793.93
merck	39378.15	6664.15	44557.15	4194.15	44033.52	3681.48	44613.18	7176.82
jp-morgan-chase	47798.65	4053.35	52977.65	7339.35	52454.02	1731.02	53033.68	9661.68
sbc-communications	43720.40	5810.60	48899.40	2474.60	48375.77	2467.77	48955.43	5817.43
proctor-and-gamble	35622.15	2502.85	40801.15	850.15	40277.52	1033.52	40857.18	619.18
at-and-t	41694.40	13278.60	46873.40	23.40	46349.77	4152.77	46929.43	9102.43
johnson-and-johnson	27681.15	325.85	32860.15	3688.15	32336.52	19.52	32916.18	3381.82
microsoft	20323.65	576.65	25502.65	2546.65	24979.02	316.98	25558.68	2806.32
united-technologies	22937.40	1189.60	28116.40	1533.40	27592.77	304.23	28172.43	39.57
intel	25337.15	4051.85	30516.15	3209.85	29992.52	3453.52	30572.18	3808.18
walt-disney	21047.90	2387.10	26226.90	901.90	25703.27	531.27	26282.93	953.93
international-paper	22255.65	2317.35	27434.65	745.35	26911.02	548.02	27490.68	2514.68
dupont	22212.15	4705.85	27391.15	876.85	26867.52	2141.52	27447.18	3441.18
am-express	19068.15	2209.85	24247.15	572.15	23723.52	1141.52	24303.18	496.18
honeywell	19903.65	3831.35	25082.65	59.65	24559.02	907.02	25138.68	2864.68
alcoa	16668.15	345.15	21847.15	811.85	21323.52	1173.48	21903.18	1640.18
caterpillar	16352.40	3349.60	21531.40	1356.40	21007.77	557.77	21587.43	1435.43
coca-cola	11622.40	2783.60	16801.40	2142.40	16277.77	672.77	16857.43	31.57
3m	12440.90	3307.10	17619.90	920.90	17096.27	1042.27	17675.93	1343.93
mcdonalds	10676.90	2582.10	15855.90	1612.90	15332.27	462.27	15911.93	506.93
eastman-kodak	9769.40	4319.60	14948.40	954.40	14424.77	1195.77	15004.43	2169.43
riaa	10047.90	4536.10	15226.90	903.90	14703.27	963.27	15282.93	2668.93
riaa-cd	8978.40	3837.60	14157.40	943.40	13633.77	724.77	14213.43	2169.43

Immediately we see a bit of a problem. The estimates are consistently off, oftentimes by large amounts; the 2002 sales estimates for Wal-Mart alone are off by almost forty billion dollars, more than the sales of the three smallest companies combined. First off, why is this happening; secondly, what can we do about it?

Let us assume that the largest companies are skewing the results, simply given the scale of their sales. Wal-Mart is large enough that a small blip in their sales -- even less than one percent -- is more than a two-and-a-half billion dollar shift in the statistics. This is natural given the scale of their distribution chain. But what can be done?

Let us assume that it is more worthwhile to work with companies whose net sales are more on par with those of the recording industries. As an arbitrary cutoff value, let us remove any companies whose 2002 sales exceeded fifty billion dollars, the amount by which Wal-Mart's sales grew from 2000 - 2002. This pruning still leaves us with a broad cross-section of the market: banks, construction equipment manufacturers, the IT sector, and food vendors.

Recalculating the data, we get the following.

Average Sales Per Company Per Year, Take 2 (1999-2002)

Avg Annual Sales (millions of dollars)
25965

Average Effect of Economy Each Year, Take 2 (1999-2002)

Avg Effect of Economy (millions of dollars)
1999	2000	2001	2002
-412	1242	-157	-673

Average Effect of Each Business/Group, Take 2 (1999-2002)

Difference from Overall Average (millions of dollars)
Business/Group	Difference
jp-morgan-chase	25600.61
sbc-communications	21522.36
proctor-and-gamble	13424.11
at-and-t	19496.36
johnson-and-johnson	5483.11
microsoft	-1874.39
united-technologies	739.36
intel	3139.11
walt-disney	-1150.14
international-paper	57.61
dupont	14.11
am-express	-3129.89
honeywell	-2294.39
alcoa	-5529.89
caterpillar	-5845.64
coca-cola	-10575.64
3m	-9757.14
mcdonalds	-11521.14
eastman-kodak	-12428.64
riaa	-12150.14
riaa-cd	-13219.64

Expected Net Sales & Error, Take 2 (1999-2002)

Business	1999		2000		2001		2002
	Expected	Error (abs)	Expected	Error (abs)	Expected	Error (abs)	Expected	Error (abs)
jp-morgan-chase	51153.80	698.20	52808.51	7508.49	51408.70	685.70	50892.99	7520.99
sbc-communications	47075.55	2455.45	48730.26	2643.74	47330.45	1422.45	46814.74	3676.74
proctor-and-gamble	38977.30	852.30	40632.01	681.01	39232.20	11.80	38716.49	1521.51
at-and-t	45049.55	9923.45	46704.26	145.74	45304.45	3107.45	44788.74	6961.74
johnson-and-johnson	31036.30	3029.30	32691.01	3519.01	31291.20	1025.80	30775.49	5522.51
microsoft	23678.80	3931.80	25333.51	2377.51	23933.70	1362.30	23417.99	4947.01
united-technologies	26292.55	2165.55	27947.26	1364.26	26547.45	1349.55	26031.74	2180.26
intel	28692.30	696.70	30347.01	3378.99	28947.20	2408.20	28431.49	1667.49
walt-disney	24403.05	968.05	26057.76	732.76	24657.95	514.05	24142.24	1186.76
international-paper	25610.80	1037.80	27265.51	914.49	25865.70	497.30	25349.99	373.99
dupont	25567.30	1350.70	27222.01	1045.99	25822.20	1096.20	25306.49	1300.49
am-express	22423.30	1145.30	24078.01	403.01	22678.20	96.20	22162.49	1644.51
honeywell	23258.80	476.20	24913.51	109.49	23513.70	138.30	22997.99	723.99
alcoa	20023.30	3700.30	21678.01	980.99	20278.20	2218.80	19762.49	500.51
caterpillar	19707.55	5.55	21362.26	1187.26	19962.45	487.55	19446.74	705.26
coca-cola	14977.55	571.55	16632.26	1973.26	15232.45	372.55	14716.74	2172.26
3m	15796.05	48.05	17450.76	751.76	16050.95	3.05	15535.24	796.76
mcdonalds	14032.05	773.05	15686.76	1443.76	14286.95	583.05	13771.24	1633.76
eastman-kodak	13124.55	964.45	14779.26	785.26	13379.45	150.45	12863.74	28.74
riaa	13403.05	1180.95	15057.76	734.76	13657.95	82.05	13142.24	528.24
riaa-cd	12333.55	482.45	13988.26	774.26	12588.45	320.55	12072.74	28.74

This model predicted the CD sales of RIAA members to within thirty million dollars (less than one-quarter of one percent) given the performance of the economy in 2002, as a function of nineteen other similarly-sized corporate sales, and the performance of the RIAA three previous years.

Conclusions (?)

So what exactly does this tell us? Perhaps not as much as we'd like. This particular analysis does not tell us exactly how accurate the rest of the model is, and several other professional statistician shortcomings. Remember, there are lies, damned lies, and statistics; this is just another statistic. I would assert, however that it does make the case in cold, hard numbers that the RIAA's claim of digital piracy ravaging their sales must be taken with a rather large grain of salt. The CEOs of Eastman-Kodak are in a nearly identical economic situation as the RIAA, yet do not have the luxury of blaiming digital piracy.

But perhaps they should try. It'd be interesting.

-jdm

P.S.: Feel free to contact me with comment, suggestions, corrections, or just to say "hello": justin at cs dot duke dot edu.

P.P.S.: You can download the original spreadsheets for both the complete coporate set and the "small"-business corporate set in Microsoft Excel format and comma-separated-values (CSV) format.

• Full Corporate Spreadsheet: Excel (30KB), CSV (5KB).
• Small Corporate Spreadsheet: Excel (26KB), CSV (3KB).