Physics

Someone will tell you nature runs on 80/20. A small cause, most of the effect, in every lab.

About 20% of a pile sometimes carries about 80% of the total. Sometimes the split is harsher. Sometimes it is softer. The ratio is not a constant of nature.

Physics is full of heavy tails. It does not owe you one magic fraction. Name the pile before you quote the slogan.

The slogan is one exponent

Mark Newman reviewed where power laws show up, in physics and far outside it (Contemporary Physics, 2005; arXiv:cond-mat/0412004). A power law means the chance of a large value falls as that value raised to a power. Pareto's wealth curve is one of those laws. Zipf's word counts are another. They are the same family. They are not the same split.

Newman takes the exponents people had already fitted and asks what share the top holds. For wealth, the calculation lands near the famous line: about 80% of the wealth in the richest 20%. He notes that more detailed wealth counts agree. That is the 80/20 rule as a result for that exponent, not as a stamp for every histogram.

80/20 example: wealth, in Newman's calculation, is the case that really is about 80 and 20. The same page of the review immediately leaves that case behind.

The next piles are not 80 and 20

Using the same method, Newman writes that the top 20% of websites get about two-thirds of the hits, not 80%. The largest 10% of U.S. cities hold about 60% of the population. Both are concentration. Neither is the slogan. If you force them into 80/20, you are editing the exponent to fit a proverb.

What concentrates is the tail. A few sites, a few cities, a few fortunes. What does not concentrate is the ratio. It moves when the exponent moves. The wealth page is about the fortune case, where the proverb happens to land. Do not carry that proverb back into the lab as if the cities had agreed to it.

Quakes follow a power law, not a proverb

Newman plots earthquake sizes the same way. California magnitudes from 1910 to 1992 fall on a power-law plot. Magnitude is already a logarithm of size, so a straight line there means a heavy tail in the energy. Many small quakes. Few large ones. The large ones matter out of proportion to how often they appear.

Read the plot this way. On log paper, a power law is a straight line. The slope is the exponent. A steep line means the giants are rarer. A shallow line means the giants still show up often enough to own the total. Newman's point about exponents at or below 2 is the sharp version: the average is then set by the largest events in whatever window you kept. Cut the biggest quake out of the year and the average quake changes. That is the opposite of a classroom mean, where one outlier is a nuisance.

He does not convert that plot into "20% of quakes release 80% of the energy." When the exponent is small enough, the average is dominated by the largest events in the window you chose. Change the window, or the upper cutoff, and the share moves. A catalog with one great quake will look steeper than a quiet decade. That is the finding. The proverb would hide it.

Solar flares and crater sizes sit on the same kind of plot in the review. The lesson repeats. Count the events. Fit the slope. Then say what share the tail holds. Skip the middle step and you will report 80/20 because you already believed it.

Why the tail keeps appearing

One mechanism Newman spends time on is self-organized criticality, from Bak, Tang, and Wiesenfeld. Some systems drift until they sit at a critical point, the way a sandpile slips when one more grain is too many. The slips come in all sizes. The size counts follow a power law without anyone tuning a dial to a special value. Forest-fire models do a similar thing. Newman is careful: a model that makes a power law is not a claim that the forest is the model.

The useful part for a reader is smaller. Uneven totals are ordinary in these systems. A single ratio is not. If someone says physics proved 80/20, ask which exponent, and which total. Energy, money, hits, and counts of words are different totals. They do not have to share a fraction.

Three lines before you say 80/20

• What is one item? A quake, a city, a dollar of wealth, a hit.
• What is the total? Energy, population, money, clicks.
• Did anyone fit the slope, or did the sentence arrive with 80 and 20 already in it?

8020 move: if the third line is blank, do not publish the ratio. Describe the tail, or go measure it. The slogan is allowed only when the exponent actually produces it, as Newman shows for wealth, and not for the websites on the same page.

Bring this to a dataset you already have, not to a new belief. Take last year's incidents, invoices, or sensor spikes. Sort them. Ask what share of the total sits in the top tenth, and in the top fifth. Write those two shares down. If they are 60% and 70%, say that. If they are 90% and 95%, say that. Only if they land near 80 and 20 do you get to use the proverb, and even then name the total you summed. Newman did this for wealth, for hits, and for city population, and he got three different answers. Copy the method. Do not copy the first answer onto the other two.

A student lab fails this in a familiar way. Ten runs, nine of them quiet, one of them huge. The average gets reported as if the ten runs agreed. The power-law habit is to show the tail beside the mean. Say which run owns the sum. Then decide whether that run is the phenomenon or a broken instrument. Those are different decisions. The proverb collapses them into "focus on the vital few," which is how a bad sensor survives the write-up. The tail is a clue. It is not permission to drop the other measurements before you know why they are small.

If you teach or explain this, lead with the counterexamples Newman already printed. Websites are not 80/20. Cities are not 80/20. Wealth is close, for the exponent he used, and even wealth is steeper in some countries than in others. The honest sentence is short. Many totals live in a few events. The fraction is a result, not a law. Anyone who needs the fraction can fit the line. Anyone who will not fit the line should not quote 80 and 20.

One more caution, because the review is easy to over-read. Newman is collecting reported plots, not running a new experiment in your building. Some famous straight lines on log paper later turned out to be log-normal, or a power law only in the middle, with a cutoff the chart cropped off. If your data bend, believe the bend. A forced straight line is the same mistake as a forced 80/20. The question to take home is modest. Is the total owned by the largest few, and by how much, in this window? Answer with the shares you calculated. Leave the proverb for the wealth exponent, where he showed it belongs.

Keep a scrap of paper with the three lines for one real pile you care about this month. A budget, a bug list, a set of storm dates, a lab notebook. Fill the shares. If you cannot fill them, you are not ready to teach the ratio. That scrap is the whole assignment. It is also the correction to the old page, which started from 80 and 20 and went looking for a science that would agree.

Misreads that flatten the tail

"Every power law is 80/20." Newman's own examples are 80/20, about two-thirds, and about 60%. The family is real. The fraction is local.

"A few big quakes are exactly a fifth of the catalog." The plot says they are rare and they dominate. It does not say they are 20% of the count. Rarity is the point.

"The lab should ignore 80% of the measurements." A heavy tail tells you where the total sits. It does not tell you which instrument to leave uncalibrated. Safety and error bars are not a proverb.