Category Archives: Politics

How do we decide how many representatives there are for each state?

The US House of Representatives has 435 voting members (and 6 non-voting members: one each from Washington DC, Puerto Rico, American Samoa, Guam, the Northern Mariana Islands, and the US Virgin Islands). Roughly speaking, the higher the population of a state is, the more representatives it should have.

But what does this really mean?

If we looked at the US Constitution to make this clear, we would find little help. The third clause of Article I, Section II of the Constitution says

Representatives and direct Taxes shall be apportioned among the several States which may be included within this Union, according to their respective Numbers … The number of Representatives shall not exceed one for every thirty thousand, but each state shall have at least one Representative.

This doesn’t give clarity.1 In fact, uncertainty surrounding proper apportionment of representatives led to the first presidential veto.

The Apportionment Act of 1792

According to the 1790 Census, there were 3199415 free people and 694280 slaves in the United States.2

When Congress sat to decide on apportionment in 1792, they initially computed the total (weighted) population of the United States to be 3199415 + (3/5)⋅694280 ≈ 3615923. They noted that the Constitution says there should be no more than 1 representative for every 30000, so they divided the total population by 30000 and rounded down, getting 3615983/30000 ≈ 120.5.

Thus there were to be 120 representatives. If one takes each state and divides their populations by 30000, one sees that the states should get the following numbers of representatives3

State          ideal    rounded_down
Vermont        2.851    2
NewHampshire   4.727    4
Maine          3.218    3
Massachusetts  12.62    12
RhodeIsland    2.281    2
Connecticut    7.894    7
NewYork        11.05    11
NewJersey      5.985    5
Pennsylvania   14.42    14
Delaware       1.851    1
Maryland       9.283    9
Virginia       21.01    21
Kentucky       2.290    2
NorthCarolina  11.78    11
SouthCarolina  6.874    6
Georgia        2.361    2

But here is a problem: the total number of rounded down representatives is only 112. So there are 8 more representatives to give out. How did they decide which to assign these representatives to? They chose the 8 states with the largest fractional “ideal” parts:

  1. New Jersey (0.985)
  2. Connecticut (0.894)
  3. South Carolina (0.874)
  4. Vermont (0.851)
  5. Delaware (0.851)
  6. Massachusetts+Maine (0.838)
  7. North Carolina (0.78)
  8. New Hampshire (0.727)

(Maine was part of Massachuestts at the time, which is why I combine their fractional parts). Thus the original proposed apportionment gave each of these states one additional representative. Is this a reasonable conclusion?

Perhaps. But these 8 states each ended up having more than 1 representative for each 30000. Was this limit in the Constitution meant country-wide (so that 120 across the country is a fine number) or state-by-state (so that, for instance, Delaware, which had 59000 total population, should not be allowed to have more than 1 representative)?

There is the other problem that New Jersey, Connecticut, Vermont, New Hampshire, and Massachusetts were undoubtedly Northern states. Thus Southern representatives asked, Is it not unfair that the fractional apportionment favours the North?4

Regardless of the exact reasoning, the Secretary of State Thomas Jefferson and Attorney General Edmond Randalph (both from Virginia) urged President Washington to veto the bill, and he did. This was the first use of the Presidential veto.

Afterwards, Congress got together and decided on starting with 33000 people per representative and ignoring fractional parts entirely. The exact method became known as the Jefferson Method of Apportionment, and was used in the US until 1830. The subtle part of the method involves deciding on the number 33000. In the US, the exact number of representatives sometimes changed from election to election. This number is closely related to the population-per-representative, but these were often chosen through political maneuvering as opposed to exact decision.

As an aside, it’s interesting to note that this method of apportionment is widely used in the rest of the world, even though it was abandoned in the US.5 In fact, it is still used in Albania, Angola, Argentina, Armenia, Aruba, Austria, Belgium, Bolivia, Brazil, Bulgaria, Burundi, Cambodia, Cape Verde, Chile, Colombia, Croatia, the Czech Republic, Denmark, the Dominican Republic, East Timor, Ecuador, El Salvador, Estonia, Fiji, Finland, Guatemala, Hungary, Iceland, Israel, Japan, Kosovo, Luxembourg, Macedonia, Moldova, Monaco, Montenegro, Mozambique, Netherlands, Nicaragua, Northern Ireland, Paraguay, Peru, Poland, Portugal, Romania, San Marino, Scotland, Serbia, Slovenia, Spain, Switzerland, Turkey, Uruguay, Venezuela and Wales — as well as in many countries for election to the European Parliament.

Apportionment Act of 1792

Measuring the fairness of an apportionment method

At the core of different ideas for apportionment is fairness. How can we decide if an apportionment fair?

We’ll consider this question in the context of the post-1911 United States — after the number of seats in the House of Representatives was established. This number was set at 433, but with the proviso that anticipated new states Arizona and New Mexico would each come with an additional seat.6

So given that there are 435 seats to apportion, how might we decide if an apportionment is fair? Fundamentally, this should relate to the number of people each representative actually represents.

For example, in the 1792 apportionment, the single Delawaran representative was there to represent all 55000 of its population, while each of the two Rhode Island representatives corresponded to 34000 Rhode Islanders. Within the House of Representatives, it was as though the voice of each Delawaran only counted 61 percent as much as the voice of each Rhode Islander7

The number of people each representative actually represent is at the core of the notion of fairness — but even then, it’s not obvious.

Suppose we enumerate the states, so that Si refers to state i. We’ll also denote by Pi the population of state i, and we’ll let Ri denote the number of representatives allotted to state i.

In the ideal scenario, every representative would represent the exact same number of people. That is, we would have
$$\text{pop. per rep. in state i}
= \frac{P_i}{R_i}
= \frac{P_j}{R_j}
= \text{pop. per rep. in state j}$$

for every pair of states i and j. But this won’t ever happen in practice.

Generally, we should expect $\frac{P_i}{R_i} \neq \frac{P_j}{R_j}$ for every pair of distinct states. If
$$
\frac{P_i}{R_i} > \frac{P_j}{R_j}, \tag{1}
$$

then we can say that each representative in state i represents more people, and thus those people have a diluted vote.

Amounts of Inequality

There are lots of pairs of states. How do we actually measure these inequalities? This would make an excellent question in a statistics class (illustrating how one can answer the same question in different, equally reasonable ways) or even a civics class.

A few natural ideas emerge:

  • We might try to minimize the differences of constituency size: $\left \lvert \frac{P_i}{R_i} – \frac{P_j}{R_j} \right \rvert$.
  • We might try to minimize the differences in per capita representation: $\left \lvert \frac{R_i}{P_i} – \frac{R_j}{P_j} \right \rvert$.
  • We might take overall size into account, and try to minimize both the relative constituency size and relative difference in per capita representation.

This last one needs a bit of explanation. Define the relative difference between two numbers x and y to be
$$
\frac{\lvert x – y \rvert}{\min(x, y)}.
$$

Suppose that for a pair of states, we have that $(1)$ holds, i.e. that representatives in state j have smaller constituencies than in state i (and therefore people in state j have more powerful votes). Then the relative difference in constituency size is
$$
\frac{P_i/R_i – P_j/R_j}{P_j/R_j} = \frac{P_i/R_i}{P_j/R_j} – 1.
$$

The relative difference in per capita representation is
$$
\frac{R_j/P_j – R_i/P_i}{R_i/P_i} = \frac{R_j/P_j}{R_i/P_i} – 1 =
\frac{P_i/R_i}{P_j/R_j} – 1.
$$

Thus these are the same! By accounting for differences in size by taking relative proportions, we see that minimizing relative difference in constituency size and minimizing relative difference in per capita representation are actually the same.

All three of these measures seem reasonable at first inspection. Unfortunately, they all give different apportionments (and all are different from Jefferson’s scheme — though to be fair, Jefferson’s scheme doesn’t seek to minimize inequality and there is no reason to think it should behave the same).

Each of these ideas leads to a different apportionment scheme, and in fact each has a name.

  • Minimizing differences in constituency size is the Dean method.
  • Minimizing differences in per capita representation is the Webster method.
  • Minimizing relative differences between both constituency size and per capita representation is the Hill (or sometimes Huntington-Hill) method.

Further, each of these schemes has been used at some time in US history. Webster’s method was used immediately after the 1840 census, but for the 1850 census the original Alexander Hamilton scheme (the scheme vetoed by Washington in 1792) was used. In fact, the Apportionment Act of 1850 set the Hamilton method as the primary method, and this was nominally used until 1900.8 The Webster method was used again immediately after the 1910 census. Due to claims of incomplete and inaccurate census counts, no apportionment occurred based on the 1920 census.9

In 1929 an automatic apportionment act was passed.10 In it, up to three different apportionment schemes would be provided to Congress after each census, based on a total of 435 seats:

  1. The apportionment that would come from whatever scheme was most recently used. (In 1930, this would be the Webster method).
  2. The apportionment that would come from the Webster method.
  3. The apportionment that would come from the newly introduced Hill method.

If one reads congressional discussion from the time, then it will be good to note that Webster’s method is sometimes called the method of major fractions and Hill’s method is sometimes called the method of equal proportions. Further, in a letter written by Bliss, Brown, Eisenhart, and Pearl of the National Academy of Sciences, Hill’s method was declared to be the recommendation of the Academy.11 From 1930 on, Hill’s method has been used.

Why use the Hill method?

The Hamilton method led to a few paradoxes and highly counterintuitive behavior that many representatives found disagreeable. In 1880, a paradox now called the Alabama paradox was noted. When deciding on the number of representatives that should be in the House, it was noted that if the House had 299 members, Alabama would have 8 representatives. But if the House had 300 members, Alabama would have 7 representatives — that is, making one more seat available led to Alabama receiving one fewer seat.

The problem is the fluctuating relationships between the many fractional parts of the ideal number of representatives per state (similar to those tallied in the table in the section The Apportionment Act of 1792).

Another paradox was discovered in 1900, known as the Population paradox. This is a scenario in which a state with a large population and rapid growth can lose a seat to a state with a small population and smaller population growth. In 1900, Virginia lost a seat to Maine, even though Virginia’s population was larger and growing much more rapidly.

In particular, in 1900, Virginia had 1854184 people and Maine had 694466 people, so Virginia had 2.67 times the population as Maine. In 1901, Virginia had 1873951 people and Maine had 699114 people, so Virginia had 2.68 times the number of people. And yet Hamilton apportionment would have given 10 seats to Virginia and 3 to Maine in 1900, but 9 to Virginia and 4 to Maine in 1901.

Central to this paradox is that even though Virginia was growing faster than Maine the rest of the nation was growing fast still, and proportionally Virginia lost more because it was a larger state. But it’s still paradoxical for a state to lose a representative to a second state that is both smaller in population and is growing less rapidly each census.12

The Hill method can be shown to not suffer from either the Alabama paradox or the Population paradox. That it doesn’t suffer from these paradoxical behaviours and that it seeks to minimize a meaningful measure of inequality led to its adoption in the US.13

Understanding the modern Hill method in practice

Since 1930, the US has used the Hill method to apportion seats for the House of Representatives. But as described above, it may be hard to understand how to actually apply the Hill method. Recall that Pi is the population of state i, and Ri is the number of representatives allocated to state i. The Hill method seeks to minimize
$$
\frac{P_i/R_i – P_j/R_j}{P_j/R_j} = \frac{P_i/R_i}{P_j/R_j} – 1
$$

whenever Pi/Ri > Pj/Rj. Stated differently, the Hill method seeks to guarantee the smallest relative differences in constituency size.

We can work out a different way of understanding this apportionment that is easier to implement in practice.

Suppose that we have allocated all of the representatives to each state and state j has Rj representatives, and suppose that this allocation successfully minimizes relative differences in constituency size. Take two different states i and j with Pi/Ri > Pj/Rj. (If this isn’t possible then the allocation is perfect).

We can ask if it would be a good idea to move one representative from state j to state i, since state j‘s constituency sizes are smaller. This can be thought of as working with Ri′=Ri + 1 and Rj′=Rj − 1. If this transfer lessens the inequality then it should be made — but since we are supposing that the allocation successfully minimizes relative difference in constituency size, we must have that the inequality is at least as large. This necessarily means that Pj/Rj′>Pi/Ri (since otherwise the relative difference is strictly smaller) and
$$
\frac{P_jR_i’}{P_iR_j’} – 1 \geq \frac{P_iR_j}{P_jR_i} – 1
$$

(since the relative difference must be at least as large). This is equivalent to
$$
\frac{P_j(R_i+1)}{P_i(R_j-1)} \geq \frac{P_iR_j}{P_jR_i}
\iff
\frac{P_j^2}{(R_j-1)R_j} \geq \frac{P_i^2}{R_i(R_i+1)}.
$$

As every variable is positive, we can rewrite this as
$$
\frac{P_j}{\sqrt{(R_j – 1)R_j}} \geq \frac{P_i}{\sqrt{R_i(R_i+1)}}. \tag{2}
$$

We’ve shown that $(2)$ must hold whenever Pi/Ri > Pj/Rj in a system that minimizes relative difference in constituency size. But in fact it must hold for all pairs of states i and j.

Clearly it holds if i = j as the denominator on the left is strictly smaller.

If we are in the case when Pj/Rj > Pi/Ri, then we necessarily have the chain Pj/(Rj − 1)>Pj/Rj > Pi/Ri > Pi/(Ri + 1). Multiplying the inner and outer inequalities shows that $(2)$ holds trivially in this case.

This inequality shows that the greatest obstruction to being perfectly apportioned as per Hill’s method is the largest fraction
$$ \frac{R_i}{\sqrt{P_i(P_i+1)}} $$
being too large. (Some call this term the Hill rank-index).

An iterative Hill apportionment

This observation leads to the following iterative construction of a Hill apportionment. Initially, assign every state 1 representative (since by the Constitution, each state gets at least one representative). Then, given an apportionment for n seats, we can get an apportionment for n + 1 seats by assigning the additional seat the any state i which maximizes the Hill rank-index $R_i/\sqrt{P_i(P_i+1)}$.

Further, it can be shown that there is a unique apportionment in Hill’s method (except for ties in the Hill rank-index, which are exceedingly rare in practice). Thus the apportionment is unique.

This is very quickly and easily implemented in code. In a later note, I will share the code I used to compute the various data for this note, as well as an implementation of Hill apportionment.

Additional notes: Consequences of the 1870 and 1990 Apportionments

The 1870 Apportionment

Officially, Dean’s method of apportionment has never been used. But it was perhaps used in 1870 without being described. Officially, Hamilton’s method was in place and the size of the House was agreed to be 292. But the actual apportionment that occurred agreed with Dean’s method, not Hamilton’s method. Specifically, New York and Illinois were each given one fewer seat than Hamilton’s method would have given, while New Hampshire and Florida were given one additional seat each.

There are many circumstances surrounding the 1870 census and apportionment that make this a particularly convoluted time. Firstly, the US had just experienced its Civil War, where millions of people died and millions others moved or were displaced. Animosity and reconstruction were both in full swing. Secondly, the US passed the 14th amendment in 1868, so that suddenly the populations of Southern states grew as former slaves were finally allowed to be counted fully.

One might think that having two pairs of states swap a representative would be mostly inconsequential. But this difference — using Dean’s method instead of the agreed on Hamilton method, changed the result of the 1876 Presidential election. In this election, Samuel Tilden won New York while Rutherford B. Hayes won Illinois, New Hampshire, and Florida. As a result, Tilden received one fewer electoral vote and Hayes received one additional electoral vote — and the total electoral voting in the end had Hayes win with 185 votes to Tilden’s 184.

There is still one further mitigating factor, however, that causes this to be yet more convoluted. The 1876 election is perhaps the most disputed presidential election. In Florida, Louisiana, and South Carolina, each party reported that its candidate had won the state. Legitimacy was in question, and it’s widely believed that a deal was struck between the Democratic and Republican parties (see wikipedia and 270 to win). As a result of this deal, the Republican candidate Rutherford B. Hayes would gain all disputed votes and remove federal troops (which had been propping up reconstructive efforts) from the South. This marked the end of the “Reconstruction” period, and allowed the rise of the Democratic Redeemers (and their rampant black voter disenfranchisement) in the South.

The 1990 Apportionment

Similar in consequence though not in controversy, the apportionment of 1990 influenced the results of the 2000 presidential election between George W. Bush and Al Gore (as the 2000 census is not complete before the election takes place, so the election occurs with the 1990 electoral college sizes). The modern Hill apportionment method was used, as it has been since 1930. But interestingly, if the originally proposed Hamilton method of 1792 was used, the electoral college would have been tied at 26914. If Jefferson’s method had been used, then Gore would have won with 271 votes to Bush’s 266.

These decisions have far-reaching consequences!

Sources

  1. Balinski, Michel L., and H. Peyton Young. Fair representation: meeting the ideal of one man, one vote. Brookings Institution Press, 2010.
  2. Balinski, Michel L., and H. Peyton Young. “The quota method of apportionment.” The American Mathematical Monthly 82.7 (1975): 701-730.
  3. Bliss, G. A., Brown, E. W., Eisenhart, L. P., & Pearl, R. (1929). Report to the President of the National Academy of Sciences. February, 9, 1015-1047.
  4. Crocker, R. House of Representatives Apportionment Formula: An Analysis of Proposals for Change and Their Impact on States. DIANE Publishing, 2011.
  5. Huntington, The Apportionment of Representatives in Congress, Transactions of the American Mathematical Society 30 (1928), 85–110.
  6. Peskin, Allan. “Was there a Compromise of 1877.” The Journal of American History 60.1 (1973): 63-75.
  7. US Census Results
  8. US Constitution
  9. US Congressional Record, as collected at https://memory.loc.gov/ammem/amlaw/lwaclink.html
  10. George Washington’s collected papers, as archived at https://web.archive.org/web/20090124222206/http://gwpapers.virginia.edu/documents/presidential/veto.html
  11. Wikipedia on the Compromise of 1877, at https://en.wikipedia.org/wiki/Compromise_of_1877
  12. Wikipedia on Arthur Vandenberg, at https://en.wikipedia.org/wiki/Arthur_Vandenberg
Posted in Data, Expository, Mathematics, Politics, Story | Tagged , , | Leave a comment

Segregation, Gerrymandering, and Schelling’s Model

[This note is more about modeling some of the mathematics behind political events than politics themselves. And there are pretty pictures.]

Gerrymandering has become a recurring topic in the news. The Supreme Court of the US, as well as more state courts and supreme courts, is hearing multiple cases on partisan gerrymandering (all beginning with a case in Wisconsin).

Intuitively, it is clear that gerrymandering is bad. It allows politicians to choose their voters, instead of the other way around. And it allows the majority party to quash minority voices.

But how can one identify a gerrymandered map? To quote Justice Kennedy in his Concurrence the 2004 Supreme Court case Vieth v. Jubelirer:

When presented with a claim of injury from partisan gerrymandering, courts confront two obstacles. First is the lack of comprehensive and neutral principles for drawing electoral boundaries. No substantive definition of fairness in districting seems to command general assent. Second is the absence of rules to limit and confine judicial intervention. With uncertain limits, intervening courts–even when proceeding with best intentions–would risk assuming political, not legal, responsibility for a process that often produces ill will and distrust.

Later, he adds to the first obstacle, saying:

The object of districting is to establish “fair and effective representation for all citizens.” Reynolds v. Sims, 377 U.S. 533, 565—568 (1964). At first it might seem that courts could determine, by the exercise of their own judgment, whether political classifications are related to this object or instead burden representational rights. The lack, however, of any agreed upon model of fair and effective representation makes this analysis difficult to pursue.

From Justice Kennedy’s Concurrence emerges a theme — a “workable standard” of identifying gerrymandering would open up the possibility of limiting partisan gerrymandering through the courts. Indeed, at the core of the Wisconsin gerrymandering case is a proposed “workable standard”, based around the efficiency gap.

 

Thomas Schelling and Segregation

In 1971, American economist Thomas Schelling (who later won the Nobel Prize in Economics in 2005) published Dynamic Models of Segregation (Journal of Mathematical Sociology, 1971, Vol 1, pp 143–186). He sought to understand why racial segregation in the United States seems so difficult to combat.

He introduced a simple model of segregation suggesting that even if each individual person doesn’t mind living with others of a different race, they might still choose to segregate themselves through mild preferences. As each individual makes these choices, overall segregation increases.

I write this post because I wondered what happens if we adapt Schelling’s model to instead model a state and its district voting map. In place of racial segregation, I consider political segregation. Supposing the district voting map does not change, I wondered how the efficiency gap will change over time as people further segregate themselves.

It seemed intuitive to me that political segregation (where people who had the same political beliefs stayed largely together and separated from those with different political beliefs) might correspond to more egregious cases of gerrymandering. But to my surprise, I was (mostly) wrong.

Let’s set up and see the model.

(more…)

Posted in Expository, Mathematics, Politics, Programming, Python | Tagged , , | Leave a comment

The Hawaiian Missile Crisis

I read an article from Doug Criss on CNN yesterday with the title “Hawaii’s governor couldn’t correct the false missile alert sooner because he forgot his Twitter password.”1 It turns out that Governor Ige knew within two minutes that the alert was a false alarm, but (in the words of the article) “he couldn’t hop on Twitter and tell everybody — because he didn’t know his password.”

There are a couple of different ways to take this story. The most common response I have seen is to blame the employee who accidentally triggered the alarm, and to forgive the Governor his error because who could have guessed that something like this would happen? The second most common response I see is a certain shock that the key mouthpiece of the Governor in this situation is apparently Twitter.

There is some merit to both of these lines of thought. Considering them in turn: it is pretty unfortunate that some employee triggered a state of hysteria by pressing an incorrect button (or something to that effect). We always hope that people with great responsibilities act with extreme caution (like thermonuclear war).

How about a nice game of global thermonuclear war?

So certainly some blame should be placed on the employee.

As for Twitter, I wonder whether or not a sarcasm filter has been watered down between the Governor’s initial remarks and my reading it in Doug’s article for CNN. It seems likely to me that this comment is meant more as commentary on the status of Twitter as the President’s preferred 2 medium of communicating with the People. It certainly seems unlikely to me that the Governor would both frequently use Twitter for important public messages and forget his Twitter credentials. Perhaps this is code for “I couldn’t get in touch with the person who manages my Twitter account” (because that person was hiding in a bunker?), but that’s not actually important. (more…)

Posted in Politics, Story | Tagged | Leave a comment

We begin bombing Korea in five minutes: Parallels to Reagan in 1984

On a day when President and Commander-in-Chief Donald Trump tweets belligerent messages aimed at North Korea, I ask: “Have we seen anything like this ever before?” In fact, we have. Let’s review a tale from Reagan.

August 11, 1984: President Reagan is preparing for his weekly NPR radio address. The opening line of his address was to be

My fellow Americans, I’m pleased to tell you that today I signed legislation that will allow student religious groups to begin enjoying a right they’ve too long been denied — the freedom to meet in public high schools during nonschool hours, just as other student groups are allowed to do.1

During the sound check, President Reagan joked

My fellow Americans, I’m pleased to tell you today that I’ve signed legislation that will outlaw Russia forever. We begin bombing in five minutes.

This was met with mild chuckles from the audio technicians, and it wasn’t broadcast intentionally. But it was leaked, and reached the Russians shortly thereafter.

They were not amused.

The Soviet army was placed on alert once they heard what Reagan joked during the sound check. They dropped their alert later, presumably when the bombing didn’t begin. Over the next week, this gaffe drew a lot of attention. Here is NBC Tom Brokaw addressing “the joke heard round the world”

The Pittsburgh Post-Gazette ran an article containing some of the Soviet responses five days later, on 16 August 1984.2 Similar articles ran in most major US newspapers that week, including the New York Times (which apparently retyped or OCR’d these statements, and these are now available on their site).

The major Russian papers Pravda and Izvestia, as well as the Soviet News Agency TASS, all decried the President’s remarks. Of particular note are two paragraphs from TASS. The first is reminiscent of many responses on Twitter today,

Tass is authorized to state that the Soviet Union deplores the U.S. President’s invective, unprecedentedly hostile toward the U.S.S.R. and dangerous to the cause of peace.

The second is a bit chilling, especially with modern context,

This conduct is incompatible with the high responsibility borne by leaders of states, particularly nuclear powers, for the destinies of their own peoples and for the destinies of mankind.

In 1984, an accidental microphone gaffe on behalf of the President led to public outcry both foreign and domestic; Soviet news outlets jumped on the opportunity to include additional propaganda3. It is easy to confuse some of Donald Trump’s deliberate actions today with others’ mistakes. I hope that he knows what he is doing.

Posted in Politics | 1 Comment