Isaac Laquedem

Isaac channels his inner Nate Silver on Clinton v. Sanders

I've been loath to estimate the relative chances of Secretary Clinton and Senator Sanders to win the nomination, though the odds favor Clinton, who leads in pledged delegates.  The question that interested me is to determine how voters' opinions must change if Sanders is to win.  I made a few simplifying assumptions: (a) Among likely Democratic voters, Clinton leads Sanders 52% to 48%, (b) Clinton has an additional edge in New York (4%), Maryland (4%), Indiana (5%), West Virginia (5%), Kentucky (5%), and the District of Columbia (18%), (c) Sanders has an additional edge in Oregon (10%), Montana 5%), and New Mexico (5%), and (d) pledged delegates will be awarded exactly proportionately to the vote.  I also assumed that Senator Sanders will not attract enough superdelegates to win unless he defeats Secretary Clinton in the race for pledged delegates by a margin of at least 200, i.e., the superdelegates will stick with Secretary Clinton unless Sanders defeats her by a wide margin, but that if the margin is wide enough, they will change.

Taking today's pledged-delegate totals of 1,307 for Clinton and 1,097 for Sanders, with 1,647 yet to be chosen, means that Senator Sanders needs to win 1,029 of the remaining pledged delegates.

I then added an assumption that for each remaining week of the campaign, there will be some flow of voter support from Clinton to Sanders, and then meshed this with the primary schedule, meaning that, other things being equal, Sanders will do better in the two primaries that are four weeks away than he will in one of tomorrow's three primaries, because the voters in the two later states have more time to be persuaded to change their minds.  A little fiddling with a spreadsheet reveals that, on these assumptions, each week Sanders must attract 7.5% of Clinton's supporters to change their minds, net of any flow in the opposite direction, in order to win 1,029 pledged delegates. )"Attracting 7.5% of her supporters" translates to a shift of 3.75% of the voters in a state in which Clinton and Sanders are tied.)

What might we learn from tomorrow's primaries?  In my model, Clinton will receive 56% of the New York vote and 52% in Connecticut and Delaware.  If she wins only 52% of the New York vote, Sanders's task doesn't become much easier.  Sanders still has to attract 7.2% of her supporters each week to get to 1,029 more pledged delegates.  If Secretary Clinton has the 4% edge in New York that I'm crediting her with, but she and Sanders are tied nationally, Sanders must attract 6.8% of her supporters each week to get to 1,029 delegates, including winning 64% of the California vote.  That's a lot of votes to change in seven weeks.

The first test of my model will be next week, in the primaries for Maryland, Pennsylvania, and Rhode Island, and in comparing those results with tomorrow's results in Connecticut and Delaware.  New York is its own beast.

Venerable Mom asks for help on a math test

Though advanced age has made her eyesight mostly theoretical, Venerable Mom still maintains a lively interest in the affairs of the day, and we talk every week or so about something that's caught her roving mind.  Her question of a few days ago was off her regular pattern: she called to ask me how to count backward from 100 by sevens.  "100, 93, 86, 79, 72," I said, continuing until she insisted that I stop.  "No, no," she said, "I don't need to know the numbers, I want to know how to do it quickly."  I said, "Imagine that you're looking at a number line that shows all the numbers from 1 to 100 (my fourth-grade classroom had one).  Then start at 100 and move 7 places to the left and call out the numbers as you go."

"That won't work," VM told me, "you know I can't see a thing."  "No, Mother," I replied, "you don't have to see it; you just have to imagine it."  She rejected the idea, and also quickly rejected arranging 100 dots in a 10-by-10 array and visualizing 7 dots at a time being removed.

"By the way, Mother," I asked, "why do you need to count backward by sevens?"  She told me that it's part of the mental test that her physician gives her to see if her mind is functioning properly, and she wants to be able to score well.  "I can't believe this, Mom," I told her.  "You're asking me to help you cheat on a senility test?"

After a fit of laughter, she replied, "Well, that shows I haven't lost my mind yet."

 

Should the Planning Bureau ask compound questions in a land use survey and oppose building a freeway through Alameda?

As state law requires it to do every five years, the City of Portland is updating its comprehensive plan, the document that sets out in broad terms what land uses the City will allow, and where, and on what  conditions.  (The zoning ordinance, which by law must be consistent with the comprehensive plan, provides more specific day-to-day rules.)

As part of this process, the Planning Bureau solicits public comment, both through traditional means such as open houses and mailed questionnaires and through modern methods, including e-newsletters, online surveys, and the Bureau's Facebook page.  One of the Bureau's surveys ran through May 1, and the Bureau told the good citizens that their answers would "help refine and amend many of the goals and policies in the Working Draft Part 1 of the Comprehensive Plan" and "help staff answer some critical questions around infrastructure investments, watershed health, industrial land development and housing choices."

The survey was composed by people knowledgeable about planning, but not so much about surveys.  One responder noted that almost all of the questions were compound questions that combined two questions into one, so that you could answer "yes" to both parts, or "no" to both parts, but not "yes" to one and "no" to the other.  (An example of a compound question, not from the survey: "Do you support protecting social security for the elderly and instituting the death penalty for jaywalkers?"  The title of this post is another example.)  For example, one of the survey questions asked if you want to "promote 'ecologically friendly' industrial sites through monetary incentives and technical assistance."   I couldn't answer that question; I favor promoting environmentally friendly industrial sites through technical assistance (providing knowledge) and not through monetary incentives (paying people not to mess up the City).  Yet I would have to support both or neither. 

Another question asked if the City should "subsidize the cleanup and redevelopment of brownfields."  These are two different things: the first part asks if the City should pay for part of the cost to clean up old industrial sites; the second part asks if the City should pay people to build new projects on the cleaned-up sites.  I might support the City paying part of the cost to clean up someone's environmental mess, in collaboration with DEQ and EPA and then pursuing the responsible party to repay the cost, but be less excited about the City subsidizing a project on private land, whether clean or dirty, under the benign if mistaken belief that the City is a real estate developer.

The Bureau says that it will release the results this month.  The results will be interesting, not so much for the answers as for the questions.  Also interesting will be learning whether the Bureau hired an outside consultant to write the rather sloppy questions.  For the record, I favor protecting Venerable Mom's social security and I oppose the death penalty for jaywalkers.

General semantics solves the "Which Romney" problem

General semantics, an intellectual discipline mostly devised by Alfred Korzybski (1879-1950), encourages us to look into the meaning of the words we use and recognize when we use one word to mean two different things, including when the word has two meanings because it's referring to the same object at two different times.  For example, "Congress" means the legislative body of the United States, but "Congress" means one thing if you're using it to refer to Congress in 2012 and another if you're using it to refer to Congress in 1948 (the 80th Congress of electioneering fame).

One problem my centrist friends face in this election is to form their opinions about Mitt Romney.  Mr. Romney, as governor of Massachusetts from 2003 to 2007, promoted government-mandated health care, increased fees and the gasoline tax, closed tax loopholes, opposed imposing restrictions on abortions, and banned assault weapons.  Mr. Romney's positions on these issues are different today.  Thus the problem: are Mr. Romney's supporters voting for his positions of this campaign, or of the positions he advocated ten years ago?  Mr. Obama's supporters do not face this problem, as his positions are generally the same as what he advocated in 2008; their question is whether he's sufficiently achieved his goals in his first term as President.

General semantics offers a solution through what it calls "extensional devices," including one that recognizes the problem underlying my "Congress" example.  Words can be tagged with the time to which they refer, so that it's clear that "Congress(1950)" and "Congress(2012)" mean different things.

I suggest that we print Mitt Romney's name on the ballot twice, once as "Mitt Romney (2003)" and once as "Mitt Romney (2012)."  All the votes would count toward his total.  He, and we, could then see whether Mr. Romney (2012) is in fact an improved and more popular version of Mr. Romney (2003), or whether the oddities of the Republican Party's nominating process have imposed upon him an unplanned obsolescence.

Isaac updates the Portland math curriculum

The Laquedemitasse's math homework includes some simultaneous-equation story problems, along the lines of "Three years ago Fred was three times as old as Paul was, and two years from now Fred will be twice as old as Paul.  How old are Fred and Paul now?"

It's not a very realistic problem, for in the real world, we would just ask Fred and Paul to tell us their ages.  It occurred to me, however, with some inspiration from Professor Bogdanski, that our metropolitan area offers us raw material for these problems.  Consider this one. 

A transit district operates X miles of bus service and Y miles of light rail.  For every additional mile of light rail that it constructs and operates, it must cut Z miles of bus service to pay for the additional light rail.  How many more miles of light rail can it build and operate before it will have no bus service at all?

The road to the White House leads through your dentist's office

Yesterday I fell into conversation with a woman of statistical bent, who told me that a few years ago she had done a research project on the results of American presidential elections, with the goal of determining what characteristics of the candidates were most useful to predict the winner.  The standard political model is that the candidate who is closer to the political center is more likely to win.  The taller candidate also has an advantage, and it's well-known that bald candidates are at a disadvantage.  My companion told me two results of her research that I hadn't heard anywhere, and that suggest that the American voter isn't as interested in the candidates' political views as they think they are.  

Her research and analysis showed, she said, that voters prefer candidates with larger heads, and with larger teeth, especially large regular symmetrical teeth.

Height, hair, head, teeth: it's not so much the candidates' minds, as it is what's on top of, in front of, and beneath their minds that makes us vote for them.

The unbreakable code is a prime topic of conversation

Some years ago I invented an unbreakable cipher.  I had the good fortune to encounter a cryptographer a few days ago, and told him the method.  He excused himself for five minutes, and when he came back he told me how to break it. 

They've made some progress in cryptography in the last 30 years.  The most significant development was the RSA method, which depends on identifying two long primes (originally, about 60 digits each), multiplying them to get a 120-digit number, and then using that number to encode the message in such a way that you can't decode it unless you can factor it into the two 60-digit primes.  It was and is possible to factor a 120-digit number into its primes, but at the time you needed a lot of patience; if I recall rightly the computational power of the early 1980s was such that it would take approximately the lifetime of the earth to do it.  One side effect of the RSA system was that you could announce a portion of your code to the world, and others could use it to encode messages to you, which only you could decode because only you knew your two primes that produced your public key.  It was something as if you could announce that your public key was 91, but only you knew (or could calculate) that its two prime factors are 13 and 7.

Computers got faster and more powerful, and then someone invented a method to factor 120-digit numbers more efficiently, meaning that 60-digit primes weren't strong enough.  My cryptographer friend said that Microsoft now uses 300-digit numbers for its encrypting, requiring two primes of 150 or so digits each.  No doubt eventually encryption will use primes in the thousands of digits.