Financial Sense   Home  l  Market Monitor  l  Market WrapUp  l  Storm Watch  l  About Us  l  Contact Us

PRIMER ON PRESIDENTIAL POLLS

Tap into the growing interest, if not excitement, of the upcoming presidential elections. Party conventions lie just over the horizon, sure to produce a frenzy among our sharply divided population. Events slow on the financial front at this time of year. So let’s take a break, step back from economic and financial concerns, to cover some polling technicals, some specific presidential tally consequences, and other nuances pertaining the ubiquitous polls. The most prominent nowadays are managed by CNN/ USAToday/ Gallup. Some statistical detail will be covered, hopefully to satisfy longstanding curiosity of readers. To be sure, certain topics will be mentioned superficially if only to acknowledge their importance. All discussion applies directly to other types of polls besides voters, such as parents of high school students, or Social Security recipients, or members of sports health clubs, or people who suffer from arthritis, or the daily topics posed by CNBC. No personal political views will be expressed. My role is the marketing researcher, the statistical analyst. Sections at the end of this essay can be skipped, if the reader wants only technical information on polls. They are added for completeness.

Popular questions relevant to the upcoming national election put to polls have been:

            In a two-man race, do you favor President Bush or Senator Kerry?
            Which candidate do you favor to best handle the economy?
            Which candidate do you favor to best handle the war on terrorism?
            Are you satisfied with the progress of the Iraqi War?
            Do you approve generally in the president’s performance?

Presidential polls persist in reporting aggregate national percentages for Bush and Kerry, while the country still elects a president on a state-by-state basis. In my view, this is a serious dropped ball by reporting staffs. News media focuses on the swing states like Pennsylvania and Ohio and Florida, without breaking down poll results to reveal estimated Electoral College tallies. The most important poll so far is missing, on the issue of how likely Democrats and Republicans are to make the effort to vote and show up at the voting booths. That factor helps to explain every past disconnect from pre-election polls.

SAMPLING PLAN, PROTECTION FROM BIAS

A random sample affords any member of the population an equal chance of being selected for query and inclusion. The target is the entire United States, its eligible voters. A plan must be laid out for capturing the desired number of respondents, which satisfy some criterion for inference accuracy. An inference is what we can safely conclude about the entire population, in the form of specific conclusions and decisions within bounds of accepted error. Bias is a major bugaboo, capable of eroding the entire study to the unsuspecting or ill-prepared analyst experimenter. Bias is revealed as unfair tendencies within the sampling and selection procedure. If one refuses to travel to the deep suburbs, then wealthier voters will be missed. If secure guarded apartment buildings are inaccessible near colleges and universities, then liberal minded voters will be missed. If retirement communities are overlooked, then conservative voters will be missed. If inner city areas are actively avoided for fearful reasons, then minority voters will be missed.

No sampling plan is perfect and free from all bias. The objective is to minimize bias. Oftentimes, a bias might be revealed after the fact, and thus require some sort of adjustment. Some studies are fraught with hidden agendas. Others are simply amateurish, like the famous Atlantic Monthly poll, which predicted Dewey would overcome Franklin Roosevelt. They sampled with a heavy bias toward the wealthy & conservative, and were embarrassed. Blatant examples of bias include a home telephone-based sampling strategy to determine customer desire to use a cellphone exclusively, and forego the wireline phone. Relying on daytime calls to households will oversample parents at home with small children, and miss working adults.

Most studies begin with a grand list of the population, which contain registered voters, or else a list from known registered voters. From the start, a potential bias surfaces. Newly registered voters might be unavailable to the researcher. If they tend to be younger and more liberal, then a bias is evident. Exiting voters from the elderly ranks would add to that bias, since they tend to be more conservative. The further one delves, the more biases become clear. If several research assistants are used, they might possess slight differences in approach, which could expose biases. A person might seem big and intimidating when canvassing voters. A person might be short, meek, and not explain the nature of the question to voters. Another person might skip certain houses with many steps or scary looking kids out front. Entire books are written about bias. My favorite came from primary marketing research of computer system professionals in a previous job. The busiest customers turned out to be most unavailable by telephone, but at the same time were the most demanding software or hardware users, with the highest repurchase rate. We attempted to devote a certain proportion of the sample to those who were difficult to reach. We called them back four or five times to complete the survey of questions, so that their views would be recorded.

SEGMENTATION & STRATIFIED SAMPLES

An edge of accuracy can be achieved whenever the population can be broken down into a multi-dimensional grid. Voters can be segmented according to various dimension factors, with certain statistical benefits. Commonly used factors are gender (women/ men), race (white/ non-white), age (under 50 yrs/ 50 or over), city (urban/ suburban/ rural), labor (professional/ govt/ blue collar), marital status (single/ divorced/ married), and idealism (liberal/ moderate/ conservative). When the researcher has the added benefit of knowing the population breakdown on these dimensions, with cell counts, the accuracy of the study can be markedly improved. A “divide and conquer” mentality prevails in mathematical statistics theory. When reliable voter response is known in these grid cells, the accuracy of the study can be made optimal in what is called a stratified random sample. Most professional polls are of this type, or strive to be. Voting records from previous elections can provide reasonably strong indications useful for stratified sample plans. Generally each sampled cell count should be designed to include a proportion equal to the overall target population cell count. If women make up 52% of the population, then a similar percentage of women must be represented in the sample. The number of sampled non-white married men over 50 yrs old from the urban centers who are blue collar workers and moderate in their views should be dictated by the percentage of people in the population of the same description. In some studies, a technical problem arises when certain grid cells cannot be covered in the sample, despite attempts.

If a good indication of response data is known from a previous year, or from a pilot study with an initial piece of the sample, an optimal stratification can be achieved. It requires that each sampled cell count be proportional to the target population cell count times the cell group variability (standard deviation). A statistical theorem states that, for studies which employ such a segmented grid, the optimal strategy is uniformly most accurate, and yields the lowest sample variation in the final estimates for a given fixed sample size. Take two particular cell groups, like urban liberal women and suburban moderate men. If those women are 20% more numerous than those men in the general population, but those women have a standard deviation 50% higher for the measured item than those men, then that women subgroup should be allocated a sample which is 80% greater than that men subgroup.

arithmetic shows  1.2  x  1.5  =  1.8

If white married men numbered 30% greater than non-white divorced women, but those men have a standard deviation 15% less than those women, then that men subgroup should be allocated a sample which is 10.5% greater than that women subgroup.

arithmetic shows  1.3  x  0.85  =  1.105

Additional optimal stipulations react to fluctuating costs to conduct the survey. The cost might be much greater to canvass rural areas than densely populated urban centers. Segments, which cost more are to be sampled less, according to an inverse square root rule (for those with math curiosity). The fixed elements are total sample size and total cost. This means a group, which costs four times more, should be optimally sampled half as much.

THE ELECTORAL COLLEGE MEANS 51 RACES

As a nation, the United States is unique. We elect our president by winning states, wherein the winner of each state takes ALL the Electoral College votes. A state carries EC votes equal to the number of House representatives plus its two senators, thus a minimum of 3. Some antiquated procedural rules exist and remain on the books, amazingly. Like some states permit the Electoral College designate to vote freely, and depart from the actual popular vote. Wow. Who selects the designate? The process might have been set up to ensure that distant states have their integrity protected in colonial times. It sometimes took months to get word back and forth to states. The District of Columbia has no senators, but does have three House Reps to make 3 EC votes. It can be regarded as it is with marketing research studies, the 51-st state.

In a “winner take all” process, the biggest states gather the most attention. They are called the prize states. They are led by California, Texas, New York, Florida, Illinois, Pennsylvania, Ohio, and Michigan. If the majority of the prize states are won, along with a split of the smaller states scattered across the map, then the election is usually won. Today, Pennsylvania, Ohio, and Florida are regarded as critical swing states, which will probably see enormous attention by both candidates. A few large states are so clearly tilted one way or the other, that the candidates will spend little time campaigning there. It is safe to say that tiny Vermont, Rhode Island, Delaware, Wyoming, Montana, and the Dakotas feel a bit left out. Ironically, little New Hampshire gets attention far out of proportion from its meager population. Outsized attention given to it during the primaries is enormous, the exact opposite as during head-to-head campaigns after the conventions, when it is largely ignored.

This archaic and non-federal manner of presidential selection renders national polls as either irrelevant or mere preliminary teases to the main event. State polls rule the day. The most relevant and captivating polls collect the Electoral College tallies from the individual state polls in order to reveal the leader. From what is shown in such analysis, the early results are very different in the size of the lead from what is publicized in the single national aggregate poll. As with the actual voting, state polls can focus almost entirely on the largest states and sometimes determine a forecasted winner, but only for more lopsided races. The new president must gather 270 EC votes to win office. Carry the argument to an extreme to see the importance of big states. If Bush wins only Texas among the top 8 states, then Kerry must win only 78 EC votes of the remaining 312 up for grabs in the remaining 43 states (including WashDC).

CONFIDENCE INTERVAL, SAMPLE SIZE

Now we need some math to explain the basic formulas. Keep it simple in this portion. Consider no segments are cut, and a simple sample is used. For illustration, suppose our target population is voters in the United States, and we measure preference for XXX versus ZZZ. In what follows, the people who want neither candidate, who like some other candidate, who prefer some obscure candidate or protest vote, they will be labeled “Other” for the purpose of this particular example poll. Prominent third party candidates are measured as often as they are ignored in major polls.

Suppose we have the following general parameters and examples, which we will demonstrate on results.

            the population:

            N = entire population size                                  150 million people

            p = true proportion preferring XXX                         50% want XXX

            q = true proportion preferring ZZZ                         44% want ZZZ

            r = true proportion undecided or preferring Other     6% want Other 

(true values of  p,  q,  r are each unknown)

            the sample:

            n = sample size                                                 1000 people

            P = sampled proportion preferring XXX                  49% like XXX

            Q = sampled proportion preferring ZZZ                  46% like ZZZ

            R = sampled proportion undecided or Other           5% undecided

Notice that  “P + Q + R = 100%” and “p + q + r = 100%” also. So the full population across the US is comprised of 150 million registered voters. Unknown to the researcher is that 50% of the people prefer XXX. Our sample contains 1000 people, a nice round number. We found 49% of the 1000 sampled voters expressed preference for XXX. We do not have the advantage of knowing that  “p = 50%”, which is the whole purpose of the study, to determine the unknown “p” which answers “what pct prefer XXX?” However, we do know the value of “N” from the start. Let’s leave alone how the real “N” might actually shift, like with deceased registered voters going out, and newly registered voters coming in. Generally, if N is huge relative to the “n” sample size, like over 100 times larger, it is not essential to know the exact value of  N. Its value will not influence the analysis.

Our “P” estimates the unknown “p” and came in equal to 49%, “Q” estimates the unknown “q” and came in equal to 46%. We do not know exactly how far our estimated “P” strays from the actual “p” but we can use statistics theory to make certain statements. We enter the topic of confidence intervals. Typically, polls cite the “P-Q” estimated lead, but make statements which shed light on the accuracy of each individual estimated percentage. When the full population size is more than 100 times larger than the sample, the statistical formula for a sample of “n” people, is 

Prob { difference between estimated P and true p  <  1.96 * sqrt [p*(1-p) / n] }  =  95%

The 1.96 number comes from the normal “bell-shaped curve” distribution, to account for plus or minus two standard errors. We do not have the luxury of knowing the value of “p” though. Some very complex techniques can address this obstacle. Fortunately though, an extremely convenient mathematical fact helps us. The component of the variation swing, that function of the “p” value, remains in a nice tight range throughout the entire relevant percentage for elections. When the true “p” value swings from 40% to 60%, the value of  “p times (1-p)” remains between 24% and 25% conveniently. The “sqrt” denotes the square root operation, where for instance  sqrt(9)=3  or  sqrt(4)=2 . If we lift the constant by rounding to 2 and use 25% to approximate that function, we get a very useful statement.

Prob { difference between P and p  <  1 / square root of n }  =  95%

Back to our example. We saw a XXX preference  P = 49%  with a sample size of  n=1000 people. We can make the following confidence interval statement. Our sampled 49% for XXX preference differs from the underlying actual true proportion within the entire population who prefer XXX by less than 3.1%, and we have 95% confidence in this statement. The formal confidence statement is below.

Prob { our  P=49%  differs from the real p  <  3.1% }  =  95%

our sample of 1000 reveals the true XXX preference is 49% plus or minus 3.1%,
a statement we make with 95% confidence.

Well, the preference for ZZZ is 46%, which is within the stated variation 3.1% swing for XXX preference. This is where it can get even more complicated. The statistical variation for the preferred difference between XXX and ZZZ (“P-Q)” is actually greater than the variation for XXX alone. The difference contains a second variable, thus more variation. The plus or minus swing in the confidence interval is  sqrt(2)  times greater, i.e. 1.414 times greater. So if we have a 3.1% swing on XXX estimated proportion alone, then we have a 4.38% plus or minus swing on the preference for the difference. Pollsters never talk about this extra inaccuracy when focused on differences. If you focus on a single candidate and estimation accuracy, then fine, go with their stated plus/minus figure. But if you wish to make comparisons of who is ahead of whom, as they always do in a sloppy fashion, you must multiply the plus/minus figure by the square root of 2.

Students of the trade pay attention to the square root relationship. They can notice diminishing returns with greater sample size. A sample four times larger gains us only twice as much accuracy. Such is the nature of the sampling beast. Most samples offer plus/minus accuracy of 3% in national polls in the press & media. This comes from a sample of  n=1067. Most likely, they sampled one thousand and rounded the accuracy to 3%. Notice the payoff in greater accuracy falls off as we go past 2000 in a sample. It is not worth it. One needs 4000 in a sample to cut that 3% accuracy in half, down to 1.5%. The chart below can aid the reader to work backwards. If the accuracy of a poll is 2%, what is the sample size involved? The answer is  n=2401. To achieve accuracy of only 1%, one needs  n=9604, off the chart.

LIKELIHOOD TO VOTE, VOTER TURNOUT

Over the past many years, my curiosity has been piqued by the physical voting process and the occasional disconnect between polls and final results. The link between voter preference and voting results lies in voter turnout. Pollsters refer to certain groups of supporters as “soft” or “dedicated” which helps to explain the disconnect. Loyalty is difficult to measure or quantify. So are motivation and initiative to go the extra mile to arrive at voting booths. Some intangibles are not subject to measurement. In the aftermath of surprisingly wide victories, like Clinton over elder Bush, or Carter over Ford, or Reagan over Carter, voter support for the loser was found to be soft. The presidential polls had failed to predict the decisive victory. The missing link was in voter turnout and initiative to make the effort. Pollsters have attempted to address the softness or dedicated characteristics of voter segments associated with certain candidates. The best researchers have actually made great progress in integrating the voter preference in conjunction with voter turnout likelihood. The statistical methods are greatly complicated, but the payoff is worth the effort in accuracy. Lazy or untrained researchers completely overlook this critically important factor.

UNDECIDEDS & NON-RESPONDENTS

Certainly, voter preferences are fluid, and change over time for good and bad reasons. Certain candidates are entertained in our minds, investigated on the issues, and are found to be alluring or unattractive. Wrong reasons wear off over time. Campaign advertisements are intended to present beneficial information, which sways preference. Mass media exposure can also be used to distort reality at times. We know all too well the smear tactics and negative agendas. However, many voters simply cannot make up their minds. Neither candidate looks all that good, or particular issues are totally unaddressed, or an obscure person is preferred but not seriously. The “undecided” group is an integral part of the fluid center. Candidates curry favor to this uncommitted class of voters. Sometimes an issue is resolved and loyalty is won. Sometimes a given candidate “grows on a voter.” Some voters actually choose not to make up their minds until late in the game when candidates reveal themselves more fully.

Apart from how or why voters change their minds, a curious phenomenon has emerged in this election season. It might have been present in the last election. In many previous elections, at this time of July, four months from the November official voting, the “undecided” group typically had gathered a percentage of polled votes between 8% and 20%. The size of the uncommitted typically has been large, usually much larger than the margin of leadership by the front runner. Not this time. We have a polarized population this time around, and might have had a similar polarization with few undecided voters in the last election. The undecideds in this election are remarkably small in number. In previous years, the big polling question was “how will the undecideds go this time?” Some polls of yesteryear were entirely directed at the large undecided group. In the current race, we have the advantage of delving deeper into issues and demographic disparity. Deeply divided public is hardly ever a positive development in a democracy. However, a small fluid center enables greater revelation of the voting population by means of polls, which has its own value.

Party conventions often help the undecideds to make up their minds. With all the speeches, publicity, public appearances, and media reporting, a preference often emerges. Let’s not forget the balloons and endless music in what often appears to be a circus atmosphere. Usually, a lift is seen in the Democratic candidate in the polls leading into and during the Democratic convention. Ditto for the Republican candidate and their convention. The lift can be a couple percentage points or more. Party planks (main issues and objectives) can be persuasive. Agendas are established. Hitting the proper chords can be successful. Reaction to current events can be critical.

Another simple sounding factor involves non-respondents to the polling questioner. It takes a moment to be clear that a vote of “none of the above” is very different from “no reply.” Those who do not cooperate with the sampling effort serve as a wooden shoe in the machinery, in a sense, sabotage. Their reasons vary from outright defiance, dislike of the researcher’s style, offense to disruption of the non-respondent’s leisure down time, or not caring about politics one iota. When it comes to politics, the non-respondent is more likely to be a no-show at the voting booth. From a purely statistical standpoint, the Non-respondent Problem is a hole in the process, a swing and a miss within the sampling leveraged machinery. It is one of the most pernicious problems in survey sampling. It really cannot be ignored. Instead, it requires a separate sample to cover the hole in the outcome. Truly competent researchers, in very important polls, will re-sample the non-respondents. Apart from politics, in a health study on certain risk exposure, for instance, passive subjects might not interfere with the outcome as they contract a disease or die.

EXIT POLLS

An impressive pollster practice in the last 20 years has been exit polls. They get voters coming and going. While summertime and early autumn polls can address the likelihood of election results, exit polls offer an immediate access to voters who are fresh on the scene. Some do not reveal their votes, which puts the Non-respondent Problem back on the analytic burner. Exit polls can conveniently delve deeply into why voters voted the way they did, what issues turned their vote, which demographic groups they come from. Certain traits are obvious like gender, racial group, and city factors. Other traits must come from query. Exit polls can directly measure the previously debated undecided group, provided the exiting voters reveal once being undecided. They can also reveal the softness of candidate support, even flip-flopping late in the game. Personally, my favorite part of watching election results on television is the reporting from exit polls. These are typically much less sophisticated than early season polls, since they tend to be more impromptu and more subject to voter lack of cooperation. The more outgoing and sociable voters cooperate. In the 1972 election, I told an exit pollster that I voted for “Captain James T Kirk” then smiled. When asked why, I said “leadership,” but revealed my real vote over my shoulder <G>. Others do not cooperate who feel an invasion of a very private constitutional right. Exit polls directly intrude into the sacrosanct voting process.

TIME ZONE EFFECTS

The sheer size of this country, combined with advanced projection methods, have resulted in actual legal changes to election news coverage. First, time zones can cause problems. In the last 2000 election, voters on the Florida panhandle, who live in the central time zone, were improperly told by television broadcasters that the voting centers had closed. Many voters were outraged, but no sinister motive was evident, given the makeup of panhandle demographics. Election reform came after Clinton’s second term election. He was projected to have won a landslide victory long before the California and west coast states were near closing time. This infuriated western voters, who were made to feel irrelevant. The California outcome, needless to say, was skewed from previous polls. The new guidelines require the major media chains to forego projecting an overall national winner before western voting centers close. However, this is a murky requirement. They are permitted to project winners in individual states. If a collection of projected states tilts the overall result, then the media network must remain reluctant to add up the overwhelming tallies. So they tend to hold back in projecting the final states which would tilt the election result. I suppose our founding fathers never considered states would stretch over four time zones in the lower 48, plus one more zone for Alaska, and three zones westward for Hawaii. It could be worse; Russia has 12 time zones. Polls cannot anticipate such time zone events.

3^rd-PARTY CANDIDATES, PLURALITY, RUN-OFFS

The first presidential third party candidates in the modern era were Eugene McCarthy and John Anderson. McCarthy challenged Richard Nixon and Hubert Humphrey. Can anyone remember the battle cry of the hippies in 1968: “get clean for Gene?” Anderson challenged Richard Nixon and George McGovern in the second term. They each had their reasons, usually the major parties disconnected from a significant slice of the population and constituency. More recently Ross Perot and Ralph Nader spoiled the brew. Perot played a role in pulling critical votes from the elder George Bush, aiding Clinton’s victory. Nader played a role in pulling critical votes from Al Gore, aiding the younger Bush’s victory. The present debate centers on whether Nader will once more affect the outcome of the upcoming election between the younger Bush and Kerry.

The United States should seriously consider the European procedural style of runoff elections. In a superior setup, the top two candidates enter a subsequent run-off election two weeks later, which excludes the other also-ran finishers. If such a procedure had been employed in 2000, we could have had a different president, but we also could have had a repeated fiasco in Florida. Since the USA has had the remarkable balance of two major parties, no constitutional amendment has ever been pursued. European parliamentary govts have traditionally included several key parties, and ruling coalitions. Prime ministers can lose power when votes of no confidence are entered by the parliament, which force new national elections. My opinion is that Europe’s method is superior to ours in the USA. The day has come when splintered factions, disaffected voter blocks, diverse interest groups, and those who object to entrenched political machinery do not feel properly represented, even disenfranchised. Small parties must be included. After the initial vote, their voters must be directed to select between the leading two candidates. My favorite obscure political party, true story, is the tiny faction in Italy which represents gay truck drivers.

When the California governor post was up for election last autumn, much talk focused on the sheer multitude of candidates who satisfied the criteria for ballot inclusion. It was actually possible that the new governor would receive under 20% or 30% of the vote, and not even remotely achieve plurality. Technically, neither Clinton in 1992 nor younger Bush in 2000 received a plurality, tallying just under 50%. Run-off elections solve the problem. The risk of democratic chaos presents itself if the majority harbors deep opposition to a winning candidate who fails to achieve plurality.

Polls often report both 3-sided and 2-sided results, but only the complete 3-sided races matter. Head-to-head polls are easier for the public to digest, but not a proper reflection of reality.

OBSTACLES & WEATHER FACTORS

Certain unpredictable factors often come into play. On an individual basis, a parent might not be able to get a sitter for the kids. An employee might have to work late. A voter might get wrapped up in the day’s events and forget. The line at the voting booth might extend around the block. Physical problems might arise from voting booth machines. Voters might forget to bring proper ID to permit voting. Voters might not know where they are supposed to vote. We will not go into the punch cards and hanging chads, oy veh! Things happen. On a wider scale, power outages might interfere with voter turnout. An access road might experience a traffic accident. As long as people are involved, snafus happen. Polls cannot adapt.

Natural factors very often play a role. A rule of thumb is that bad weather favors the Democrats. The reasons why are not so palatable, and sound stereotyped. Experts claim elders and conservatives tend not to brave the bad weather, but youngers and liberals will allow themselves to go through discomfort. Enough. Weather plays a role, check for yourself. Bad weather might enable those who live closer to voting centers to arrive to vote more easily. As long as people must negotiate nature, strange things happen. Polls cannot adapt, and ignore all such factors. They must avoid weather, or else they are entangled with weather forecasts.

ACTIVISM, CORRUPTION

Other factors can greatly influence an election outcome, both legally and illegally. Activism can result in precinct managers and campaign directors going house to house in “getting the vote out” and actually assisting. Assistance and active promotion is fully legal. They can remind voters, especially the less mentally sharp. They can drive caravans loaded with voters who do not own or have access to cars. This is an urban and suburban and rural practice. They can simply engage in campaigning on the issues or urge party loyalty. A name is given to the entire apparatus for reaching out to voters and facilitating their votes. It is called the “party machine” and typically is an urban phenomenon. In many cases, the assistance is welcome and appreciated. In other cases, like with city govt workers or labor unions, pressure exists to vote for a certain party or candidate. This is not peer pressure, but rather employer pressure, and lies directly in the gray area of legality. Polls have difficulty in separating voter stated preference (conscience) from their likely vote (reality). Polls might not adequately address the distinction.

Examples of corruption are as long as your arm. My favorite is the registration and alleged Chicago voting of perhaps over 300 thousand dead people for John Kennedy in 1960. Without a victory in Illinois, Nixon would have served his first term eight years earlier than he did. In some backwater towns far from the madding crowd, entire batches of written ballots are burned or discarded which might tally for perceived “city slicker” candidates. In some cases, the party in power of major cities omit entire precincts which are traditionally in opposition to their favored candidate. Examples of tampering with voter registration lists are numerous. Minority voters are typically the victims, as are poorer voters, since they have more difficulty in challenging the disenfranchisement through due legal process. Claims are alleged that perhaps 100 thousand African-American voters were improperly removed from the Florida registration lists in 2000. Some used the cover of legal means to do their nasty deeds, as Florida prison inmates are not permitted to vote. Free citizens share some of their names, the confusion from which enabled the removal from registration improperly. Whether or not corruption in voter fraud was the case in Illinois or Florida is not certain, but accusations and prima facie cases are easy to build. Polls cannot come anywhere near this problem. Reports of past violations involve electronic voting and precinct collection. Access to machine program software is not easily shared for inspection. Some election process laws pertain to inspection of machine hardware but not software, which seems incomprehensible. Potential for voter fraud is ripe.

OTHER TOPICS

Incumbents have a clear advantage, as they mix the business of running the administration govt with actual campaign efforts. The presidents in the past have often resorted to tax cuts and other favorable ploys to curry to voters. Recognition and respect for the institution of president carry a positive rub. Attention can turn negative if things go badly, whether inside the economy or outside with foreign activities.

In times of domestic economic trouble, we as a nation send the president packing with little concern over whether fairness is applied. A president might inherit a difficult climate (the subject of a later essay on blame & credit) from predecessors. Voters and even experts are often clear on what factors are responsible. Few sitting presidents are re-elected when the economy and jobs are in distress or not thriving. Others point to the four-year term as providing plenty of time to turn things around. However, in times of foreign conflict, we as a nation rally around the president, our commander in chief, and support our troops. This reminds me of a family, which argues incessantly among themselves, hunkering down when attacked from outside.

Political polls are a huge investigative topic, the subject of statistical volumes. When hired in my last job, the opening filled by me came from an analyst who departed to start his own political poll consulting business. Several other topics are important, interesting, and relevant. When multiple polls are made public, readers want to know which is the right result. There is no perfect result, since they are all sampled snapshots. Some, more than others, might be undermined by bias, which is hardly stated with the results. They might be consistent with each other, each lying within the bounds of sample variation. One poll might be entirely inconsistent, which would invite query as to how they sampled in their strategy.

Another interesting topic is longitudinal trend. To study how voter preference changes over time is a fascinating and challenging task. Historical trends might come to the fore, such as if incumbents gather support over the summer months, or new challengers become more gradually known as the campaign rolls onward. It leads to early attempts to extrapolate to election day, when gains could continue as a pattern is revealed.

NEWS TIDBITS

A blank week for economic reports. So, nothing to distort. This is a heavy week for earnings reports though, as 40% of S&P firms and 14 of 30 Dow firms report. KMart was beaten up in a Barrons article, as they were criticized for investing little and selling portions of the business. Their stock sale was suggested based upon valuation, after a significant multi-month rise. In the same publication, Mike Santoli describes the Dow30 as stuck in an 8% range over the last 120 days, in what he calls “Dullsville.” The NYSE trading volume was 1.187 billion shares last week, compared to a 20-day average of 1.775 billion shares. Is Fed Gradualism going to undermine the stock market for months to come? The Retail Holders index (RTH) is down for five consecutive weeks. WalMart reports that sales remain in the 2 to 4% growth range, but cites $16 billion in tax refunds one year ago which make for tough comparisons. The Fed Funds futures contract is pricing in a 25% chance of a 25 basis point rate hike in August. Greenspan faces Congress in the next two days for Humphrey Hawkins testimony. One side will make speeches disguised as questions, while the other will make eloquent excuses. Attorney Spitzer filed a motion to put the NYSE Grasso case into state court.

Gov Schwartzenegger angered California state legislators in a “girlie men” reference, as he accused them of catering to special interests, labor unions, and trial lawyers. Health & Human Services Secy Thompson has recommended certain obesity related medical treatment to be covered by Medicare, after movement has progressed to label obesity as an illness. A CNBC poll indicates 82% of voters said “NO” on whether Medicare should pay for obesity treatments.

Todd Hamilton, a 38-year old newcomer to the PGA Golf Tour, overcame Ernie Els in a 4-hole playoff at Royal Troon to win the British Open. Dale Earnhardt Jr was injured in a burn incident in Sonoma, Calif. He suffered moderate burns on the face and legs. “I, Robot” topped the box office this weekend with $52.3 million (including me), see my own movie review. “Spiderman 2” came in second with $24.2 million.

TODAY’S MARKET

Today the Dow Jones Industrials wrapped up at 10,094 (-46), S&P at 1101 (-0.5), Nasdaq at 1884 (+1), TENS yield 4.36% (unch bpt). Currencies closed with Euro at 124.23 (+1.11), JYen at 92.64 (+0.88), Can$ at 76.34 (+0.90). Metals finished with gold at 406.4 (unch), silver at 660.0 (+5.2), copper at 130.50 (+2.05). Energy ended with crude oil at 41.44 (+0.30), natural gas at 586.4 (+2.7), unleaded gasoline at 127.18 (-1.20). Prices are at major futures contracts.

Jim Willie CB

Home  l  Broadcast  l  Market Monitor  l  Storm Watch  l  Sitemap  l  About Us  l  Contact Us