The Bureau of Labor Statistics (BLS) is close to announcing the results of last week's presidential election. The Bureau, chosen last week as the arbitrator of the election conflict, has been working around the clock since then to apply their statistical models to the vote counts that will determine the next President. Their announcement is expected at noon tomorrow.
Although votes were cast one week ago, the winner has not yet been determined. The two sides have been mired in a legal struggle resulting from a series of lawsuits initiated by the Kerry campaign, met by the Bush administration's counter-suits. The legal standoff has brought into question both the vote counts in various closely fought states and the legality of voter registration procedures. The number and complexity of the suits threatened to tie up the result of the election for weeks or even months.
The BLS was an unlikely latecomer to the conflict. In an unprecedented move, teams of lawyers from the two campaigns met in a series of closely guarded meetings under the supervision of Supreme Court justices to agree upon a neutral third party. A surprise emerged from these meetings that stunned even seasoned Washington observers when the Bureau of Labor Statistics (BLS) was given the job.
The BLS was chosen because of their statistical expertise. Several well-known and closely followed index numbers are produced by the Bureau on a monthly basis, most prominently the consumer price index (CPI), the producer price index (PPI), and the unemployment report. They are well known among economists and financial analysts for their number crunching abilities.
Since taking on the task, the Bureau's job for the past week has been to sort through the conflicting claims using their widely used statistical models to determine who really won: Bush or Kerry.
The BLS is not without its critics – some have suggested that the choice of the Bureau may in the end prove even more controversial than had the results been fought out in the courts. Many of the Bureau's statistical procedures have come under attack since the mid-90s. Critics of the BLS have charged that the inflation and unemployment numbers are manipulated for political reasons.
A prominent Austrian economist explained, "There are really two means of funding government expenditures: taxation and inflation. Government debt can be financed for some time by borrowing, but it only postpones the decision whether the debt will be paid back by future taxation or future inflation."
"Taxation is the most direct means, but also the most easily resisted. Governments often choose inflation because the mechanism is not well-understood by the average citizen, so it is easier to blame the loss of purchasing power that people are experiencing on scapegoats – greedy oil companies, OPEC, greedy labor unions, u2018cost-push', u2018demand-pull', or by pointing to whatever price component happens to be going up at the time. In so doing, they distract people from the real cause of the problem, which is central bank money printing."
Governments, according to this economist, have a natural attraction to inflation, but also to hiding its effects. "Inflation costs the government a lot of revenues because social security and pension payouts are indexed to inflation. Because income tax brackets are also indexed, tax revenue does not go up as much as it might when there is inflation."
"And there is the effect on inflation of the government's own cost of borrowing. If credit markets believe that there is inflation, interest rates will go up to reflect the loss in purchasing power of the money that will be used to repay the debts. This will in the long run increase the government's cost of financing its debt. And finally, a high rate of inflation casts doubt upon the effectiveness of the central bank's alleged policy of fighting inflation." the economist continued.
The Austrian economist's most controversial charge was that the BLS manipulates the inflation number for the benefit of the government. "The obvious solution is to inflate but then to lie about the effects of inflation by reporting a low inflation number. This understates the the reality of the monetary debasement that is occurring."
The BLS has adopted a number of questionable procedures in figuring its final inflation number. Among the most controversial are quality-adjusted prices, altering the composition of the index, seasonal adjustments, and the proliferation of different index numbers. Also, the use of the "birth-death" model in the unemployment report has drawn criticism.
Critics of the BLS have drawn attention to the practice of adjusting prices based on improvements in quality, using so-called "hedonic adjustments". The explanation given is that a good that appears to have gone up in price because it costs more does not really cost more because the quality has improved. According to this logic, it might even cost less.
For example, computers in today's stores have faster CPUs, more memory, and more hard drive space than computers of five years ago. These specifications are used to compute a quality adjustment factor that is then applied to the price of a computer. If the factor is 0.50, then a $2000 computer will be assigned a price of $1000 for the purpose of computing the CPI.
As the use of these adjustments has spread from high-tech products like computers and cell phones to cars, washing machines, and even health care, an ever-expanding array of consumer price increases are adjusted downwards by their "hedonic multipliers" before the CPI is computed. The numbers used as input to the CPI then show the prices of many goods falling that are in reality rising. This procedure results in a lower inflation number than if the actual prices paid by consumers were used.
The BLS has been using the same techniques in computing the final vote tally in the presidential race. The Bureaus is computing quality adjustment factors for Bush and Kerry.
According to a BLS spokesperson, "Using a base year of 1980, the quality of Republicans has declined considerably," he said. "Republican candidates use to at least talk about lower taxes and smaller government, but Bush is a bigger spender than any Democrat. We have come up with a Republican quality adjustment factor of 0.45." Applying the factor to Bush's vote total of 65 million votes yields a quality-adjusted total of 29.25 million.
If the base year were set to somewhere in the 19th century, that was a time when the Democrats were in favor of sound money and opposed central banking. But a base year of 1980 brings a comparison of Kerry to the malaise of the Carter years. This comparison shows only a small quality drop-off for Democrats. The quality adjustment factor applied to Kerry is only 0.85 according to the Bureau.
The arbitrary dropping of certain components from the index has drawn attention from critics. Food and energy were dropped from the CPI some years back because these components were deemed "too volatile" and did not reflect the "core" rate of inflation.
"When the central bank prints more money, the price of something will go up. Not everything, and not by the same amount, but something," said the Austrian economist. "If you exclude those things that are going up in price from your inflation index, you can come to the phony conclusion that there is no inflation. This is equivalent to saying that there is no inflation, except for those prices that are rising," he stated.
This technique is also being applied to the presidential vote totals. The vote counts from Alaska, Hawaii, and North Dakota were dropped from the totals on the grounds that "no one important lives there", said the BLS, and Florida votes were excluded on the grounds that they are "too volatile". "Look at all the problems Florida caused for us last time", said the Bureau's spokesperson.
Seasonal adjustments are another of the controversial practices used by BLS statisticians. The idea is that heating oil might be high during December due to seasonal demand by north-eastern home owners for oil to heat their homes during the winter months, but this does not reflect inflation per se, only a seasonal condition that will be evened out over the full year with lower prices in the spring and fall.
Some critics though, have questioned the integrity of the process. According to the BLS site, seasonally adjusted energy prices declined on a quarterly basis in the 3rd quarter, in spite of record-high prices for oil and natural gas.
Another questionable aspect of the process is the way that energy and housing prices interact in the computation. When home prices are increasing by 20-30% annually in some US cities, some have questioned how the contribution of housing to the cost of living could be decreasing.
The explanation is that Bureau statisticians believe that some housing rentals include energy. Energy prices are therefore subtracted from rents to produce an energy-adjusted rental price. When energy prices go up, energy-adjusted rents go down, resulting in a lower contribution by housing prices to the total CPI. The energy price increases are then removed by the seasonal adjustment, resulting in a lower contribution of both energy and housing to the inflation index even when both are rising.
Seasonally adjusted vote totals were computed for Bush and Kerry. "The apparent vote totals of about 65 million each for Bush and Kerry are a statistical aberration", the BLS spokesperson stated. "The election only takes place every four years, so the seasonally adjusted number of votes is 16.25 million." By way of explanation he added, u2018If Hillary Clinton were to run next in 2008, we would have to adjust her vote total by a multiple of 1/12th because she has been running for president since 1996".
The proliferation of different index numbers is another area of controversy. In addition to the CPI, there is the "core CPI" and the "median CPI". The Austrian economist commented, "If one if their index numbers goes up too much during the month, policy makers can always point to another one of the indexes that was more moderate and come up with a reason why the more moderate index is the u2018real' rate of inflation. It is Orwellian because the next month, they will use a different argument to tell you why a different index measures the u2018real' CPI than the month before."
The BLS will be computing three different vote totals for Bush and Kerry: total votes cast, "core" votes, and "median votes". The "core votes" for Kerry consist of the total number of votes cast in the heavily Democratic cities and coastal areas that form the Democracts' core constituency, such as Berkeley California, Santa Monica California, north-eastern Cities, and Chicago. Core Republican votes will consist of the number of votes cast in the core Republican regions such as Texas, the Midwest, and the South.
One problem with this method is that each candidate shows a substantial lead in his own party's core vote totals, making the selection of a president more difficult rather than less. This outcome is inherent in the BLS' methodology, according to the Austrian economics, who is on record as stating, "If you can manipulate the perception of reality by putting only those things that are consistent with the result you want, you can twist reality to fit your own agenda."
Substitution of cheaper products for those that have risen in price is another controversial adjustment procedure. BLS statisticians reason that if beef goes up in price, people may eat more chicken instead, because it costs less. They then expand the weighting of the substitute good – chicken – in the index, while contracting the weighting of beef. This produces a lower number for the CPI than if the higher priced beef had been used.
"This is one of their most dishonest techniques because they are confusing inflation itself, which is the expansion of fiat money supply, with peoples' response to inflation, which is to consume less of the goods that they value more because they are losing purchasing power", the Austrian economist noted.
The application of this substitution effects to the election has proved difficult for the BLS. "Because both candidates are so awful," the BLS representative began, "the decision to vote for one or the other is a u2018lesser-of-two-evils' for most voters. We are going to determine the number of Bush voters who are only voting for him because he is not Kerry, and vice versa, then subtract these effects from the totals because these votes represent substitutions of a lesser good for a more highly valued one." If there had been a viable independent or third-party candidate this year, such as Nader in 2000, that would have confused the process of making the calculations even more.
A final adjustment was provided by borrowing a technique from the monthly unemployment rate computation. Due to alleged measurement problems in counting the number of jobs created or eliminated during any month, a computer program known as the "birth-death model" is used to estimate jobs that cannot be counted. This model was developed during the booming 1990s when it was assumed that small businesses were creating a lot of jobs under the radar of the Bureau's data collection. Government jobs, part-time jobs, and temporary positions are also counted as net new jobs in this process.
During some recent months, statistical job losses were translated into net job gains by the addition of hundreds of thousands of jobs from the birth-death model. The King Report, a frequent critic of this methodology, has written:
Of the 947,000 jobs counted by the BLS the past three months, 618,000 representing 65% of the u2018job creation' are due solely to the Birth/Death Rate. Without going into much detail, it has to do with statistically created jobs, not real jobs where people get pay checks.
The Bureau's statisticians have been hard at work to adapt the birth-death model to presidential vote counting. For the election, this model has been revised to estimate hypothetical votes that would have been cast by Democrat and Republican children not old enough to vote, projected number of children based on estimated family sizes of Republican and Democrat families, and voters who would have voted had they not died before the election.
Computing the correct birth/death numbers for the presidential election has proven to be a challenging task. According to a new study, Republicans are having larger families than Democrats. However, "Democrats have a long history as the party of welfare patronage, and the welfare system gives cash incentives to single women to have more children. We are coming up with a computer model to estimate the effects of these adjustment factors."
How the statistical process will play out is anyone's guess, leaving the election results very much in doubt. Some fear that the announcement of a winner tomorrow will spark another round of controversy as the Bureau's statistical methods become the focal point in a renewed firestorm of debate. The real problem, though, according to the Austrian economist is "when you trust government statistics, you give them the right to define what reality is."
November 1, 2004
Robert Blumen (send him mail) is an independent software consultant based in San Francisco, who does not plan to vote in the election.
