Can We Trust The Inflation Statistics?

In response to my last post calling out the Federal Reserve for engineering grotesque income inequality in America, some commentators have said that the Federal Reserve shouldn’t be blamed since there hasn’t been any meaningful inflation.

Before I continue with this post, let me say that even when relying on the official statistics, there has been massive redistribution. The average inflation rate over the past 20 years has been 2.3% while the average growth rate has been 2.53%. I realize most people who go through an economics program are not trained to think this way, but let’s remember that absent money printing, prices would fall by approximately the growth rate. In other words, if the money supply had simply remained fixed over the last 20 years, prices would have declined by about 2.5% per year (all other things equal).

Now that doesn’t mean business would have been suffering. The whole reason for the decline in prices would have been technological improvements reducing the cost of production and increasing supply. If you sell more at a lower price it’s possible to be more profitable than selling less at a higher price.

Think of what has been happening to prices in the technology sector. Prices have been falling due to production outstripping the growth of the money supply, yet tech businesses are flourishing. The same thing would have applied to the entire economy.

So because prices should have fallen by 2.5% and instead they rose by 2.3% that implies they were inflated by an average of 4.8% each year. What this means is that upwards of 5% of income each year (though probably less on average) has been siphoned away from the middle and working classes to the wealthy. Compound this over 20 years (or longer) and you get a pretty massive number for the amount of wealth that has been redistributed.

So even using the official statistics, I can rest my case. However, my intent with this post is to question the accuracy of the official statistics.

How do we calculate inflation?

Most people seem to assume that calculating inflation is a pretty straightforward task. Just go into the stores, record the prices at the start of the period and again at the end, calculate the difference, and there’s your inflation rate. Unfortunately, it isn’t that simple.

The first problem you run into is that products come and go from the marketplace. It’s likely that at least some of the products in your study will not be on the shelves when you go to record the price at the end of the period. So what do you do? Well, you can do what the BLS does and try to select a similar product as a substitute, but already you can see that the degree of accuracy in your calculations is going to take a hit.

What if there are no similar products? Well, then you’ll have to use a dissimilar substitute. But this is problematic because different products will likely be of different quality. So now you have to try to adjust for quality. For example, the iPhone 5 was introduced in the market at price point of $649 for the 16GB model. The iPhone 5s was just released also at a price point of $649 for a 16GB model. Obviously, the iPhone 5s is of a higher quality than the iPhone 5. Should we just record this as no price change in our calculations or should we make an adjustment? There is a sense in which we should reduce the price of the iPhone 5s to account for its higher quality, but by how much? Quality is subjective after all. Are we going to let our subjective judgements influence our reported inflation rate?

The BLS tries to adjust for quality by using regression analysis. This may be the best they can do, but it isn’t perfect. Given that the phenomenon is subjective, there really isn’t any good way to go about measuring it. So our numbers are bound to be off by some degree.

The second problem you run into is weighting. Should all products that you measure be given equal weight in your inflation statistic? If so, then you might understate (or overstate) the degree to which people actually experience inflation in their everyday lives. For example, suppose the price of housing is skyrocketing, but housing in just one of 10,000 prices that you measured in your study. As such, the skyrocketing price of housing isn’t likely to affect your inflation statistic that much. However, in the real world, housing makes up a large portion of most people’s budget — meaning they would experience much higher inflation than reported by your statistic.

So there is another sense in which we want to apply weighting to products to get a more accurate measure of the inflation that is actually felt. But how much weighting? Not everyone spends their income on the same things. Some people may spend more on education while others spend more on entertainment. Each person is likely to experience inflation differently. How do we decide what the best weight is to apply to each category? Any weight you chose is likely going to affect the accuracy of your statistic.

So you can see, calculating inflation isn’t a straightforward task. If you gave 50 people the price data collected by the BLS as asked them to calculate the inflation rate, you would probably get 50 different answers. As it stands, the BLS has changed the formula for how they calculate inflation several times throughout the years, resulting in significantly different inflation numbers. Here’s an example of the difference as reported by Peter Schiff:

As reported in our Global Investor Newsletter, we selected BLS price changes for twenty everyday goods and services over two separate ten-year periods, and then compared those changes to the reported changes in the Consumer Price Index (CPI) over the same period. (The twenty items we selected are: eggs, new cars, milk, gasoline, bread, rent of primary residence, coffee, dental services, potatoes, electricity, sugar, airline tickets, butter, store bought beer, apples, public transportation, cereal, tires, beef, and prescription drugs.)

We know that people do not spend equal amounts on the above items, and we know their share of income devoted to them has changed over the decades. But as we are only interested in how these prices have changed relative to the CPI, those issues don’t really matter. We chose to look at the period between 1970 and 1980 and then again between 2002 and 2012, because these time frames both had big deficits and loose monetary policy, and they straddle the time in which the most significant changes to the CPI methodology took effect. And while the CPI rose much faster in the 1970’s, the degree to which the prices of our 20 items outpaced the CPI was much higher more recently.

Between 1970 and 1980 the officially reported CPI rose a whopping 112%, and prices of our basket of goods and services rose by 117%, just 5% faster. In contrast between 2002 and 2012 the CPI rose just 27.5%, but our basket increased by 44.3%, a rate that was 61% faster. And remember, this is using the BLS’ own price data, which we have already shown can grossly underestimate the true rate of increase. The difference can be explained by how CPI is weighted and mixed. The formula used in the 1970’s effectively captured the price movements of our twenty everyday products. But in the last ten years it has been quite a different story.

As an aside, I can’t seem to find it now, but I remember hearing this story earlier this year: Someone said to Paul Volcker, “Boy you really overreacted to inflation in the late 70s/early 80s.” Volcker said “What do you mean?” He said, “Well, if they calculated inflation back then the way they do today, there wouldn’t have been any inflation at that time”.

That’s likely very true. And if they still calculated inflation the same way, right now the press would be bombarding Bernanke with questions about how he’s going to deal with the double digit inflation.

So we should ask, do they have the formula correct now? Did they have it right before? I don’t even know if there is such a thing as a “correct” formula.

Problems with current formula

There is some reason to doubt that the current formula is correct, however. Consider that health care costs have been skyrocketing in the recent years, yet the BLS only applies a less than one percent weight to health care in it’s formula. Less than one percent!? Yet, a 2012 Kaiser study found that the health care costs were a full one third of the median family income. By giving such a low weighting to health care, the BLS very well may be underreporting inflation.

This is just one example of many, but it goes to show how small decisions made by the BLS can have major impacts on the overall number.

We also have to question how accurate their collection methods are. The BLS reported that health care costs rose by 4.3% between 2008 and 2012, yet that same Keiser survey found prices increased by 24.2% (5.5 times faster). So who’s right? Judging by my own two eyes I would say it’s the Kaiser survey. 4.3% just seems ridiculously low compared to my own experience.

Here’s another example reported by Peter Schiff:

Magazines and newspapers provide a good case in point. The truth has not been exposed through the economic reporting that these outlets provide, but in the prices that are permanently fixed to their covers. For instance, from 1999 to 2012 the Bureau of Labor Statistic’s (BLS) “Newspaper and Magazine Index” (a component of the CPI) increased by 37.1%. But a perusal of the cover prices of the 10 most popular newspapers and magazines (WSJ, Washington Post, Time, Sports Illustrated, U.S. News & World Report, Newsweek, People, NY Times, USA Today, and the LA Times) over the same time frame showed an average cover price increase of 131.5% (3.5 times faster than the BLS’ stats). This is not even in the same ballpark.

Interesting to note, economist Jeff Herbener tells a story from when he was in grad school. They had a BLS employee come speak to his class about how they produce the inflation statistics. This employee let slip that sometimes when they collect the data and crunch the numbers, the model spits out an inflation number that is just obviously wrong. Like showing a 6% inflation rate when the quarter before it was only 1.2%. Not knowing what to do about it, they just send the number up to their bosses. A few days later, they see it reported in the papers that the inflation rate was 1.5%.

I think you can chalk that up to sheer incompetence. Pretty typical for government employees.

So to conclude, calculating inflation isn’t nearly as scientific as people seem to think it is. It’s one of those things that doesn’t really lend itself well to measure and where small changes in the methodology can significantly alter the results. Even if we assume that BLS employees are honestly doing the best job they can (an assumption that is dubious if you believe bureaucrats respond to incentives like the rest of us), the reported inflation rate isn’t likely to accurately reflect reality.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s