Archive for the ‘Economy’ Category

Understanding the U.S. GDP

Tuesday, August 12th, 2008

This essay will provide an overview of the Gross Domestic Product (GDP) of the United States, Subsequent essays will provide more detail on the components of U.S. GDP.

The GDP is an attempt by the Bureau of Economic Analysis (BEA) to measure the value of final goods and services produced in the US in a given period of time (see here for a pdf primer by the BEA that explains the GDP as well as other measures of economic activity).

A few key points: GDP measures mostly market production, that is activities that people pay for (leading to the old economist joke that if a man marries his housekeeper-or a woman marries her accountant-it will result in a decrease in the GDP). GDP measures production, not sales. GDP also tries to capture the services generated by owner-occupied housing. The idea here is that when one rents a house the BEA can measure the service flow (the apartment produces, say, $800 of services a month), but unlike the housekeeper/accountant example above, when the renter buys his apartment the BEA tries to continue to measure the service flow. The BEA does this by estimating the rent on comparable housing and imputing (or attributing) the income to homeowners. In other words, BEA counts the rent payments that homeowners would have paid as part of the GDP.

The BEA estimated that the US GDP (see here for the BEA’s latest release) was around $14,250 billion in the second quarter of 2008 (see Table 1.1.5 here where GDP is measured in current Dollars). This number is “seasonally adjusted annual rates” or SAAR, meaning that if the whole year were like the second quarter then the total GDP for the year would have been $14,250 billion, or that quarterly GDP was roughly 1/4 of that figure.

The GDP is calculated both in current dollars and in “chained” year 2000 Dollars. “Chained” Dollars represent the BEA’s attempt to remove inflation from the calculation and express GDP in so called real terms; this is a complex theoretical and statistical effort (see here for BEA’s introduction of the chained data). When analysts speak of economic growth they usually refer to real or chained data, so that they ignore the effects of price changes and focus on changes in quantities.

Using the expenditures approach to GDP (familiar from Introductory Macroeconomics), GDP is equal to the sum of consumption plus investment plus government spending plus net exports (exports minus imports). I briefly discuss here the data as they appear in the GDP report and how the data are released to the public. Linked essays will discuss these data in more detail.

Consumption of goods and services is revealed in two parts: first retail sales (see my discussion here), which represents an estimate of the goods part (28% of GDP in the second quarter of 2008); personal consumption expenditures (monthly consumption of goods and services, 70% of GDP) comes out about ten days later. Goods consumption tends to be more variable than services consumption, so much of the information in PCE is already known before the release. Consumption as a whole is not highly variable but is important due to its size; many analysts have made erroneous forecasts predicting the collapse of the US consumer.

Investment is divided into three categories: nonresidential (companies buying equipment and buildings); residential (home construction) and change in inventories. There are multiple sources for these data; survey data is produced more quickly but measures of spending are more accurate. Important releases include housing starts, construction spending, durable goods, factory orders and business inventories. Investment is a smaller component than consumption (15% of GDP) but is more variable; untangling these reports is hard work, and a way that careful analysts distinguish themselves from the pack.

Government expenditure: the Monthly Treasury Statement gives the monthly spending (and revenue) numbers for the US (Federal) government (7% of GDP); the problem is that much of the government budget is transfer programs (such as social security) that do not directly enter the GDP; further, these data are missing the spending by state and local governments (12% of GDP).

Net exports: Trade reports come out much later than other data; for example, June trade data are released in August. Net exports is a relatively small component of GDP (-5.1% of GDP) but can be highly variable; when advance (the first estimate) GDP comes out, only two months of trade have been released, and the BEA actually forecasts the third month of trade in producing their GDP estimate. These numbers are difficult to forecast. Note that in recent years imports have been larger than exports, so net exports are negative and subtract from total GDP.

Understanding the Twin Deficits

Monday, August 4th, 2008

“Twin deficits” (occasionally also called the double deficits) is a shorthand summary for two related economic problems, the government budget deficit and the current account (or international trade) deficit. The government budget deficit is the difference between government revenue (mostly taxes) and government spending; the current account deficit is the difference between exports and imports (there are some adjustments for items such as funds sent abroad). Both deficits occur when someone is spending more than they earn; during the last 25 years the US government has tended to spend more than it collects in taxes and US residents have tended to spend more on imports than they export.

If an individual spends more than he earns he must either borrow money or sell off some assets. Advice columns are filled with stories of people who regularly spend more than they owe and finance the difference with multiple credit card loans that they will never be able to pay off; eventually they face bankruptcy and a reduced standard of living. Presumably a country that follows the same path will experience the same problems.

But it is well understood that individuals tend to have life-cycle consumption patterns: young people spend more than they earn (incurring student loans and mortgage debt), those in middle age tend to be net savers (investing in 401(k)s and building up portfolios and paying off their mortgage); and retirees return to net spending after they stop working.

So it is not necessary to exactly match income and spending in each period. But there are some limits to how much you can borrow if you are going to have a realistic chance of escaping financial ruin. If your lender sees that you are spending more than you will ever be able to pay then it is likely he will start charging you a higher price to reflect the increased risk associated with your debt. The same should be true for a country.


Starting in the 1980s (during the Reagan Administration) the US had both large budget and trade deficits. The budget deficit (see here for data) had been around around $50 to $75 billion in the late 1970s and grew to over $200 billion in 1983. The current account deficit (see the last line of the table here; you can select annual data for any recent period) was around zero through the early 1980s but exceeded $100 billion in 1985.

Some economists believed that the large budget deficits and large current account deficits of the early 1980s would result in higher interest rates. See here for a February 1986 expression of this view by Walter Heller, who argued that the deficits could “send interest rates shooting up”; yet yields on 3 month Treasury Bills fell from 7.29% in February 1986 to 5.75% in February 1987 and yields on 30 year Treasury Bonds fell from 8.93% in February 1986 to 7.54% in February 1987 while the budget deficit remained around the same size and the trade deficit increased a bit).

The relationship between the twin deficits and interest rates is complex. Instead of directly addressing it, I will try to provide some tools to examine the question of whether the deficits are “too large”.


It is a fact of life that in recent years virtually every national government spends more than it takes in, resulting in a deficit. These deficits are financed by the issuance of government bonds; the sum total of all past deficits is the current debt. It is important to carefully distinguish between the deficit, the gap between government revenue and spending in any given year, and the debt, or the outstanding obligations of the government.

Making the situation more complicated is that the government makes promises for future spending (for example, social security and Medicare) that do not directly enter into the calculation of government debt. And if that were not enough, the government often borrows money on behalf of others, incurring debt without increasing the deficit by an equal amount. So the data are, to use a precise economic term, messy. See here for a recent speech by a Federal Reserve Bank President who believes that the future obligations are much larger than the measured debt.

To determine whether the US budget deficit is a problem we must know (1) how much does the US government owe, (2) how much is this debt likely to increase in future years, and (3) how much is the government likely able to pay. A way to restate this question to ask whether the US debt to GDP ratio is likely to be stable or growing indefinitely.

As of 2007 the US had a debt to GDP ratio of around 61% (see here for a table of estimates of the debt to GDP ratios of 126 countries), about the same as France (64%) and Germany (63%), much smaller than Italy (104%) and Japan (195%), but larger than the UK (43%) and Spain (35%).

But the level is not the problem, it is the growth rate. For the growth rate to be stable, the size of the deficit must be less than or equal to the growth of the economy. The deficit to GDP ratio is expected to be around 2.7 to 3.2% in 2008 and 2009 (see recent projections in a pdf here at page 1), decreasing to around 1% in 2010 and 2011. The deficit projections are dependent on political and economic conditions and are far from certain (earlier this year, before the troubles in the real estate and financial markets, the CBO estimated that the budget deficit would be 1.4% in 2009 (see this pdf, until the new report is published). As long as the budget deficit is around the same magnitude as GDP growth, the debt to GDP ratio should not increase.

One big problem will occur when baby boomers (people born between 1945 and 1965) start to retire en masse around 2018. There a number of programs that pay for benefits for retirees such as Social Security and Medicare that, if the current levels of benefits are maintained (and tax revenues do not increase sharply), will cause the budget deficit to reach 10% of GDP by 2030, when debt to GDP would be over 100% (see here for several scenarios).

So, for the budget deficit, the short answer is things are fine for the next few years, but things could go very badly in twenty or thirty years. One way to look at this problem is to look at the government bond yield and to note that the interest rate on 30 year bonds is only a bit higher than the interest rate on 10 year bonds, meaning that bond investors do not seem that worried about the events of the late 2030s. Markets seem to expect the government to “do something” about the problem, possibly some combination of benefit cuts and tax increases.


The current account deficit measures the difference between the exports of US goods and services and US imports of foreign goods and services (the trade deficit measures only goods, which are increasingly less important in recent years). If the US has a deficit, this means that we are buying things from the rest of the world, and must pay for them in some way. The ratio of the current account deficit to GDP has been as high as 6% in recent years; in 2007:III the current account deficit was around $178 billion, implying a deficit to GDP ratio of around 5.1%.

The high current account deficits in recent years have led to an increase in US international indebtedness, although this is a controversial topic (briefly, foreigners who invest in the US tend to look for safety and buy bonds, which have relatively low rates of return; Americans who invest abroad tend to look for riskier assets that have higher rates of return; it is extremely difficult to track the value of these investments over time; see the pdf article on “Dark Matter” or possible inconsistencies in the measurement of the US international indebtedness here).

In any event, measured US international indebtedness to GDP (see here for some estimates) is probably around 25% and increasing. A crude estimate reveals that if the US current account deficit decreases to around 3%, then indebtedness to GDP will converge to around 50%. If the deficit to GDP ratio remains around 5 or 6%, the indebtedness ratio could rise to 100% of GDP. Most analysts believe that such a ratio around 50% would be sustainable, but a ratio of 100% would have dire consequences. So if the US current account to GDP ratio remains at current levels (or gets larger) then the US will eventually face a crisis. The Dollar has lost considerable value in recent years against some currencies, helping exporters and modestly reducing imports. Credit markets seem willing to absorb US debt at reasonable prices, indicating that they believe that the situation will revert to a sustainable level.

Understanding the Retail Sales report

Thursday, July 24th, 2008

Retail sales is, after the employment report, the second major piece of economic data to be released (see here for the date of retail sales releases) typically arriving 9 working days into the month (around the 14th), about ten days after the employment report.

Personal consumption of goods is about 28% of GDP (personal consumption of goods and services or PCE is about 70% of GDP, also including, among other items, housing and medical care). So by the middle of the next month we have a rough estimate of the performance of a bit over 1/4 of the GDP. But, as we will see below, the estimate is rather rough.

The Advance retail sales report (that is the data that are published first) surveys 5,000 firms, including 1,300 large firms that are always sampled and the remainder randomly selected to represent the entire retail industry (see here for details of the Advance report). A month later the revised report called the Monthly Retail Trade Survey is published; the Monthly report surveys 12,000 retail firms, including the 3,000 largest firms and another 1,000 firms that are the largest in their sector.

Due to the change in the sample, there is often a significant amount of revision in the data; in the most recent release the Census Department estimated that the 90% confidence interval of the Advance number was ±0.5%, meaning that if the monthly estimate is, say, +0.4%, then the true number is likely to be between -0.1% and +0.9%, but there is a small chance (10%) that it is outside that interval. The 90% confidence interval for the Monthly report, that is the data of two months before, is considerably smaller, ±0.2%. If you are confused by the language of statistical inference, the message is that the Advance data are less trustworthy than the revised data, but are better than nothing.

The questionnaire that firms fill out is fairly simple (see a pdf of an Advance Report Survey here), asking for the amount of sales, e-commerce and the number of retail establishments. An important point is that the form does not ask anything about price, just sales. A related point is that the retail sales data are nominal data, that is not adjusted for inflation; careful retail sales analysts will adjust retail sales by the appropriate inflation measure to obtain measures of real activity.

Finally, careful GDP analysts will examine the parts of the retail sales reports that the Commerce Department uses to calculate PCE. Commerce uses other sources for some of the data included in the retail sales report (notably cars, trucks and gasoline); see here for details.

Understanding Inflation Data: PPI

Tuesday, July 15th, 2008

[If you would like a brief review of ways to measure inflation and the most important U.S. monthly inflation indices you should read Understanding Inflation Data: An Introduction first and then return to this article that is about the PPI, the producer price index]

The PPI or the Producer Price Index is published by the Bureau of Labor Statistics (the BLS) an independent agency of the U.S. federal government. The BLS publishes PPI data in the middle of the month after the data were collected (data for June 2008 will be released on July 15; see here for the release schedule).

The PPI set of indexes measures price changes from the perspective of the seller, in comparison to the CPI that measures price changes from the consumer’s perspective. These two price indices may differ because of government subsidies, sales or excise taxes and distribution costs.

The BLS calculates the prices of about 100,000 prices from 25,000 estabishments every month (see here for the BLS handbook explaining the methods used to assemble the data). There are three separate reports:

  • Prices by industry (see here for the most recent table): there are price indexes for 600 industries, ranging from manufacturing to services and health care.
  • Prices by commodity (see here for the most recent table): there are price indexes for 2000 commodity groups.
  • Prices by stage of processing (see here for the most recent table): there are price indexes for finished goods, intermediate goods and crude goods, with and without food and energy.

The PPI weights prices using data from the 1997 economic census (see here) and some 2002 data.

The PPI does not attract the same degree of attention as the CPI, partially since it is not used as frequently by policy makers. In the past, the PPI was regularly published a few days before the CPI and was viewed as a forecast of consumer prices; now the CPI published before the PPI around half the time. Even so, the BLS appears to spend considerable energy improving the report. The BLS believes that the PPI is more of a work in progress, in transition from a report that measured the prices of goods to a report that will measure the entire domestic output prices of goods and services.

The most interesting use of the PPI (in this economist’s view) is as a way to view the transmission of inflation through the economy, the so-called “pass through”. By looking at the stage of production indices you can see whether increases in commodity prices in crude goods result in increases in prices of finished goods. A great puzzle of recent years has been the lack of pass through, meaning that increases in the prices of oil and food have not, so far, resulted in increases in finished goods prices.

Understanding Consumer Confidence Measures

Wednesday, July 9th, 2008

One of the biggest problems facing an economic analyst is the lack of timeliness of published data. Most economic numbers are published a month or two after the data are collected, due to the difficulty in collecting and assembling the data (for example, the CPI measures the prices of 80,000 goods and the employment report surveys 60,000 households and 160,000 businesses; both reports are released between three weeks and a month after the surveys are conducted). Measuring what people have done in an economy of 300,000,000 residents is hard work. It is much simpler to contact a small group of Americans and ask them how they are feeling.

The reports on how Americans are feeling are called surveys of consumer confidence or sentiment. There are two main measures of consumer sentiment, the Conference Board and the University of Michigan. The Conference Board publishes (among other data) the Consumer Confidence Survey (available by subscription, see here for a sample issue) that asks 5,000 households by mail how they are feeling (see here for a description of the Conference Board methodology, which I can no longer find on their website). The University of Michigan Survey Research Center (see here for past data and survey description, registration required) does 500 telephone interviews asking people in the mainland U.S. (residents of Alaska and Hawaii are excluded) about their perception of the current and future state of the economy.

The obvious advantage of these data is that they can be published almost immediately. The obvious disadvantage is that the data are measuring what people say and not what they are doing. While both entities have published studies demonstrating the usefulness of their data, there are numerous skeptics (see here for one) who argue that at best confidence surveys reflect the current economic situation but are not good at predicting the future. An earlier, more careful study (see here) finds some evidence that the Conference Board data can be useful at predicting consumer spending (but note that one of the co-authors is a former employee of the Conference Board, see here). But more recent studies (see here) are considerably more skeptical. Recent Conference Board data (see here) show that confidence was at a relatively high level as recently as July 2007, just before the subprime problems affected markets and the price of petroleum started to climb.

There is a third confidence measure published by the Washington Post/ABC News that appears weekly that does not attract as much attention as the two monthly surveys. The survey does telephone interviews with 1000 randomly selected adults (see here for details) and data are released every Tuesday evening at 5pm. You can examine past polling data here or see a table of past data here. If you look at the historical data, you will find out that the answer to the question “Would you describe the state of the nation’s economy these days as excellent, good, not so good, or poor?” is generally “not so good” or “poor”. From 1986 to 2007, there were only five years when a majority of respondents described the economy as excellent or good, from 1997 to 2001. While there are numerous problems in the US, this view seems overly pessimistic.

There are other measures of how consumers feel about the economy (available here; there are also a large number of polls on more traditional topics such as politics) but these days markets seem to attach little weight to these surveys.

The media report the confidence data as if they represent important news; see here for a newspaper article about the July 2007 surge in the Conference Board consumer confidence index. While forecasting the economy is always difficult, economists who believed that “[the] rebound in confidence suggests economic activity may gather a little momentum in the coming months” probably made worse forecasts than economists who ignored the report.

Understanding US Inflation Data: CPI

Tuesday, June 24th, 2008

[If you would like a brief review of ways to measure inflation and the most important U.S. monthly inflation indices you should read Understanding Inflation Data: An Introduction first and then return to this article that is about the CPI, the consumer price index]

The CPI or the Consumer Price Index is published by the Bureau of Labor Statistics (the BLS) an independent agency of the U.S. federal government. The BLS publishes CPI data in the middle of the month after the data were collected (data for June 2008 will be released on July 16; see here for the release schedule).

The CPI-U index (often referred to as the CPI or headline inflation number; there is much information here) measures prices for the average urban household, representing about 87% of US consumers (the index does not measure prices paid by people living on farms or in rural areas).

The BLS calculates the basket of goods consumed by urban residents conducting interviews of around 30,000 households; they include about 200 categories of goods including a sample of 80,000 individual items. The weights of these items in the basket is now adjusted every two years. There is a fair amount of controversy about the categories contained in the sample (in particular concerning how the price of housing is measured, see here for the detailed explanation by the BLS; the issue is that the BLS tries to compute how much it costs to live in a house but tries to exclude the gains and losses experienced by homeowners as investors in housing), but in general the BLS chooses certain items, such as a 4.4 pound bag of golden delicious fancy grade apples to represent the apples category and then adds all the prices together (divided by their weight in the consumption basket) to arrive at a price index.

As technology changes it becomes very difficult to measure the market basket. New products are introduced every year (how much did an iPhone cost 2o years ago? see here for a nice comparison of a 2008 model iPhone and a state of the art 1988 Motorola DynaTAC 8000X that cost $4000 when first released) and the BLS must change the basket of goods frequently to keep up with consumers.

In 1996 the Boskin Commission report found that the CPI overstated inflation by about 1.1% due to biases related to substitution, new goods and quality change. At the time there was general consensus that the CPI overstated inflation, but much disagreement about the magnitude. Substitution bias means that when the price of one good (say Coca Cola) rises, consumers will buy more Pepsi Cola or other similar soft drinks without being much worse off; an index that assumes that the basket does not change will overstate inflation. New goods and quality change speak to the problem of measuring the price difference between the latest iPhone and its predecessors.

Other studies found a bit less bias but still believed that CPI was too high (see here for a review of later studies). More recently critics of the CPI such as Bill Gross (see here and here) have argued that the BLS overcorrected and now the CPI understates inflation; but there are others (see here) who believe that the CPI fairly accurately reflects the inflation picture.

It is important to remember that the CPI is an attempt at measuring the level of prices for an urban consumer and that the composition of the urban consumption bundles has changed dramatically in the last 40 years. There has been an explosion in the number of new products; the number of products in supermarkets has increased from around 8,000 in 1970 (see here) to around 45,000 today (see here) and the internet allows consumers access to a much wider set of retailers. While the CPI might accurately capture the average prices paid by urban consumers, differences in consumption bundles have certainly increased. Further, consumers are especially aware of food and gasoline prices and seem less aware of service prices (how much are you paying per channel for cable/satellite TV versus a few years ago? The price has probably gone up but so has the number of channels) . Complaints about the accuracy of the CPI are likely to be greatest when (as is the case now) food and energy prices are rising rapidly. It is probably safe to conclude that complaints about the CPI’s accuracy are likely to increase over time along with the diversity of consumer bundles purchased.

Understanding US Inflation Data: An Introduction

Tuesday, June 24th, 2008

Measuring inflation is more complicated than it would seem at first. It is fairly easy to view the price of gasoline from the big signs at many major intersections. Average gas prices are almost double the level of a year ago (see here for US retail gasoline prices); many Americans buy gas regularly and are aware of the price rise. The prices of some other goods have also risen sharply, from vegetable oils and rice and wheat to cable television and health insurance. If all these prices are rising (as well as others like college education and airfare) then it seems hard to believe the government’s published numbers that claim that overall prices (as measured by the Consumer Price Index) were up 4.2% in the last 12 months; the claim that core prices (excluding food and energy) were up only 2.3% is likely to be treated derisively by those who believe that the price of the things that went up from the inflation calculation (see here for one of the more carefully thought out expressions of this view).

Yet official inflation series present fairly low estimates of inflation, raising speculation about secret government efforts to understate the inflation rate (see here for a fairly mainstream version of this argument by a famed bond fund manager; articles such as this one from Bloomberg News use slightly more provocative language).

To better follow these arguments I will first discuss some of the problems of measuring inflation. Future topics will include a review of several popular monthly inflation indices, the CPI, PPI and the PCE deflator.

First a bit of definition: a price index tries to capture the level of prices at a specific time; the change in the level of prices is called inflation (or deflation, if the change in prices is negative). Note that inflation and deflation are merely statements about the change in an index and not any statement about other economic conditions. Because deflation has occurred during some difficult economic times (such as the 1930s in the US or more recently in Japan) some people confuse deflation with recession or depression. But deflation and inflation are merely statements about the direction of change of the price level and nothing more.

There are numerous problems with calculating price levels and inflation; I will briefly speak of three of them, though there are others. Consider first a very simple economy with one good that does not change over time; the price level in this economy is just the price of the one good. It is very easy to compare prices over time; some years the price of this good will be higher and there will be inflation, other years it will be lower and there will be deflation.

The three major problems when we try to move from the simple economy described above to the actual economy of today arise from the observation that there are many goods (and many prices), so (1) our consumption basket changes over time, (2) the goods change over time and (3) not everyone consumes the same basket.

The people who believe that inflation is overstated argue that the price of gasoline (as noted above) has doubled in the last year. But (1) many of us consume less gasoline now than we did a year ago, driving less or using more efficient cars; (2) gasoline has changed somewhat over time, with the addition or removal of certain ingredients to increase the octane rating or reduce the negative effects of gasoline on the environment and (3) different people consume different amounts of gasoline, depending on where they live, what kinds of cars they own and how much they drive. So if the gas price is up 100%, but there is a new additive that has slightly increased the octane level and I drive much less (but still much more than you) how exactly do we compute the appropriate inflation number?

Also, many of us use more computer services than we did, say, ten years ago; some of those services are far cheaper than they were (if they even existed then). High speed internet was fairly new and expensive in the late 1990s and it is now cheaper and faster; I probably spend more each year on computer-related services (including hosting this web site) than I do on gasoline.

Without getting into the mathematics (but there is a reasonably clear explanation in Wikipedia if you are interested) there has been a lot of time devoted to trying to carefully deal with these issues but each method involves serious compromises.

As noted above, there are a number of different inflation measures (calculated by different government agencies) that are used to describe inflation; I will provide a brief overview of the different measures here with more detail in specific essays about each measure.

The CPI (consumer price index) calculated by the BLS (Bureau of Labor Statistics) is the best known U.S. price index. CPI data are published once a month and try to measure the prices paid by urban consumers for goods and services. Two numbers attract the most attention, the overall inflation rate and the “core” rate (excluding food and energy); the first number measures the price of the entire basket and the second subtracts the cost of food and energy that are typically more volatile than other prices, meaning that large changes tend to be reversed over time. Some analysts believe that recent price increases in food and energy are unlikely to be reversed and they tend to ignore the “core” rate (which is much lower than the overall inflation rate).

The PPI (Producer Price Index) is also published by the BLS. In contrast with the CPI (which tries to measure prices from the perspective of the ultimate buyer of the goods and services) the PPI measures prices from the perspective of the seller. To the degree that retailers pass through the prices that they pay for goods, changes in the PPI should be a good predictor of future changes in the CPI (if the price of fish sold by wholesalers rises, the likelihood is that stores will raise their prices to retail customers). But in recent years the link between wholesale and retail prices has been less clear, possibly due to changes in the way business is done. Financial markets appear to pay less attention to PPI than CPI (one curious note is that up until a few years ago PPI data were routinely released prior to CPI data, but now CPI data are released first; the earlier release of the PPI led some to use it as an aid in forecasting the CPI).

The PCE deflator starts with the CPI data but also incorporates other data. The Bureau of Economic Analysis (the group within the Commerce Department that calculates the GDP) calculates monthly estimates of Personal Consumption Expenditures as part of the Personal Income report. The most important difference between the two measures of inflation is that the CPI uses a fixed-weight market basket (reset every ten years or so) that assumes that consumers continue to buy the same goods even if prices change; in contrast the PCE uses a so-called “chain link” method that reflects the monthly changes in the basket. Also, the basket for the PCE is somewhat larger than the basket include in the CPI (see here for a brief description of the differences). The PCE data are reported after the CPI\ data (data for April 2008 CPI were released on May 14, data for the PCE deflator were released on May 30); the Fed has used the PCE core index (ex-food and energy) as its measure of inflation (see footnote one in this PDF document)

Understanding US Unemployment Claims Data

Tuesday, June 24th, 2008

Every Thursday, at 8:30 AM US Eastern Time, the Bureau of Labor (BLS) releases the data on initial claims for unemployment insurance (see here or here for recent reports). In the US most employers pay a small unemployment insurance tax; in general, if a worker has worked for more than a year and loses a job through no fault of his own, the worker can apply for unemployment benefits, but not all applications will result in benefits (see here for more). To continue to receive benefits the worker must be available for work (ready and willing to accept suitable work, and make a personal and continuing effort to find work). The benefits are related to the salary the worker earned, roughly half the weekly salary for most workers.

The great advantage of these data is their timeliness: every Thursday the data for the week ended the previous Saturday (five days earlier) are released, along with revisions to the previous week’s data. In contrast, the employment report is released around three weeks after the data are collected, and most other data considerably later.

The principal disadvantage of the data is their noisiness: there are many reasons for employers to fire workers that are distinct from movements in the general economy. There are strong seasonal patterns to employment (for example, there are many retail workers hired before Christmas who are not needed shortly thereafter; large industrial employers often shut down factories for maintenance or to prepare the factory for the preparation of a new product and temporarily lay off their workers; this allows employees to try to get some money from the government when employers do not need them).

Without getting into too much detail, the BLS seasonal adjustment procedure (correcting the data for typical seasonal patterns) is a bit mysterious. The seasonal factors to adjust data for the last few years through April 2009 can be found here. In early January, the seasonal factor peaks at 173, meaning that the actual number of claims is divided by 1.73 to produce the seasonally adjusted number; in September the seasonal factor of 76 means that the actual number of claims is divided by 0.76, so that actual claims of 300,000 would be seasonally adjusted up to around 395,000. Once you factor in the complexity of four day weeks (there are around 12 national holidays plus the occasional state holiday), you can see that it is hard to put too much emphasis on a single weekly number.

For this reason, some analysts prefer to look at the four-week average of the data as a way of smoothing out some of the variability of the data. There have been a number of cases in recent years when claims were up (or down) sharply for a few weeks but then returned to the range of the previous few months.

In general, in recent years when the economy is doing well the average initial weekly claims number is around 300,000; when the economy is in recession the number is closer to 400,000. An amazing part of the U.S. economy (especially relative to some European economies) is the amount of “churn” in the job market. Consider a month when the job market is poor, and employment is down by 100,000 and weekly jobless claims are 400,000. Rough calculations would imply that 1.6 million people lost their jobs (4 weeks of 400,000) and 1.5 million people found jobs (resulting in net jobs decreasing by 100,000). So even in a bad month, over 1.5 million Americans are hired for a new job. The dynamism of the U.S. economy means that firms fire workers on a regular basis, but there are usually other firms that are hiring.

Along with initial claims the BLS also releases continuing claims data. A worker who loses his job files an initial claim for unemployment insurance once, but may continue to receive benefits for months before he finds another job. To continue to receive benefits an worker must file regular reports (see here under continued eligibility). In general, during good economic times unemployed workers will find jobs quickly and stop receiving benefits; analysts thus view a rising continuing claims number as a signal that the economy is weak. But you must be careful because there are limits for how long workers can receive benefits, generally around 26 weeks (six months); after this time the worker can no longer receive benefits, even if he has not found a job. However, during difficult economic times Congress will extend this period to a longer period (39 weeks).

Understanding the US Employment Report

Tuesday, June 24th, 2008

The employment report is the single most important US economic data release. The importance of this release is due to its timeliness (the report comes out ahead of most other monthly economic data) and its scope (it is the result of a two large surveys: the household survey of 60,000 households and the establishment survey of 160,000 businesses and government agencies covering 400,000 work sites). The report is viewed by some analysts as a sort of monthly GDP number, due to the strong correlation between employment and growth.

This article will not make you an expert analyst of this report (if you want to be an expert I would suggest a year or more of graduate school in economics followed by a year or more working at the economics department of a bank or investment bank as a better start). But it will tell you a few things to look for and provide you with some context to understand what it means.

First, as was noted above, there are two surveys. One of them (the establishment survey) gets most of the attention from traders. The key number (announced, with the rest of the data at 8:30 am Eastern Time on the first Friday of every month and available shortly thereafter here) is the change in nonfarm employment, or the change in the number of workers as calculated by employers and adjusted by the Labor Department. There is much speculation about this number, and if it is greater than expected (and greater than the so-called “whisper” number that certain traders circulate just before the official number is released) then typically equities will rise (in general, strong economy, strong stocks) and bonds will fall (strong economy, more likely that the Fed will raise rates).

But while markets react quickly to this number, it is somewhat controversial, for several reasons: (1) the imprecision of the initial estimate; (2) problems in counting new businesses and the self-employed; (3) the timing of employment relative to the business cycle. I will now discuss the criticisms of the headline number in turn, and then discuss other parts of the report:

Imprecision: The establishment survey is revised in the two months following the initial release (as well as several more times in later years). For example, in September 2006, the initial number was +51,000, a number described as “dire” by one analyst. But in October, this number was revised to +148,000 and in November it was revised again to +203,000. Presumably the analyst viewed the revised number a bit differently. During 2006, the range of the revisions (see here for a table) from the initial announcement to the second revision was from -43,000 to +152,000. The Bureau of Labor states that a 104,000 monthly change is statistically significant. All this implies a bit more randomness in these numbers than some analysts seem to believe. A further complication is that the news from the headline number is often offset by recent revisions; if the headline number is below expectations, but the two previous months have been revised up, is that good news or bad? Thoughtful traders may conclude that markets tend to overreact a bit to these numbers.

New Business and the Self-Employed: Since the establishment survey works with existing businesses it cannot measure newly-formed businesses or count the self-employed. The Department of Labor attempts to correct for these omissions by including an adjustment to the numbers for these factors, based on a model. Analysts who believe that the economy is weak tend to argue that these adjustments overstate employment.

Timing: Many companies may attempt to keep workers even after business has turned down. It is expensive to find, hire and train workers; despite all the announced layoffs by firms who have experienced losses, the evidence suggests that many firms prefer to maintain workers on their payrolls even when demand for their products is down so that they will be prepared for the next upturn. Thus, the economy may turn down before employment actually drops. Similarly, when the economy has started to turn up, firms may be reluctant to hire workers until there is ample evidence that business is good. Economists believe that the decision to hire and fire a worker is viewed as an investment project, differently from the decision to buy supplies for the office. A business that is losing money due to a lack of customers can sell off their extra supplies that will not be needed, confident that when business is better they can buy new supplies at the going price. But a company will try to avoid firing workers because there is a significant cost to finding and training new workers.

As I have noted above, the employment report is much more than the headline number from the establishment survey. I will briefly mention two other parts of the report that attract attention, the household survey and the index of hours worked.

The household survey is used to calculate the most politically important number, the unemployment rate. The household survey asks people who are not employed whether they were available for work and have made specific efforts to find employment during the past four weeks; they further state that whether an individual is eligible for unemployment benefits does not enter into their analysis. The household measure of employment is much less accurate than the establishment survey, with a 90% confidence interval of plus or minus 430,000, so it is usually ignored by most market analysts, but the unemployment rate (particularly among certain groups like minorities or women) can be a very important number for political analysts. Fed Chairman Greenspan used to pay attention to the rate of job leavers (see Table A8), the percentage of workers who left a job voluntarily; he believed that this was an indicator of the strength of the labor market.

The index of hours worked (see Table B5) attempts to measure not just how many people are working but how much total work they are doing. The Labor Department calculates how many hours per week nonsupervisory workers are working and then creates an index. While heroic assumptions must be made to create an index (similar to the assumptions of any index, in which very different things are summed together), this is an interesting monthly indicator or the US economy. It somewhat accounts for the timing problem cited above, in that firms can adjust how much work is done (via a shorter work week) more easily than adjusting the number of workers.

Understanding the US Economic Data Schedule

Tuesday, June 24th, 2008

Here is a schedule of the major US economic numbers in the order in which they are released for May 2008. The key to understanding data releases is to see how they fit in to the overall economic picture. There are many economic calendars available on the internet for free such as here and here.

The best overall picture of the economy is the GDP (approximately the measure of all goods and services produced in the US during the calendar quarter); as you can see from the table below, the first estimate of GDP for the second quarter of 2008 (which covers the period April 1 until June 30) will be released on July 31, 2008 and then be revised twice, producing a final number on September 26. But there will be much information available to the market about the economy in May 2008 before then, starting with the unemployment claims data released on May 8.

Economists who analyze the data follow all these releases (some more than others, based on their experience and research) and form a continually changing picture. Releases in red attract more than the usual amount of attention. Alert investors will be aware of upcoming data releases that can affect the markets and plan accordingly. As this site is updated, there will be links to various specific releases such as the employment report; by clicking on these links you can learn more about the informational content of the report. There are many ways to use these data (there are many economists who find employment analyzing these data at different institutions) and there are almost as many views as to which releases are most important. What follows are one economist’s views.

Below is the release schedule for May 2008 data; while the order of data release varies a bit from one month to the next, thinking about these releases as a group should help you understand how the market learns about the economy.

The general pattern is that survey data is available first; these include surveys of businesses by regional Federal Reserves and surveys of consumers. Surveys of the labor market are also available relatively early, including weekly jobless claims (the number of workers who lost jobs who are applying for unemployment benefits, see here for my take) and the employment report (a survey of businesses and households that measures the amount of work done and the number of people who would like to work but have not found jobs; see here for a fuller description).

In general, economists pay more attention to data about what people and businesses have done (employment, retail sales and so forth) than what they say they might do (consumer and business surveys). But economists are willing to work with whatever they have and use the survey data until the actual data arrive later.

Date Release
May 8 Unemployment Claims (description)
May 15 Philadelphia Fed Survey
May 15 New York Fed Survey (Empire State)
May 16 Michigan Sentiment Index
May 27 Conference Board consumer confidence
May 30 Chicago PMI
June 2 ISM index
June 3 Auto sales
June 4 Nonmanufacturing ISM
June 6 Employment report (description)
June 11 Treasury Budget
June 12 Export and Import Prices
June 12 Retail Sales
June 13 Consumer Price Index
June 17 Producer Price Index
June 17 Housing Starts
June 17 Industrial Production
June 19 Leading Indicators
June 25 Durable Orders
June 25 New Home Sales
June 26 Existing Home Sales
June 27 Personal Income and Outlays
July 1 Construction Spending
July 2 Factory Orders
July 10 International Trade
July 12 Business Inventories
July 31 Advance 2007:III GDP
August 28 Preliminary 2007:III GDP
September 26 Final 2007:III GDP

Bad Behavior has blocked 35 access attempts in the last 7 days.

FireStats icon Powered by FireStats