The monthly BLS Nonfarm Employment Forecast will now be made available as an offline service to new or existing WANTED Analytics™ subscribers. The monthly forecast will not be part of the standard subscription to the WANTED Analytics platform of applications – the Nonfarm Forecast is offered as a stand-alone product subject to its own service level agreement. Anyone interested in this service is invited to contact us.

WANTED will still provide guidance on US labor market dynamics on this blog through a monthly 'National Hiring Demand Outlook' article.

This decision is driven by three considerations:

• Most professional forecasters already maintain models of their own. An additional forecast or model output is of little use to practitioners. Instead, WANTED's services have focused on integrating Hiring Demand into existing forecasting equations. Particularly relevant is the specification of Hiring Demand autoregressive terms (AR terms), as well as their interaction with other labor market variables. What's more, WANTED provides both weekly and monthly data, so forecasters can use either datasets depending on their preferences. Only a separate subscription service can allow sufficient technical support to properly incorporate Hiring Demand data into sophisticated econometric models.

• BLS revisions to Nonfarm Employment counts are a fact of life in the employment forecasting world. "What the forecast would have been" under revised data is different than what is produced given the preliminary data. As such, the same Hiring Demand data produces different forecasts as the BLS revises its data once, twice, and finally thrice. This is because most models include previous values of employment. In other words, monthly Nonfarm Employment counts are a moving target. Providing a public forecast, once and only once, does not adequately accommodate the fact that the same forecasting model produces different results as the BLS revises its data. Once integrated into forecasting models, however, the marginal impact of Hiring Demand can be quantified even as the BLS revises its data.

• Hiring Demand is so closely related to published measures of economic activity that providing derived output free of charge is not a sustainable long-term business practice. Hiring Demand explains 94% of future changes in Canadian Employment levels, explains 80% of future changes in new Unemployment Insurance Claims, and explains 78% of S&P 500 index returns.

Hiring Demand has proven to be a valuable tool in predicting or forecasting changes in employment, unemployment insurance claims, and even stock market returns. The number of online job ads for the companies comprising S&P 500 Sector Indices are also highly predictive of returns for that index. What's more, Hiring Demand leads changes in US Nonfarm Employment by two to three months.

The number of online job ads is highly correlated with published measures of labor market activity and aggregate economic output. Using new online job ads as a labor market signal – specifically an intent to hire – WANTED has published 13 forecasts of changes in US employment levels. Each forecast predicted changes in Nonfarm Employment counts as reported by the US Bureau of Labor Statistics in its monthly Employment Situation Summary.

WANTED's forecasting accuracy approaches the bounds permitted by the noise inherent to the BLS time-series:

• Forecasting models based on univariate seasonal ARIMA models of the form (3,1,1)x(1,1,1,12) have a forecast standard error of 108,000 workers/month.
• WANTED's forecasting model, which layers in Unemployment Insurance Claims and Hiring Demand data, has a forecast standard error of 84,000 workers/month. This is an improvement of 24,000 workers/month or a 22% improvement in forecasting efficiency.
• Finally, remember that the BLS reports employment data for the same month on three separate occasions: there are two preliminary releases and then a final tally, with successive updates coming at monthly intervals. If the BLS's "first preliminary" release is interpreted as a forecast of its own third and final employment report, the BLS has a forecast standard error of 78,000 workers/month. In other words, the BLS has a forecast standard error of 78,000 workers/month relative to itself, it is very difficult for any forecasting model to better that level of accuracy.

The predictive value of Hiring Demand is derived from the length of the hiring cycle, the duration of a typical online job posting, and the nature of online job advertising:

• New online job ads signal an intent to hire. WANTED only uses "paid-for" job postings in its forecasting models (free sites such as Craig's List are omitted). There is no incentive for a business to pay for an online job posting if it's not going to hire a new employee – '30-day' job postings on national job boards usually cost around $350. In other words, online job ads are a significant labor market signal, particularly when companies have to pay for that signal.

• Most major online job boards sell '30-day' job postings. 60-day options are also available. Companies will want to utilize their 30-day advertising window to collect the greatest number of resumes/candidates possible. The first set of interviews may occur before the job ad expires, but because the job ad is paid-for, there is no reason for an employer not to wait until the job ad expires until it hires somebody. So the 30-day ad length provides at least a 4-week predictive signal on hiring because companies should wait until the job ad expires before it hires someone.
• The 4 to 8 week long hiring cycle can either coincide with or be added to the 4-week ad duration. Once the job ad expires, a company will usually proceed to first round interviews for 2 to 4 weeks, depending on candidate and hiring manager availability. Then, the next set of second round interviews can last 2 weeks. Once an offer is extended, a new hire who is currently employed may or may not have to provide the typical 'two-weeks notice'. Econometric models confirm that Hiring Demand leads changes in Nonfarm Employment by 2 to 3 months, depending on the type of job being advertised.

© Copyright 2010 WANTED Technologies Corporation

In the last 4 weeks, 1.35 million new job ads were posted on paid-for online job boards. This includes 158,000 postings on CareerBuilder, 157,000 on Monster, 64,000 on HotJobs, and 71,000 job ads on Dice, an IT specialized job board.

There were 37,000 more online job ads posted in April compared to March. Considering the data collected so far for May, those monthly gains have remained stable at 33,000 new online job ads.

This growth comes after a solid first part of 2010 in terms of labor demand. The US was finally able to start generating gains in employment, even if we remove temporary hiring at the Census Bureau.

Compared to January, even the number of new Unemployment Insurance Claims fell from 490,000/week to around 450,000.

Source: WANTED Analytics

However, recent growth in Hiring Demand has decelerated a bit, although it remains positive in absolute terms. April did see an increase of 37,000 online job ads – but we had seen gains of 100,000 for March, 130,000 for February, and 143,000 for January. Even the number of new UI claims showed weakness last week, rising by 25,000 new claimants.

Wall Street has also noticed this deceleration. The S&P 500 is still up 19% compared to last year, but it has fallen 12.4% in the past month.

Hiring Demand tracks S&P 500 stock price returns quite well. The correlation between the two over the past three years has been 0.70. In other words, 70% of the variation in S&P 500 index returns can be explained by changes in Hiring Demand. The following graph shows that yearly gains in stock prices and labor demand have receded somewhat, after showing strong acceleration following the long decline into recession. Both lines are above the 0% line, marking year-over-year growth, but both of them have shown weakening in the past two months.

Source: WANTED Analytics, Google Finance

© Copyright 2010 WANTED Technologies Corporation

Last month, the BLS said that US Nonfarm Employment dropped by 36,000 workers.

Full-size chart available on Jobs. Trends. Insight.

This month's forecast is based on the following labor market dynamics:

• Particularly meaningful for this month's forecast are offsetting changes in new Unemployment Insurance Claims:
  • There were consecutive increases of 32,000 and 24,000 new UI Claims for the weeks ending February 13th and February 20th 2010. The week ending February 20th had a total 498,000 new UI claims. The BLS measures employment on the 12th day of each month, so technically the report ending February 13th can be excluded since it overlaps with February's employment count. However, it still provides a signal on what happened in the labor market and is a close straddle on the March employment reporting date.
  • More recent UI claim reports have been positive. New Claims fell by 30,000 for the week ending February 27th, settling at 468,000. In the past two weeks UI Claims have improved marginally as well, falling to 462,000 and 457,000 for the weeks of March 6th and March 13th. This is a good signal as it indicates continued positive momentum in the labor market. The 4-week moving average has returned to levels around 470,000 and will fall precipitously when the value of 498,000 drops out the 4-week window used this week's release.
  • The 4-week moving average now stands at 471,250. This is an increase of 3,500 claims relative to the 4-week moving average of 467,750 seen a month ago. Over a 4-week period, that means a cumulative total 14,000 more new Unemployment Insurance Claims.
• New online job ads on major national job boards grew by 9,500 ads in March. This is a slight softening after 4 consecutive months of improvements in the range of 12,000 to 14,000 online job ads. Changes in Employment depend two variables – separations and hires – and we've discussed in a previous post how it's possible to have both growing hiring and falling employment.

Both of these factors combined – a slight increase in new UI claims, and a slight softening of labor demand – lead us to forecast a drop in US Nonfarm Employment of 52,000 workers. This is a relative worsening of 16,000 workers compared to last month's drop of 36,000 workers. There was no surge in Hiring Demand that would lead us to forecast employment gains.

Employment changes will fluctuate around zero as the economy and labor market have reached a steady-state equilibrium.

Over the past three years, the average difference between the BLS's first preliminary data release and its third and final data release has been 78,000 workers a month. The BLS's first preliminary report is treated as a forecast of its third and final report. The WANTED historical standard of error of forecast is 84,000 workers per month, which is consistent with the BLS's average error to its own data.

The BLS will release its Employment Situation Summary for March on Friday, April 2, 2010, at 8:30 a.m

© Copyright 2010 WANTED Technologies Corporation

December and January employment data is highly volatile in terms of seasonal fluctuations. Job ads usually drop heavily during Christmas-time and then pop back up in January.

What's more, there's something we call "the January bounce" – not only does January Hiring Demand come back from the Holiday slump, it usually jumps above previous December levels. New budgets are usually the source of this "January Bounce" – there's a sort of pent-up demand for labor that's waiting for a new budget cycle to kick in.

During the last week of December, UI Claims benefited from a smaller than expected seasonal dip in Hiring Demand (UI claims improved by 22,000). This week, however, UI Claims suffered from a smaller than expected January rebound in Hiring Demand.

The two tables below show historical December/January seasonal fluctuations, compared to what we've seen so far this year.

Whereas we expected Hiring Demand to drop around 28.2% at the end of December because of the Holidays, it in fact dropped only 19.7% – this is what drove new UI claims down by 22,000 new claims. We were expecting a 24.8% jump in Hiring Demand this week, yet we only saw a 21% jump – and UI claims rose 11,000. (We label the UI claims as "Improvements" in the table below : a negative sign in front of UI changes is actually a good thing, as there are fewer unemployed, so the sign has been flipped momentarily to facilitate interpretation…). It could be that the current economic climate "dampens" seasonal movements. The good news is that over the past couple weeks we've still seen a net improvement of 11,000 new UI Claims  (22,000 for last week minus 11,000 for this week's release). (The bounce in 2008 was a little bit more spread around. The difficulty in seasonally adjusting weekly data is were "how exactly" a holiday falls matters greatly. January 1st falling on a Friday has a different effect than if January 1st fell on a Wednesday, for example – that entire week may have to be discounted more heavily in terms of seasonal adjustments).

Table : December/January Hiring Demand Seasonal Movements

Source: WANTED Analytics data

In 2006 and 2007, Hiring Demand rebounded by over 55% in the first two weeks of January. January 2008 was the start of the recession, so the bounce was less pronounced at 51.7%. For 2009, the financial crisis had just struck, and we saw a measly bounce of 35.2% . We'd need a jump of over 30% for the second week of January this year to compensate for last week's soft rebound. This is possible, but we're more likely to see something around 25%.

Table : January Hiring Demand Rebound, Historical

Source: WANTED Analytics data

Below is a 4-year Hiring Demand time-series, where a small horizontal hash has been inserted to show the break between the first and second week of January. This coming week will be a clear signal of how businesses will hire in 2010, as new budgets are rolled out and hiring activity finds its way to online job boards.

Source: WANTED Analytics

(**: WANTED Analytics is a highly flexible data querying tool. The sites which are used to construct Hiring Demand Indicators can change, depending on, for example, significant price changes for a job board. If a job board drops its price by 50%, the rise in online job ads isn't macroeconomic, rather it's a price-elasticity response to cheaper advertising. Such modifications were carried out for this week's analysis, and the time-series graph above is not directly comparable to those seen in previous posts).

© Copyright 2010 WANTED Technologies Corporation

This news also confirms our expectation that the Bureau of Labor Statistics will announce increases in levels of US employment for December.

The correlation between the number of new online job postings (Hiring Demand) and unemployment insurance claims has been -0.72 over the past 4 years. As more new online job ads appear on the internet, more people can find work, and fewer must file for unemployment insurance – hence the <negative> sign on the correlation coefficient. What's more, the correlation improves to -0.80 when looking at the one-week lagged value of Hiring Demand, supporting the hypothesis that Hiring Demand leads UI claims by a week. (Remember that the "best correlation" possible is -1.0 when two variables move in opposite directions).

Given the strength and robustness of this relationship, we can forecast changes to UI Claims data using the number of new online job ads. WANTED Technologies predicts that new UI Claims will fall by 18,000 on a seasonally adjusted basis for the week ending January 2^nd.

The following scatter plot, where each dot represents a weekly data point, shows the structural relationship between new UI claims and the number of new online jobs ads. It is this relationship, estimated in an econometric model, that allows us to produce a forecasted value.

Source: WANTED Technologies Hiring Demand Indicators

© Copyright 2009 WANTED Technologies Corporation

The correlation between the number of new online job ads (Hiring Demand) and unemployment insurance claims has been -0.72 over the past 4 years. As more new online job ads appear on the internet, more people can find work, and fewer must file for unemployment insurance – hence the <negative> sign on the correlation coefficient. What's more, the correlation improves to -0.80 when looking at the one-week lagged value of Hiring Demand, suggesting that Hiring Demand leads published measures of labor market dynamics. (Remember that the "best correlation" possible is -1.0 when two variables move in opposite directions).

Given the strength and robustness of this relationship, we can forecast movements in the seasonally adjusted value of new Unemployment Insurance Claims: WANTED Technologies predicts that new UI Claims will fall by 6,000 on a seasonally adjusted basis for the week ending Dec.19.

The following graph shows the historical relationship between the number of new online jobs and new Unemployment Insurance Claims, on a seasonally unadjusted basis.

Full-size chart available on Jobs. Trends. Insight.

© Copyright 2009 WANTED Technologies Corporation

This week, Hiring Demand grew only 3.9% on a seasonally unadjusted basis. Over the past 5 years, the first week of December usually sees a little bit of a rebound in Hiring Demand. This year's "post-Thanksgiving rebound" was the lowest we've seen in 5 years. This is what drove new UI claims upward. Even last year, when the economy took a huge hit, the first week of December saw Hiring Demand rebound almost 25% after Thanksgiving.

Table: Week-over-Week % Growth, Hiring Demand Indicators, First week of December

The graph below shows the historical relationship between Hiring Demand Indicators (the number of new online job ads) and seasonally unadjusted Unemployment Insurance claims. The UI claims axes has been inverted: as new job ads appear online, fewer people must file for unemployment insurance. The correlation between the two variables, over the past 4 years, has been -0.72. What's more, the correlation between new UI claims and the one-week lagged value of Hiring Demand improves to -0.78, suggesting that Hiring Demand leads new UI claims.

Full-size chart available on Jobs. Trends. Insight.

© Copyright 2009 WANTED Technologies Corporation

Since new UI claims reached a peak in late March 2009, they have trended downwards 3% a month on average. Hiring Demand (the number of new online job ads) has improved at an average rate of 1% a month since then too.

Last week's report of a drop of 39,000 claimants came as a surprise. Some have suggested that the economic conditions we're in have "tricked" the seasonal adjustment factors. Bradford DeLong, for example, suggests that because there aren't as many construction workers employed now compared to historical levels, the seasonal increase in new UI claims that comes from them being laid off right before Thanksgiving didn't really happen this year.

Our data, however, confirms that last week's improvment of 39,000 UI claims was in fact "real". (The initial report was for 35,000). Over the past several years, the number of new online job ads usually drops about 24% during Thanksgiving week. This year, the drop was only 11.6%. In other words, given the seasonal patterns we've seen in our data over the past five years, the improvement in the UI Claims report make sense:

What makes us think that there's any relationship between this lower than expected drop in new online job ads and the sudden improvement in new UI claims?

There is, in fact, a close relationship between the number of online job ads and new UI claimants: the correlation between the two has been -0.71 over the past two years. What's more, the correlation between new UI Claims and the one-week lagged value of Hiring Demand is -0.78, which means that the number of online job ads leads new UI claims. (The two-week lagged value has a correlation of -0.73). The correlation coefficient is negative because as new job ads go up, UI claims go down. Analysts forecasting macroeconomic performance can gain an advantage by incorporating this data into their models.

The following graph shows the historical relationship between these two variables – the UI Claims axis has been inverted to facilitate visual interpretation.

Full-size chart available on Jobs. Trends. Insight.

What can we expect in next week's report? This week marks the start of the fairly aggressive seasonal decline in the number of online job ads. Without explicitly seasonally adjusting our data, it becomes hard to do the little mental exercise we usually present here (i.e. "does the number of online job ads correspond to the expected drop in NSA UI claims?). However, given the positive trend we've seen across the board in the labor market, we can expect another positive report next week. The 4-week moving average, at the very least, will resist any slight increase in next week's report.

Source: WANTED Analytics

© Copyright 2009 WANTED Technologies Corporation

The level of Hiring Demand has also remained unchanged on a seasonally unadjusted level, even as new UI claims dropped by 53,000. New online job ads fell by only 1,700.

The following graph, where the UI Claims axis has been inverted to facilitate the visual interpretation of the relationship, shows how the two variables have been moving in tandem over the past 4 years:

Full-size chart available on Jobs. Trends. Insight.

The correlation between Hiring Demand (new online job ads, weekly) and New Unemployment Insurance Claims has been -0.71 over the past four years. As new job ads appear online, more people can find work, and fewer must file unemployment insurance claims – hence the negative sign on the correlation coefficient. What's more, the correlation between the one-week lagged value of new online jobs and and new UI claims is -0.77, which supports the hypothesis that Hiring Demand is a leading indicator of published labor market data. Remember that a correlation coefficient is bounded between -1 and 1, where values of -1 or 1 imply a perfect relationship between two variables (you can usually only get a perfect correlation if you correlate a variable with itself, which isn't of much use…). So Hiring Demand closely tracks new UI Claims.

What's our forecast for next week's Unemployment Insurance Claims report?

Visibility into next week's report is clouded by the high level of seasonal fluctuations, which is usually why the 4-week moving average of the seasonally adjusted data is usually more closely followed than the week-over-week changes. Indeed, following our usual process for forecasting the direction of next week's UI claims report:

• Assume next week's report will have the same seasonally adjusted count ("Initial Claims – SA"): the week-over-week change is zero.
• Given next week's multiplicative seasonal correction factor ("Multiplicative SF");
• What would the level of unadjusted claims ("Initial Claims – NSA") have to be to keep us steady in SA terms?

We see that we're expecting an increase of 110,000 new UI claims – or a 22% jump – on a seasonally unadjusted basis, due to seasonal fluctuations alone.

<Computed values are marked as such> in the table below.

© Copyright 2009 WANTED Technologies Corporation

Seasonally adjusted counts are driven by seasonally unadjusted counts and a seasonal correction factor. On a seasonally unadjusted basis, new UI Claims rose 47,000. This was driven by a drop of 40,000 new online job ads two weeks ago. Last week, new online job ads rose by only 16,000.

The correlation between Hiring Demand (new online job ads, weekly) and New Unemployment Insurance Claims has been -0.71 over the past four years. As new job ads appear online, more people can find work, and fewer must file unemployment insurance claims – hence the negative sign on the correlation coefficient. What's more, the correlation between the one-week lagged value of new online jobs and and new UI claims is -0.77, which supports the hypothesis that Hiring Demand is a leading indicator of published labor market data. Remember that a correlation coefficient is bounded between -1 and 1, where values of -1 or 1 imply a perfect relationship between two variables (you can usually only get a perfect correlation if you correlate a variable with itself, which isn't of much use…). So Hiring Demand closely tracks new UI Claims.

The following graph, where the UI Claims axis has been inverted to facilitate the visual interpretation of the relationship, shows how the two variables have been moving in tandem over the past 4 years:

Full-size chart available on Jobs. Trends. Insight.

What can we expect for next week's report? Three factors are at play: new UI claims on a seasonally unadjusted basis, the seasonal correction factor (which produces the seasonally adjusted count), and the number of new online job ads.

The method for "estimating" the direction of next week's report has four steps:

- Assume next week's report will have the same seasonally adjusted count ("Initial Claims – SA"). The week-over-week change is zero.

- Given next week's multiplicative seasonal correction factor ("Multiplicative SF");

- What would the level of unadjusted claims ("Initial Claims – NSA") have to be to keep us steady in SA terms?

- Finally, does that concord with observed changes in the level of new online job ads?

<Implied calculations are marked like so> in the table below:

In order to stay "stable in SA terms", we'd need the level of new claims NSA to drop by 52,546 to 476,900. However, we saw an increase of only 13,000 new job ads this week. Unless there's a huge spike in online job ads next week to absorb the unemployed, we can expect New Unemployment Insurance Claims to rise on seasonally adjusted basis.

© Copyright 2009 WANTED Technologies Corporation

New UI Claims also fell on a seasonally unadjusted basis, falling 14,000 new claimants. This corresponds to a rise of 40,000 new online job ads the week before.

The relationship between new UI claims and online job ads is quite strong: the correlation between the two variables has been -0.70 over the past four years. As more job ads are posted online, more people can find work, and fewer must file unemployment insurance claims – hence the negative sign on the correlation coefficient. What’s more, the correlation increases to -0.76 when we examine the one week lagged value of online job ads, which supports the hypothesis that Hiring Demand leads published measures of labor market dynamics. The UI Claims axes has been inverted on the graph below to facilitate interpretation.

Full-size chart available on Jobs. Trends. Insight.

What can we expect from next week's UI claims report? We expect little to no change in next week's report. Two factors come into play: the number of online job and the seasonal correction parameter. We saw a drop of 44,000 new online job ads this week. This means that new UI claims will rise on a seasonally unadjusted basis. The multiplicative seasonal correction factor is moving from 1.07 to 0.95, so we're being accommodated there. In order to maintain the same levels on a seasonally adjusted basis, new UI claims need to rise to 539,000, a jump of 59,500 new claims. Given historical data, this week's drop in online job ads corresponds almost exactly to this expected rise in new UI claims (every new online jobs produce -1.36 new UI claims).

© Copyright 2009 WANTED Technologies Corporation

On a seasonally unadjusted basis, the number of new UI claims rose by 32,00 claimants, even as the number of online job ads rose slightly by 2,000 ads. However, this is probably due to the drop of 32,000 that we saw two weeks ago.

The relationship between new UI claims and the level of online job ads is quite strong: the correlation between the two variables has been -0.70 over the past four years. As more job ads are posted online, more people can find work, and fewer must file for unemployment insurance – hence the negative sign on the correlation coefficient. What’s more, the correlation increases to -0.76 when we examine the one week lagged value of online job ads, which supports the hypothesis that Hiring Demand leads published measures of labor market dynamics. The second lag stays strong at -0.72 as well. The UI Claims axes has been inverted on the graph below to facilitate interpretation.

Full-size chart available on Jobs. Trends. Insight.

What can we expect from next week's UI Claims report? The number of online job ads rose by 2,000 last week and again by 31,000 this week, settling to 371,000. The multiplicative seasonal correction factor is moving from 92.9 to 93.8, which is almost no change at all. The influx of new job ads should drive the seasonally unadjusted UI claims count downwards. With a fixed seasonal correction factor, new Unemployment Insurance Claims should fall again.

© Copyright 2009 WANTED Technologies Corporation

However, most of this is due to the seasonal fluctuation parameter. The variance around the "true number of new UI Claims on a seasonally adjusted basis" can be greater when a data series goes through high levels of seasonal fluctuations. That's what happened this week. The multiplicative seasonal adjustment factor went from 1.02 to 1.15, which means that we expected a drop in new UI claims of about 13% due to seasonal fluctuations alone. If new UI claims don't fall enough on a seasonally unadjusted basis, they grow on a seasonally adjusted basis. We had seen a drop in new job ads last week, which meant we were expected more UI claims. This drop in the number of job ads, in addition to the severity of the seasonal fluctuation, led to this week's "negative" report.

So, for example, CNBC interpreted the news as "indicating the labor market remained fragile despite some signs of an economic recovery." We did predict a rise in new UI Claims, but that was mostly driven by a very severe change in the seasonal adjustment factor. Had the seasonal correction factor been 1.13 instead of 1.15, seasonally adjusted UI claims would have remained unchanged.

Remember that the relationship between new UI claims and the level of online job ads is quite strong: the correlation between the two has been -0.70 over the past four years. What’s more, the correlation increases to -0.76 when we examine the one week lagged value of online job ads, which supports the hypothesis that Hiring Demand leads published measures of labor market dynamics. As more job ads are posted online, more people can find work, and fewer must file for unemployment insurance – hence the negative sign on the correlation coefficient.

The UI Claims axes has been inverted on the graph below to facilitate interpretation.

Full-size chart available on Jobs. Trends. Insight.

Hiring Demand fell again this week, dropping from 336,000 new online job ads to 330,000.  This drop of only 6,000 new job ads is much less severe than the drop of 36,000 we saw the week before. What's more, the seasonal adjustment parameter is swinging back the other way, from 1.15 to 1.075. Combining these two factors, we predict improvements in next week's UI claims report, putting us back "on trend". (Hiring Demand dropped, but the seasonal coefficient on UI claims overcomes this).

© Copyright 2009 WANTED Technologies Corporation

The relationship between new UI claims and the level of online job ads is quite strong: the correlation between the two has been -0.70 over the past four years. What’s more, the correlation increases to -0.76 when we examine the one week lagged value of online job ads, which supports the hypothesis that Hiring Demand leads published measures of labor market dynamics. As more job ads are posted online, more people can find work, and fewer must file for unemployment insurance – hence the negative sign on the correlation coefficient.

The UI Claims axes has been inverted on the graph below to facilitate interpretation.

Full-size chart available on Jobs. Trends. Insight.

The number of new online job ads fell by 36,000 last week, settling at 332,000. The seasonal adjustment factor on new UI Claims is moving from 1.02 to 1.15, which means that we’re expecting fewer new UI Claims on a seasonally unadjusted basis. These two factors – a drop in new job ads but a rise in the multiplicative seasonal factor – combine such that we predict a worsening of new UI Claims in next week's report. (We would need the multiplicative seasonal adjustment to be smaller to accommodate more UI claims when new job ads dry up).

© Copyright 2009 WANTED Technologies Corporation

This week's Unemployment Claims Report from the Department of Labor confirms this improving trend. After an increase in new UI claims from 534,000 to 554,000 in last week's report, new UI claims fell by 33,000 to 521,000 in this week's report. This substantial improvement puts new UI Claims "back on trend".

There were 19,000 fewer new online job ads last week, which made new UI claims rise by 3,757 on a seasonally unadjusted basis. Fewer online job ads means that the unemployed cannot find work, increasing Unemployment Insurance Claims. The correlation between new online job ads and new UI claims is -0.69 over the past 4 years (both are seasonally unadjusted). The graph below, where the UI Claims axis has been inverted to facilitate interpretation, shows this relationship.

Full-size chart available on Jobs. Trends. Insight.

Next week's report has quite a bit of forecast variance around it: the multiplicative seasonal correction factor is moving from 0.86 to 0.98. In other words, for next week's report, we are expecting an increase of 14% in unadjusted new UI claims because of seasonal fluctuations alone. Given recent robustness in Hiring Demand, we expect another positive report from the Department of Labor next week (new online job ads won't fall by so much that new UI claims will rise enough to increase the seasonally adjusted level).

© Copyright 2009 WANTED Technologies Corporation