Here's a sobering statistic: All nine recessions between 1948 and 1990 saw employment return to pre-recession levels within 31 months. Today, 42 months after our recession began, we've only regained about a fifth of lost jobs -- and population growth has made those meager gains irrelevant. The average duration of unemployment is now three times longer than the 1960-2007 average -- 40 weeks vs. 13 weeks. This is terrible stuff.

How we got here is well-known: bubble, financial crisis, recession, deleveraging. It's an ugly combination.

Where it happened isn't discussed as much, but the numbers tell an interesting story.

Private employers make up about 84% of all jobs, with government employing the remaining 16%. But since the recession began, the composition of job losses hasn't followed that divide evenly:

Sources: Bureau of Labor Statistics and author's calculations. Data excludes 2010 Census hiring and layoffs.

Since the recession hit in late 2007, private employers have shed 4.2 million jobs, while government payrolls have dropped by 350,000.

Government payrolls typically blunt the bleeding of the private sector during a recession. Early on during this past recession, that was the case. But then things turned. Since January 2010, private employers have added 2 million jobs, while government employment has shrunk by nearly half a million jobs. The opposite was true after the 2001 recession: Private employers shed 884,000 jobs from 2001-2003, while governments added 174,000.

Most of the recent drop in government jobs has been at the state and local level, which have been hammered by plunging tax receipts. State jobs have fallen by more than 300,000 since early 2010; local governments have fallen by roughly the same amount. Contrary to popular rhetoric, the long-term trend in federal workers has been flat to down. There are 15% fewer federal employees today than there were in 1990, and the same number as there were in 2000. Much of the decline can be traced to the Postal Service, which shed 150,000 jobs since 2007 alone, or one-fifth of its total.

Within the private sector, the overwhelming majority of job cuts fell on three sectors: manufacturing, construction, and retail. Those three plus financial services account for 80% of all private jobs lost:

Sources: Bureau of Labor Statistics and author's calculations.

Keep in mind, construction employment is down, but it's down from a time when the industry was grossly overgrown and fueled by a bubble. Nearly a quarter of all new jobs created from 2003 to 2006 were housing-related. Most of these jobs never would have been created without a bubble, and so the culling over the past three years has been more a reversion to normalcy than a slash-and-burn. The same is true for financial services. Legions of real estate agents and mortgage bankers have lost their jobs over the past three years, but the sad truth is many of those jobs never should have existed in the first place.

Then there's manufacturing. Manufacturing jobs have been disappearing for decades, declining 27 of the past 41 years. The last three years were simply an acceleration of a long-term trend -- albeit a very sharp acceleration. As I've written before, much of the decline is due to increased productivity, or technology replacing bodies. Most of the rest is due to offshoring. At any rate, the worst of the manufacturing fallout could well be over. Productivity growth looks topped out, increased shipping and labor costs have diminished Asia's competitive advantage, and -- at least anecdotally -- both GM (NYSE: GM) and Ford (NYSE: F) have been recalling laid-off workers.

Part of the problem in manufacturing is that the jobs that are available require skills most job searchers simply don't have. Incredibly, there are more manufacturing job openings today than there were in 2003. Some companies just can't fill those openings. As The New York Times wrote last year:

[S]upervisors at Ben Venue Laboratories, a contract drug maker for pharmaceutical companies, have reviewed 3,600 job applications this year and found only 47 people to hire. ... All candidates at Ben Venue must pass a basic skills test showing they can read and understand math at a ninth-grade level. A significant portion of recent applicants failed, and the company has been disappointed by the quality of graduates from local training programs. It is now struggling to fill 100 positions.

Such stories might be more widespread than you think. Former President Bill Clinton noted last year that for the first time in his lifetime, job openings are going up at twice the rate of job hires. And it's not just manufacturing. In a recent interview, Dow Chemical (NYSE: DOW) CEO Andrew Liveris said he thinks more than 1 million science and technology jobs will open up this year, but only 200,000 new graduates will have the skills necessary to apply.

Where are jobs growing? In two words: health care. The other end of surging medical costs is that there's surging amounts of money for medical employment. More than 1 million health-care jobs have been added since 2007 -- the only major industry to grow over that period. Plenty think this trend isn't anywhere close to being over. The Bureau of Labor Statistics projects health care jobs will grow by a quarter between 2008 and 2018, adding 4 million new jobs and "about 26 percent of all new jobs created in the U.S. economy."

Perhaps the most difficult figure to come to terms with is how long it could be before jobs return to normal. Barring an unforeseen miracle, most respected economists don't see employment hitting pre-recession levels until 2014 at the earliest. At current growth rates, it would be far longer than that. In an interview last week, Berkshire Hathaway (NYSE: BRK-B) Chairman Warren Buffett offered a little optimism, saying, "We will come back big time on employment when residential construction comes back. ... You will be surprised, in my view, how fast employment changes when that happens."

That, alas, could be years away.

Check back every Tuesday and Friday for Morgan Housel's columns on finance and economics.