Floyd Norris has been arguing that the Labor Department figures tend to underestimate job growth in the early part of a recovery (and underestimate job loss in the early stages of s recession.) Here's Norris's column from a month ago:

The unemployment rate declined four-tenths of a percentage point in one month. There had not been a monthly decline that large in many years, but economists were unimpressed. After all, the decline was caused in no small part by a surprising reduction in the labor force, which could be an indication that more workers were discouraged and no longer looking. That would hardly be an encouraging development.
Anyway, it was said, the unemployment rate is based on a survey of only 60,000 households, some of whom cannot be reached in any given month. It can be volatile, so you should not pay much attention to it. The president took heart from the figures, but critics said there was no real improvement.
The above describes what happened a month ago, when the unemployment rate for December came in at 9.4 percent. It was the first time in more than 11 years that the rate had declined that much in a single month, but the headline in The Chicago Tribune read, “U.S. jobs picture gets darker; Unemployment rate dips, but only because workers apparently are giving up.”
As it happened, those paragraphs also describe the situation 28 years ago. In January 1983, with President Ronald Reagan reeling from his large setback in the midterm elections the previous November, the unemployment rate fell to 10.4 percent from 10.8 percent. It was the first such decline in five years, but few thought it significant. 

David Leonhardt applies this analysis to today's job numbers:

The other main survey the department does is of households, rather than businesses. It has a smaller sample size and so usually is not as accurate as the survey of businesses. At turning points, though, the household survey can detect changes before the business survey does. As Floyd Norris has explained, the numbers from the business survey may well end up being revised later on — as more accurate data on start-ups flows in — and looking more like the sunnier household survey.
Today’s report clearly raises the odds that we’re in the middle of such a period now.
Average job growth over the last three months has been only 102,000, according to the survey of businesses. But it’s been 221,000 — significantly more than population growth — according to the survey of households. The unemployment rate, which is calculated from the household survey, has dropped sharply in the past few months in large part because of these job gains.
It’s particularly noteworthy that the Labor Department today revised upward the business survey’s estimated job growth for both December and January. That may mean more revisions are coming.

Let me add another point. The Norris-Leonhardt hypothesis is also the best explanation for the pronounced rise in President Obama's approval ratings:

Of course, political events -- like the elections giving Obama a foil, or the spate of lame-duck accomplishments in Deecember -- probably played a role as well. But I'm guessing the polls reflect an improving labor market.