Sorry for this long post, but I hope it'll clear things up a bit
Nielsen says the decline in household numbers is due to three things
1) Digital transition, many people who can't get signals simply went without. I am one of these people who lost all TV signals, despite living on 3 miles NW of Willis (Sears) Tower. In addition was the rebound effect. People bought digital TVs and coverters and this caused a "spike" in their numbers. This downward in numbers is the adjustment as people throw out their old TV and discontinue converters. Indeed on another board, one poster asked why he was sent a Nielsen survey when he got rid of his TV. Nielsen told him to leave it blanks and send it in.
2) Economy, people are losing TVs due to age and simply not replacing them. They are going without, watching at gyms, or simply having friends record TV shows and watching them on iPods or computers later on.
3) Torrents, there are three known sites which you can download shows within 15 minutes of their completion. And there are hundreds more where you can get the shows within a day
4) Multiple platforms. Nielsen says it plans to expand into measuring all ways people watch TV shows including torrents, iPods, and such but it hasn't. It admits it hasn't perfected measuring time-shifting measurements but plans to.
The 2012 TV penetration for U.S. households was estimated based on data collected during the recruitment of homes for Nielsenís People Meter panel. Nielsenís definition of a television household requires at least one TV capable of tuning to at least one channel. For example, television sets not updated for digitally transmitted content would not count under Nielsenís definition.
Under this defintion, despite having a TV and a computer with a TV tuner I would not be a household, because I can't tune any channels in my location. When I want to watch a show, I go to the gym or move to the library with my laptop.
Remember sharing also effects it. Last year Nielsen says a household has about 2.5 people. If more people move into a household due to the economy, through roommate shares or moving back with mom and dad it INCREASES the population but it doesn't increase the household. The household size just gets bigger.
For the confusion of Nielsen markets versus the census:
Nielsen defines its own markets. This is NIELSEN not the FCC and this confuses people. Nielsen controls the definition. However the FCC uses Nielsen markets (DMAs) for its purposes. Why? Because it's easier than making up their own definitions. Since Nielsen is really the only game in town, it makes sense in a way to use them. Nielsen is NOT a scientific survey. In their defense, they never once, claimed their numbers were scientific. This comes up a lot in arguments, but for all its shortcomings, Nielsen never said their polls were scientific.
Nielsen rankings differ because they rank them. Tampa is bigger than Miami, simply because it's geographically larger. But by ever definition of US Census, Miami is the bigger of the two.
For the census, the US Government defines 4 basic city groupings.
The first is a "Micropolitan Area" which is defined of a central city of 10,000 - 49,999 and surroundings. This is measured by
social and economic integration with the core as measured by commuting ties.
The Metropolitan Area, which is defined as a central city of at least 50,000 (or two or three cities together which add up to 50,000). This is measured also by social and economic integration with the core as measured by commuting ties. And it always includes at least ONE county, the county they city (cities are in). This can have odd results, especially out west where counties are large. The Reno Metro area includes Washoe County, which extends all the way up to the Idaho state line. Obviously the few people up by the Idaho state line aren't as urban as the people in Sparks, NV, next to Reno. But because the Metro area has to include a county in it's entirety, it is just how it is.
Then there is a thing called a Combined Statistical Areas. This is two or more adjacent Metro areas (or a Metro area and a Micropolitan Area) that are joined economically as measured by commuting ties
This produces conflicts as well. For instance, Kenosha Wisconsin, which has high commuter pattern into Illinois and Chicago. So Kenosha is its OWN metro area, but is also part of the Chicago, Combined Statistical Area, because of the commuting patterns.
Of course Kenosha gets Milwaukee TV, not Chicago TV. but because of the commuting and social integration of Kenosha into Illinois cities and Chicago is greater than that of Milwaukee for US Census purposes it's linked to Chicago.
The last type of US Census are is Urban Area. This is defined as a core city (or group of cities) with 50,000 people. Instead of counties they uses census tracks that are densely settled at least 2,500 people, at least 1,500 of which reside outside institutional group quarters.
The Urban area is sort of new and was developed to answer complaints of Eastern cities that are highly settled. Indeed according to the Urban area, parts of Houston, Phoenix and San Antonio, Jacksonville (and a lot of other cities, some East but most West and Southern) are actually rural not urban. So they are incorporated into a city but are rural in nature not urban.
This is what confuses people. Once you get below the big three. New York is always first, LA is always second and Chicago is always third. You start running into problems. Washington-Baltimore can be fourth by combining them. San Francisco and San Jose climb right up there. But by simple metro area definitions those examples are split and rank much lower.
Lists also confuse, because not all cities have equal components. For instance, Atlanta using a combined statistical area (CSA) is bigger than Miami, because Miami doesn't have a combine statistical area. It's hemmed in by the Everglades so it stands on it's own and doesn't appear on a list of CSAs.
By other rankings, Miami IS bigger than Atlanta.
So if this causes confusion, just remember with so many ways to define a city and it's surroundings, you can have completely different answer and all of them right, depending on which list and which definition you're using