There is no simple theory of geography accounting for the SPIN-TIGER disparities, but this paper rounds up the "usual suspects" geographers rely upon to explain mapping anomalies. Understanding them may make planning Census 2010 easier.
The financial logic applies to a state like Louisiana, but the facts do not, since the state is not a target of immigration, nor is it a center of urban growth. Rather, the state's census leakage is likely to come from suburban sprawl that has outpaced geographical documentation. This raises obvious equity issues since uncounted whites still affect the state's minority population. It also raises the central topic of this paper, which addresses problems and prospects in using cheaper and more rapidly deployed remote sensing data to document and anticipate shifts in population distribution.
The cities of New
Orleans, Baton Rouge, and Slidell are located near the base of this plain, and
are close to a major estuary of the Gulf of Mexico called Lake Pontchartrain,
the ellipsoidal water body prominent in this part of the state. 
What this means,
as a practical matter, is that the state's major population concentration is
subject to spring river flooding and fall hurricane flooding. Long-term
residents of south Louisiana use the term "the big one" in a sense similar to
southern Californians' allusion to some cataclysmic seismic event.
In Louisiana terms the "big one" is a catastrophic hurricane that would
breach the levees in New Orleans on both the Lake and River facing portions of
the city, and then march on towards Baton Rouge. One can visualize the effect in
the next map, which approximates sea level without the benefit of levees and
pumping systems, while it also greatly exaggerates landforms further north. 
It
is easy to discern the meandering natural levee of the Mississippi, but all that
appears of New Orleans is a thin crescent of swamp on the south shore of Lake
Pontchartrain. In effect, this map models most of the residential areas of this
region as resting five to fifteen feet under water following a Category IV or V
hurricane. These are rare storms, but save for evacuation, they outstrip
foreseeable defensive measures.
The resulting population density pattern defies conventional analysis (Reid, 1997),
however, because of the complications added by racial considerations. The Parish
of Orleans, coterminous with the city, is estimated (2000) to be about 65% to
70% black, while Jefferson Parish, to the west, is, possibly, 20% black. Since
Jefferson Parish grew as a dormitory suburb from the 1950s on, it is tempting to
attribute its development to white flight. But the argument is made more
complicated by the fact that since the 1980s the most dramatic growth in the
region has taken place far to the north in St. Tammany Parish, as the following
map illustrates. 
This area has made the urbanologists' term "sprawl" clearly pejorative, even though the term's connotation is debatable on a nationwide plane (Meyers and Kitsuse, 1999, Wells Fargo Bank, 1996).
This parish is too far away from New Orleans (50 miles) to fall into the "white flight" category. It is best characterized as a "squirearchy", more driven by class considerations than by racial issues. Yet New Orleans developers and politicians, less loudly, denounce St. Tammany's rapid expansion because it saps both a market and a tax base. There is plenty of room in New Orleans for housing. Depopulation of the city has left it with about thirty thousand empty, and often decrepit, dwelling units. Developers are eager to restore and fill these potential homes. But their eagerness is not matched by City officials, who have to wonder about the political impact of a substantial middle-class migration back to the urban core.
At the same time St. Tammany Parish residents denounce sprawl too, and use minimum lot sizes and various impact fees to keep entry-level housing out of much of the parish. This has the odd effect of keeping elected officials on both sides of Lake Pontchartrain at peace. St. Tammany keeps its revenues, while New Orleans is immune from the country squire values of exurbanites. But there is no end in sight for frustrated developers who cannot build "affordable" housing" in St. Tammany, nor easily market gentrified townhouses in the urban core.
The same pattern is found at the other end of the Pontchartrain Basin. Baton Rouge is another urban core that has spread laterally. In its case Livingston Parish, to the east, has become a parallel to the Jefferson Parish dormitory community. Farther to the east lies Tangipahoa Parish, which like St. Tammany, has developed a tendency to define sprawl as the next family planning to move in. In the case of both sets of urban cores and satellite parishes, there are crosscutting problems and paradoxes associated with development and population dispersion. Politicians put up with sprawl because they can trade lost revenues for homogeneous, hence predictable, electorates. Developers are frustrated because neither market nor regulatory mechanisms favor profitable housing models based on high density and tax relief in the form of federally financed housing salvage programs.
Making the Census effective through, so to speak, the front door with GIS and companion planning tools has not been a Louisiana priority for most of the 1990s. There is nothing unique or peculiar about this pattern. It merely reflects the uneven and competing technical priorities inevitable in local governments.
The background of this should be clear from the problems of physical geography referred to above. Localities south of Lake Pontchartrain are preoccupied with urban drainage problems, while north of the Lake streams and forests were a priority. GIS expertise in the area was largely restricted to questions of either hydrology or gerrymandering. It was only in the late 1990s that a regional GIS community became self-conscious, due in part to Census 2000 and the possibility of losing yet another congressional seat.
But technical issues played an important role in increasing geographical awareness. Inexpensive computing, the transition from tape to CD data storage, TIGER, USGS digital files, internet servers, and much improved software were among the most obvious elements in a geographical revolution.
One obvious and persisting problem is updating maps. At mid-decade LORAN and LANDSAT were being eclipsed by newer tools, but SPOT and positionally degraded GPS were not suitable for urban planning roles at the block or parcel level. The standard of resolution we thought necessary for census work was about six feet, or two meters - enough to resolve an SUV in a driveway. Vehicles in driveways make a presumptive case for the existence of human habitation. At the end of the decade the failure of IKONOS to deploy was a major disappointment. In Louisiana there were hopes, as early as 1998, to exploit its potential to audit new versions of the TIGER files prior to the 2000 census.
The fallback position was to turn to product from SPIN-2, a Russian panchromatic instrument designed as an intelligence-gathering tool. Its resolution is a nominal 1.57 meters, and there was leaf-off and cloud-free 1998 imagery of the Pontchartrain Basin in Russian archives. So, there were adequate resolution data available for the study area from a suitable epoch.
The main technique for exploiting the data is simple in concept: overlay the
imagery with TIGER line files as in this rendition of an area near the town of
Mandeville, Louisiana. The idea is to look for neighborhoods the KGB (or its
successor) knows about that do not show up in TIGER. 
This rendition of the image does not convey fairly what can be seen at higher resolutions. But it should be apparent that TIGER (1997) misses some inhabited areas.
There are some complications and opportunities along this route. In one sense we were just lucky to have suitable Russian data. SPIN satellites have a short orbital life and it is not always the case that a Louisiana March will be cloud-free. We never determined the spectral window SPIN uses, but it, sensibly, is more sensitive to developed land than to swamps and the undifferentiated coniferous forests of Louisiana. Roadways and excavations are fairly easy to detect.
Georeferencing the imagery presented some options. Planning GIS practitioners in the Pontchartrain Basin have informally settled on Louisiana State Plane NAD83 in feet as a common coordinate system. But the State uses UTM-15 for its crucial oil-spill response map, even though the most critical area falls in UTM zone 16. In this case we adopted a political standard. Most local officials are familiar with paper USGS 7.5' maps and they recognize DRG versions of these maps. This meant referencing most of the data to UTM-15 and NAD-27. Some images were referenced to UTM-16 and NAD-27. The principle here is that the first rule of public cartography is to make the map familiar to the audience. In this case it was important to bend Russian images to U.S. maps.
SPIN images do not follow LANDSAT rules, in the sense that there is no apparent pattern to the image footprint. This is, at least in part, due to the lower orbit SPIN employs. Normally this would not be of interest, but over an area this large it means that mosaicking becomes crucially important, since we are trying to yoke a Russian spy satellite to U.S. Census wagon. We found that IMAGINE's mosaicking tools, and its capability to define areas of interest (AOI) in terms of ArcInfo polygon coverages (such as municipalities) to be very helpful.
Another piece of luck we had lay in SPIN's degraded resolution by comparison to IKONOS. No matter how fast one's computers are, it is still going to take more time digest 1 meter pixels than 1.57 meter pixels. For this kind of task, auditing vector files at the scale of a road, SPIN's resolution is not an apparent limitation that outweighs its reduced storage and processing requirements.
The GRID rendition, a bit Mondrianesque, yields isolated (from TIGER) polygons. These are, presumably, unpopulated islands. We are attempting to model their characteristics in two ways. The first is in terms of sheer size as a function of distance decay from some point. Arc's SPATIAL ORDER tool for constructing Peano curves of the polygon centroids may helpful here. The second is in terms of settlement intrusion on these polygons as portrayed by early editions of TIGER, Russian imagery, and by later editions of TIGER. This means working out an agenda over a fairly extensive area (in terms of 1.57 meter pixels) using vector GIS, remote-sensing tools, spatial analysis tools, and time-series analysis. Only the goal is clear. If it is possible to infer rules for urban growth missed by the Census it will be possible to improve its aim.
References
Meyers, Dowell and Alicia Kitsuse. 1999. "The Debate Over Future Density of Development: An Interpretive Review." Lincoln Institute of Land Policy Working Paper.
Ewing, Reid. 1997. "Is Los Angles-Style Sprawl Desirable?" Journal of the American Planning Association. Vol. 63: 95-105.
Wells Fargo Bank. 1996. Preserving the American Dream: The Truth About Suburban Communities and Housing Choice. San Francisco: Wells Fargo Bank.