Part of its ongoing “Safety Net” series, this morning’s New York Times features Once Stigmatized, Food Stamps Find Acceptance, a story about the increasing demand for food stamps. In an earlier article, the NYT published one of its peerless national maps illustrating food stamp usage across the country. (“The number of food stamp recipients has climbed by about 10 million over the past two years, resulting in a program that now feeds 1 in 8 Americans and nearly 1 in 4 children.”)

This historic escalation is reflected in the web stats for the California Food Stamp Guide. A year ago, we drew attention to how the usage patterns at the FSG were a sign of the economic times. If so, then it has only gotten worse. Since then, as illustrated below, current Google Analytics for the FSG show monthly visitor sessions have increased to nearly 10,700 (a 52% increase), with over 101,000 page views (a 27% increase) and a bounce rate of only 0.25%. Annualized, this increase in the monthly usage rates of the FSG comes to more than 128,000 visitor sessions and 1,216,000 page views.

Other posts of possible interest...

• Web traffic as a sign of economic times
• The best GIS map. Ever.
• "A List Apart" search / usability trifecta
• Dealin’ with the down and out, at the iPanic app store

• Testing Search for Relevancy and Precision
• Internal Site Search Analysis: Simple, Effective, Life Altering! (by Avinash Kaushik of Occam’s Razor fame)
• Beyond Goals: Site Search Analytics from the Bottom Up (by Lou Rosenfeld, co-author with Peter Morville of Information Architecture for the World Wide Web)

Other posts of possible interest...

• Google Analytics "Conversion University"
• A painless way to learn basics about Google Analytics
• Web traffic as a sign of economic times
• Diggin’ data with Google Analytics

It’s like this: Two months ago, in a different context, we observed that traffic at the California Food Stamp Guide seemed to have reached an annualized plateau of about 54,000 visitor sessions and 570,000 page views. Pretty much what we were expecting based on our experience with traffic at the earlier version it replaced.

That was then. This is now: Traffic at that site has escalated dramatically in the last two months, to 7.000+ visitor sessions with 79,000 page views, per month, which annualizes to 84,000+ visitor sessions and now approaching a million page views, at 948,000+ page views per year:

Translation: In the last two months, there has been a 55% jump in visitor sessions and 66% jump in page views. And the bounce rate now falls regularly below 1%, hovering around 0.65% most days. More people are finding it and sticking to the site.

What’s going on? During this more recent 30-day period, illustrated above, 86% of the visitor sessions were driven by 3,841 different keyword searches (the proverbial long tail). People have always been looking for help getting food stamps. But a whole lot more are now out there looking.

Other posts of possible interest...

• Revisiting web stats for the California Food Stamp Guide
• "A List Apart" search / usability trifecta
• Diggin’ data with Google Analytics
• Outsourcing your web stats

Other posts of possible interest...

• GoogBurner … just so you know
• Props to FeedBurner for explaining stats
• A few more notes about the Google “full-account” transition
• Our final 15 minutes of Google fame

Here’s what happened at the recent training:

It was my responsibility during the training to explain why LSNC had changed how it records race and ethnicity. The brief history lesson for the trainees was that the Pika defaults, like our prior case management system, were configured to track client race data using four long-used specific race categories (white, black, native american, asian/pacific islander), plus two additional generic catch-alls (multi-racial and other). I explained that, while these categories are still commonly used, even if incorrectly by some federal funding sources, they are not properly calibrated to conform to current Census race categories that have been in place since the 2000 Census. Among other things, the Census Bureau dropped “multi-racial” as a survey selected category years ago, instead opting to allow the individual to self-identify with one or more of the six Census-defined racial categories. I also explained that the Census Bureau categories include an all encompassing demographic view of the world such that all persons are either Hispanic/Latino or non-Hispanic/non-Latino. It took a few moments for the gathered crowd to wrap its collective head around these concepts, but at this initial level of abstraction no one voiced any concerns or confusion. They understood the basic but familiar racial categories, even if the particular labels had changed slightly, and the dichotomy of Hispanic/non-Hispanic made sense.

So far, so good. Then I walked them through how LSNC had redesigned the fields for tracking this information, with the initial question to the client being broader, i.e., whether he or she identifies as Hispanic or Non-Hispanic, followed by a more specific racial identity question, relying on the six basic Census categories:

OK, still so far, so good. But then we all danced our way, at times somewhat gingerly, through LSNC’s Pika CMS Race and Ethnicity Quiz-o-rama! Admittedly, several of the quiz questions are intentionally tricky, to challenge folks about their perceptions. And there was the rub. Without getting too much into particulars, several staffers were startled to confront what they assumed incorrectly (in terms of Census definitions) to be the racial or ethnic identity of some of their clients. And some were visibly uncomfortable with how some clients may self-identify. For example, for some it was news that some Latinos identify as “white.” Or that clients from two bordering countries like Iraq and Iran may or may not self-identify as “white” or “asian” or both, or even something else (“other”). And then there were the inevitable funding-driven concerns, along these lines: “Wait, you mean if a client from Morocco identifies herself as being Hispanic and/or white, we can’t record the person as African-American? But our funding source really wants to see that we’re serving Blacks/African Americans.” I answered, “Not if the client does not self-identify that way.” Not a popular answer, with some. And an instructive example of how organizations are prompted to perceive clients within categories driven, if not explicitly defined, by a funding source.

The problems with tracking racial data are compounded by the failure of many funding sources to catch-up with the prevailing Census race and ethnicity categories. I use here an example familiar to LSC-funded programs, but not to single out LSC. It does what a lot of funding sources currently do, including the California IOLTA program. But it is a good example of how race and ethnicity data can get muddled by confusing Hispanic ethnicity with the basic racial categories, however labeled.

Take a look at this example of the LSC Form G-4(a). Every LSC-funded field program has to fill this form out annually:

The problem with this form is that it demands totals of data categories that are of a different type, i.e., it lumps “Hispanic” ethnicity in with all the other categories which are racial, as the terms are understood by the Census Bureau. Tallied in this fashion, one cannot possibly get a statistically accurate total. Why? Because even if the first two categories (white and black) excluded those who identify as Hispanic, the three other racial groups (native american, asian or pacific islander and other) do not. And those groups are not Hispanic because…? Who says so? And if you say they may be, then if you total the “Hispanic origin” data with these three same groups, you are double-counting, no?

How should this been done properly? Take a look at HUD’s Race and Ethnic Data Collection Form (HUD Form 27061):

The difference here is that the form asks for numbers that correspond to the basic Census racial categories (or understandable but consistent variations of them) in the second column, but then parses out each of those rows with a third column to identify how many of each presents Hispanic/Latino ethnicity. Easier to understand. Consistent with the Census definitions and demographic data. And statistically accurate.

It is precisely that type of argument I am now making to varied funding sources, including most of our seniors funding grantors, to resolve inconsistencies between us and them and among themselves about the race and ethnicity data they demand.

Other posts of possible interest...

• GIS map of the day: Impact of USA Diversity
• Lies, damned lies and the New York Times
• Revisiting web stats for the California Food Stamp Guide
• "A List Apart" search / usability trifecta

Whatever happened to 24/7, people?

Other posts of possible interest...

• First take at responsive web design
• What our tech training survey tells us
• Trouble in River City
• The Race Equity Project rebuild debuts