In the midst of all this economic strife, the need for state and local governments to cut waste and spending is urgent. Why, then, is the LA County Board of Supervisors fighting to renew a contract with an underperforming and overpriced firm?

Last year, the Los Angeles County Department of Public Social Services (DPSS) procured for vendor services to operate the county’s GAIN case management services – a program that helps welfare recipients find work. Two bids came in from the incumbent company (MAXIMUS, Inc.) and Policy Studies Inc. (PSI). The two companies were scored by a neutral third party, and PSI beat Maximus solidly in several categories, including performance and price. Maximus protested the scoring, but the findings were upheld on 3 levels and PSI was recommended by DPSS to receive the contract.

The Board of Supervisors disagreed. They rejected the recommendation with 3 votes. They claimed the process of consensus scoring somehow concealed bias from the DPSS, though no specific evidence of this bias was ever presented. Furthermore, this scoring process was documented as a valid process which had been used for years prior to 2008, and the same process whereby the incumbent Maximus had been recommended and awarded. The BOS then directed the DPSS to extend Maximus’ contract for 6 months while they reissue the RFP and devise a new scoring method.

The BOS also expressed some superficial concern that the cost of the contract may exceed county requirements (see County Prop A). Although language could’ve easily been built into the contract to ensure cost neutrality/savings, the BOS rejected that argument and asked DPSS to review their contract monitoring costs for possible reductions and eliminate or reduce pay for performance provisions that could drive up the overall contract cost should the vendor outperform expectations.

The reissue of this RFP makes no fiscal sense whatsoever, particularly given the dire state of California’s economy. What’s more, the state faces federal penalties to the tune of approximately $185 million if they do not meet a preexisting federal threshold. Why is the BOS insisting on spending MORE of our tax dollars in an effort to maintain their business relationship with Maximus – a company whose performance was scored lower and whose contract was priced higher than PSI? (The county has estimated the cost to reissue the RFP to be $250,000). PSI’s contract would save the county over one million dollars annually. What’s going on here?

An LA Times article from last year exposed just how entangled Maximus is with the BOS. In the first half of 2008, Maximus spent over $124,000 on two lobbying firms, more than doubling what they spent on marketing in year before. Perhaps even more troubling, Maximus donated $1,000 (the maximum allowed) to the campaigns to re-elect supervisors Don Knabe and Michael D. Antonovich. They even gave $1,000 to two members whose terms had 2 years left to run.

In these lean times, something else must be motivating the board’s decisions, because it’s certainly not the bottom line.

i don’t want to overstate my statistics credentials. but i caution against drawing any conclusions there exists even a slight correlation between unemployment and state tax rates based on this data. a correlation coefficient of 0.128 is really really low. a positive r-value doesn’t, in and of itself, mean a positive correlation. from what i understand, the magnitude of the r-value is what’s important. hence the rule of thumb i mentioned in my previous post. if the r-value was above .2, we would have data supporting an inference that unemployment and state taxes were slightly correlated. but our r-values don’t even support an inference of low correlation. then again, this is just a rule of thumb.

you make a good point that i had not considered–the lack of correlation between unemployment rates and state tax rates is limited to the current recession. fair enough. in the past, these variables might have been correlated, so tax cuts made sense. but from a policy perspective, i think our data is adequate. the current recession, and its underlying data, is what is important. if data on the current recession does not indicate correlation between taxes and unemployment, then a policy of tax cuts just doesn’t make sense to me. it’s a policy proposal driven by ideology and not by data.

thanks for answering my questions, pat.

i’m not sure i understand how you’re using them, since we’re not engaging in a random sampling (i.e., our universe of data is finite–only 50 states, plus d.c.). in other words, we aren’t trying to predict how accurately our regression represents the entire population, since all 51 data points are included in our model.

You’re right, this is not a sample, but rather an analysis of the whole population in consideration here for this recession. So my understanding is that the r = 0.128 does indicate that there is some positive correlation then. Granted, it is quite minimal, and that only 1.6% of the variation in unemployment can be explained by the tax rates.

The other thing is that this is only an analysis of the current recession. This is not necessarily a predictor of what happens during recessions in general, or during any period for that matter, because this data was limited to now. If I had more time, I would have done an analysis of some of the previous recessions and see if there is similar (non-)correlation.

I have a colleague who is an economist who normally believes that during a recession, the worst thing to do is to increase spending. However, he sees that this recession is different than what we’ve seen in the past, and thinks that the stimulus might be the right thing to do this time. Now, he didn’t address tax rates, but I’m wondering if he is right that this recession is “different,” maybe correlations we’ve expected in the past do not apply for this recession. But again, that’s a big maybe.

i ran a similar regression analysis, so your results weren’t surprising. i kept pushing other people to give me data re: unemployment rates and state taxes, partially because i knew the outcome and also as a demonstration of good faith interest in debate. instead, i got a lot of supply-side rhetoric, and nothing data-driven. since correlation is a statistical measurement, data is essential, much to the chagrin of previous commenters.

i used similar data–may 09 state unemployment rates (wsj, 5/22/09), and 2009 state income tax rates. to simplify the model, i used only the highest marginal tax rate, though generally i found that the highest tax rates kicked in a relatively low levels of income. for the record, i recognize the limitations of this model, which you noted–this is not an accurate reflection of the total tax burden, which would also include sales taxes, estate taxes, property taxes, luxury taxes, etc. but absent a way to accurately model these tax burdens, i think state tax rates are an acceptable proxy.

my r-value was similar to yours, but for argument’s sake, i’ll accept your figures, i.e. r=0.128. i disagree with your conclusions, because i think this figure is pretty strong evidence of non-correlation, or alternatively, a level of correlation so negligible that it should not drive tax policy.

as you note, r-values fall with a range between -1 and 1. since, presumably, gpw assumes that state tax rates and unemployment rates are positively correlated, an r-value approaching 1 would prove his assertion. an r-value approaching 0 will indicate a lack of correlation between these two variables. at 0.128, the r-value is a whole lot closer to 0 than 1. such a low r-value is strong evidence of non-correlation.

a rule of thumb used to interpret r-values is a follows: an r-value of .8 or above indicates strong correlation; an r-value above .5 indicates moderate correlation; an r-value above .2 indicates low correlation. your r-value, at 0.128, doesn’t even cross the threshold for low correlation. again, i interpret this as an indication of the lack of correlation between state taxes and unemployment.

besides the correlation coefficient, another way to determine how accurately your line of regression “fits” the data is to look at the coefficient of determination, or r-squared-value. the r-squared-value determines the percent of variation between unemployment and state taxes that is explained by the regression model. note that 0<r-squared<1. if r-squared is 1, then 100% of the variation is explained by the regression model. if r-squared is .75, then 75% of the variation is explained by the regression model.

to get the r-squared value, all you do is square r. in your case, since r=0.128, then r-squared=0.016 (i rounded). this means, in percentage terms, that 1.6% of the variation in unemployment is caused by state taxes. only 1.6%!!!!! the remaining 98.4% of variation in unemployment rates is not explained by state tax rates. these figures fail to make even a minimal case supporting a conclusion that state taxes and unemployment are correlated.

more interesting, for me anyway, is the policy-bias that this data indicates. in spite of this data, people like gpw and other commenters will continue to bang the drum in favor of tax cuts. this reaffirms the perception that conservatives have one idea when it comes to economic policy–tax cuts, consequences or efficacy be damned.

also pat, a favor, if you don’t mind and feel so inclined–explain more your use of confidence intervals in your analysis. i’m not sure i understand how you’re using them, since we’re not engaging in a random sampling (i.e., our universe of data is finite–only 50 states, plus d.c.). in other words, we aren’t trying to predict how accurately our regression represents the entire population, since all 51 data points are included in our model.

This is what I did. I made two comparisons:

A) A correlation between the state unemployment rate for May 2009 and the effective tax rate on a single person earning $50,000.

B) A correlation between the state unemployment rate for May 2009 and the top marginal tax rate.

The top marginal tax rate kicks in anywhere between $0 and $1,000,000 (California), so I thought A) might have been a better comparison. Also, I did my best to really get the typical tax burden, by including the exemptions, tax credits, as well as county or local taxes that residents have to pay in addition to their state income taxes. In about three cases, you would need to be either a tax attorney and/or CPA to figure it out, but I did my best. I did not include sales taxes, however. In many cases, a state not having income taxes may have a higher sales tax rate. I did not include sales taxes, or even compare sales taxes to unemployment rates, because sales taxes can be very confusing. In many cases, states tax different things. In NJ, we’re famous for not taxing clothes and I think we still don’t tax toilet paper. In other states, like NC, the sales tax rate is very confusing and there are different rates for different items.

Anyway, for those who are not familiar with the correlation coefficient r, it quantifies how a data of paired items are related to each other. It can range from -1 (perfect negative correlation) to 1 (perfect positive correlation). In the middle, 0, represents no correlation.

With respect to A), r = 0.128 and for B), r = 0.118.

So both indicate some positive correlation between tax rates and unemployment. However, using 95% confidence, one cannot conclude that there is correlation between the two. The best we can do is get about 63% and 61% confidence, respectively, which is pretty low in the statistics world. In order to achieve 95% confidence for 51 paired date, we would need r = 0.277.

A couple of other points. Does this “prove” that there is no correlation between tax rates and unemployment? No, not at all. It just means we can’t prove that there is correlation between tax rates and unemployment. Big difference. Kind of like saying that a jury coming back with a verdict of not guilty (using the reasonable doubt threshold) does not mean the jury concluded the defendant is innocent.

Further, this also shows that you can’t use one example and try to come up with a conclusion. Michigan has the highest unemployment rate, but it’s tax burden is on the middle or low side. Nevada has no state income taxes, but has a relatively high unemployment rate. California has the highest marginal rate (10.3%), which kicks in at $1 million, and even it’s second highest marginal rate (9.3%) kicks in at $47,000. Despite that, it’s tax burden falls in the low to middle range for a single person earning $50,000. New Jersey, like California, has a high maximum rate (8.97% for over $500,000) and a fairly low burden for the $50,000 earner. It’s unemployment rate is slightly lower than the national average.

Also, to put this in terms of what we see often during campaign times. When you see polling information, you usually see a margin of error (like +/- 3.5%). This means that the polling company is either 90% or 95% sure that the poll represents the actual percentages within the margin of error. If the poll has one candidate leading another, but within the margin of error, it is usually referred to as a statistical dead heat. So that is one way to characterize the correlation between tax rates and unemployment rates.

If you remove all the data that crushes your arguement, then you won’t see it.

Actually, you’re just ignoring everything that doesn’t fit your Obama Party theory that high taxes equal economic growth.

look, unemployment rates and state tax rates are easy to find online. if gpw is correct, you shouldn’t have trouble verifying it.

Hell, the Soviet Union had “full employment.” Cuba has “full employment.” North Korea has “full employment.” But did they have prosperity? Not so much.

50% of all the private sector job growth in the USA over the last decade has occurred in a single state; a state that has No Income Tax: Texas.

thanks guys.

FYI – it appears that Nevada and Florida are also doing well in the business market. Could it be because those states also don’t have a state income tax?

Regards,
Peter H.

Why is it a state like Texas, with no income tax, is coming through the recession in much better fiscal shape than a state like California… which has obscene income tax rates?