Comments on: Disability Insurance Recipiency Bubble Plot

By: David

David — Sat, 22 Feb 2014 23:28:37 +0000

For those of you pondering the relationship between reported health status and disability recipiency, remember that a 60-year-old reporting himself in good health is likely to overlook more afflictions than a 20-year-old. This isn’t the whole story by a long shot – there has been a rise in disability recipiency for other reasons, but the self-reported health status isn’t likely to catch the full impact of the aging boomer population.

By: Joe Mako

Joe Mako — Wed, 29 Jan 2014 21:37:40 +0000

Here is a horizontal dot and bar chart with the labels subdued. This lets us see that the Recipiency Rate is increasing each year, while there is no correlation to the Health Status.

By: Mynda

Mynda — Sat, 18 Jan 2014 02:58:24 +0000

Another approach might be to index the values to show the change over time. This gives a more dramatic effect of how they have diverged.

We can see from the chart that 2.32 times as many people are getting disability benefits in 2010 than were in 1982, while those reporting poor health in 2010 is below 1982 levels.

By: Mynda

Mynda — Thu, 16 Jan 2014 10:24:17 +0000

I think Matt is on the right track with plotting the 100-health status. The lower figure allows for clearer comparison between the two numbers.

I like the simple line chart over time using the same axis as opposed to Michael’s secondary axis for the proportion claiming insurance. I think this exaggerates the difference, in the same way starting the vertical axis above zero distorts the highs and lows.

Plotting the lower number (100-health status) and actual recipients on the same axis over time shows the disparity.

So what’s the reason? Are people simply more aware that they can claim than they were in 1982?

By: Xan Gregg

Xan Gregg — Sat, 04 Jan 2014 21:59:02 +0000

I like a scatter plot matrix for comparing variable pair relationships. For a less technical audience, presented them as separate graphs may work better. The regression line and confidence interval suggests the degree of correlation.

By: Michael W Cristiani

Michael W Cristiani — Sun, 15 Dec 2013 02:06:50 +0000

Or,

By: Michael W Cristiani

Michael W Cristiani — Sun, 15 Dec 2013 01:40:10 +0000

Hows about this? It shows the stability of the health status vlaues and makes it very clear that disability recipiency has been “on the rise”. If you wanted to explore Kim’s musing about what is really driving the recipiency rate, you could add in some other data, like average age of recipients (size or color ramping on the Recipent Rate (circles). Another interesting piece of data might be the proportion with or average number of multiple comorbd conditions among the working age population – again varying the color of the bars to make it clear; alternatively, the bars could be annotated or labeled with that information.

Peace and All Good!
Michael

By: Matt Fletcher

Matt Fletcher — Mon, 09 Dec 2013 22:17:29 +0000

Kim,

I agree that this chart shows the constancy of the health status line, but not so sure that the thickness increase helps drive home the point about the big increase in recipiency rate (it’s not absolutely clear to me without looking at the key that the rate has more than doubled)

I’m just wondering whether it would be possible to plot something more like (100-health status) – this is also fairly consistent but much lower, and presumably represents the proportion of people not reporting good health(?) which may well be the metric of more interest when looking at drivers of increased recipiency rates. There’s also less white space (though maybe still too much?) when starting both scales at 0.

By: Kim Rees

Kim Rees — Mon, 09 Dec 2013 21:26:19 +0000

I agree with Sheila (except the min/max difference is a mere 2.2%). Here’s what it looks like with the axis starting at 0 (thickness showing recipiency). Showing the chart in this way would enhance your point, Jon. I guess it leads me to ask, if health status isn’t driving disability, what is?

By: Matt Fletcher

Matt Fletcher — Sun, 08 Dec 2013 23:41:57 +0000

Agree re rotating – I did it both ways and attached the wrong one! Re variability, the suggested plot looks pretty flat if you use the original scale from the bubble plot – but like you I don’t know whether 3% is a large or small variation, so I set the axes purely based on the data.