And the answer is… we think… more or less…­

As the pres­i­den­tial elec­tion moves closer and closer, vot­ers have ques­tions that need to be answered in a clear, con­cise and unam­bigu­ous way. For exam­ple, how many jobs are being cre­ated? Or, get­ting down to the elec­tion itself, which can­di­date is likely to win, Pres­i­dent Obama or Gov­er­nor Romney?

And it’s not just at elec­tion time that ques­tions come up. Busi­nesses are inter­ested in such issues through­out the year as they are try­ing to deter­mine what pro­duc­tion should be or how many peo­ple need to be hired.

So, with regard to the ques­tion above as to how many jobs were cre­ated in Sep­tem­ber, the answer is 873,000. No, no; wait, that’s not right. The answer is 114,000.

Per­haps that one is too com­pli­cated. So, how about the pres­i­den­tial race here in Ohio? Who is ahead in the polls at this time and seems likely to win our state’s elec­toral votes? Accord­ing to the Quinnipiac/CBS News poll, the pres­i­dent is nicely ahead — it would seem — with a lead of 50 per­cent to 45 per­cent. No, no; it’s closer than that accord­ing to the Fox poll, with Obama at 46 per­cent and Rom­ney at 43 percent.

So, what gives here? Num­bers seem to be all over the place, no mat­ter if we are exam­in­ing the strength of the job mar­ket or the vot­ing habits of Ohioans. How can there be such a wide diver­gence of num­bers that should pro­duce a defin­i­tive answer? Is some­body sim­ply lousy at doing their job and we get a bunch of flakey figures?

No, actu­ally the dif­fer­ences are a func­tion of the means by which data are gath­ered in this coun­try (and all oth­ers) given lim­ited finan­cial resources to col­lect use­able information.

Con­sider the issue of how many peo­ple is this coun­try have jobs and how many new jobs are cre­ated in a par­tic­u­lar month. Obvi­ously, there is a defin­i­tive num­ber for each of the above ques­tions. But since the U.S. has (as best we can deter­mine) more than 155 mil­lion peo­ple who want a job, it is impos­si­ble to con­tact each one of them every month to inquire about their employ­ment sit­u­a­tion. As a result, each month (as dis­cussed in pre­vi­ous columns) the gov­ern­ment con­ducts two sep­a­rate ran­dom sam­ples, one of approx­i­mately 60,000 house­holds and the other of approx­i­mately 400,000 indi­vid­ual busi­ness estab­lish­ments. From those two survey-based research efforts, they infer from the spe­cific sample(s) to the gen­eral pop­u­la­tion. While there cer­tainly are defin­i­tive answers as to how many peo­ple have jobs (that fig­ure being called a “pop­u­la­tion para­me­ter”), the survey-based research pro­duces a “sam­ple sta­tis­tic” which will, hope­fully, esti­mate well the unknown pop­u­la­tion parameter.

Of course, we don’t and won’t know what the pre­cise fig­ure is (the pop­u­la­tion para­me­ter), so each sam­ple sta­tis­tic rep­re­sents a method­olog­i­cally rea­son­able best-guess. But since all ran­dom sam­ples of the same size from the same pop­u­la­tion can­not be expected to pro­duce iden­ti­cal esti­mates, there is a level of con­fi­dence and a mar­gin of error that must be con­sid­ered due to vari­abil­ity which may exist from one sam­ple to another. As such, it is always impor­tant to note the con­fi­dence level uti­lized in the sam­pling work (typ­i­cally set at 95 per­cent) and the mar­gin of error for any par­tic­u­lar sam­ple size (for exam­ple, plus or minus 3 per­cent­age points).

So, in the pres­i­den­tial con­test in Ohio, if a par­tic­u­lar poll says Obama leads Rom­ney (or vice versa, as I don’t want to show par­tial­ity) by 50 per­cent to 45 per­cent, with a mar­gin of error of 3 per­cent­age points, Pres­i­dent Obama’s sup­port should range between 47 and 53 per­cent (that is, 50 plus/minus 3 per­cent­age points) in 95 cases out of 100 (pro­duc­ing a 95 per­cent con­fi­dence level). Sim­i­larly, the level of sup­port for Gov­er­nor Rom­ney would range between 42 and 48 per­cent. Since the two sets of sup­port fig­ures over­lap within the “mar­gin of error,” the race is con­sid­ered a sta­tis­ti­cal tie.

Vir­tu­ally all data, whether mea­sur­ing sup­port for can­di­dates or the state of the econ­omy, are gath­ered in this fash­ion and rep­re­sent a best esti­mate of some pop­u­la­tion para­me­ter. The process does not involve a fudg­ing of the num­bers (such as some “Chicago guys” manip­u­lat­ing job fig­ures for polit­i­cal advan­tage), but rather an hon­est attempt to esti­mate some impor­tant vari­able. It may not be pre­cise, but the process tends to pro­duce use­able results… results that are cer­tainly bet­ter than noth­ing when help­ing some­one in a decision-making process.

Dr. James New­ton serves as chief eco­nomic advi­sor to Com­merce National Bank and is an aux­il­iary fac­ulty mem­ber in eco­nom­ics and sta­tis­tics at OSU-Marion and OSU-Newark. Dr. Newton’s views do not nec­es­sar­ily reflect those of Com­merce National Bank or OSU-Marion/Newark.

Posted by Jim Newton on Oct 23 2012. You can follow any responses to this entry through the RSS Feed. Comments can be made below.