Page 236 - Between One and Many The Art and Science of Public Speaking

Chapter 8 Supporting Your Message 203

pling. Thus, at a minimum, we should never accept a poll at face value. We
need to fi nd information about the sample on which the results are based.
• Are the differences in the poll greater than the margin of error? Good polling
re sults state the margin of error. Keep in mind that the margin of error
increases as the sample gets smaller. Whereas the margin of error for a
sample of 1,067 people is about plus or minus 3 percent, for 150 people
6
the margin of error is about plus or minus 8 percent. Suppose a poll has
a margin of error of plus or minus 4 percent. This means if the poll shows
a political candidate ahead of her opponent by 51 to 49 percent, she could
be ahead by as much as 55 to 45 percent, or behind by 47 to 53 percent—or
any number in between. When only subgroups of a larger sample are con-
sidered, there are even more chances for error. For example, on the morn-
ing of November 2, 2004, supporters of Senator Kerry were ecstatic when
early exit polls from key states such as Ohio and Florida indicated he was
defeating President Bush. What they failed to realize was that voters who
cast their votes early in the day were not representative of voters at large. In
fact, the subgroup sampled in these early exit polls was 59 percent female, a
group more likely to support Kerry than were men. When all the results were
in, of course, Kerry was defeated in both states and Bush was reelected. 7
• What are the percentages based on? “There’s been a 10 percent increase in the
rate of infl ation!” Sounds pretty alarming, doesn’t it? However, unless you
know what the underlying rate of infl ation is, this is a meaningless fi gure.
Infl ation rates are themselves a percentage. Say that infl ation is running at
4 percent. That means what cost $100 last year now costs $104. A 10 per-
cent increase in the rate of infl ation means that it would cost $104.40—not
too bad. On the other hand, a 10 percent rate of infl ation means that what
cost $100 a year ago now costs $110. Sound confusing? It is. The point
is that we need to be sure we understand what percentages are based on
before relying on them to prove a point.

• What is meant by average? One of the most frequently reported numbers is
the average, or mean. Although easily computed, the average is often mis-
leading because it is commonly distorted by numerical extremes. Consider
a newspaper report that states the average salary for new college graduates
is $40,000 a year. That doesn’t mean a majority of college graduates are
paid $40,000 a year. It simply means that when we add the salaries paid to
all college graduates surveyed and divide that sum by the number of grad-
uates in the sample, that’s the mean (the arithmetic average). The number
likely has been distorted by graduates in engineering, computer science,
and information systems management, who, though few in number, start at
salaries two to three times as much as their more numerous counterparts in
the liberal arts and social sciences. The most telling number is always the
median, which is the midpoint in a distribution of numbers. Knowing the
median tells us that half of the numbers in the distribution are larger, and
half are smaller. In Chapter 15, we discuss the mean and median in more
detail.

This list of questions is not meant to discourage you from using numerical data.
They can be a powerful form of support. The key is to know what your numbers
mean and how they were collected, and to avoid biased sources and question-

8/24/07 3:20:52 PM
M4344.indb 203 8/24/07 3:20:52 PM
M4344.indb 203