Data Trending

Repeated measurements over time are used to track changing trends. When measurements are based upon sample surveys over time, data at different points in time may be different due to different reasons. The most obvious reason is that the underlying phenomenon has changed. An alternate reason is that the changes are due to sampling error, even though the underlying phenomenon has not changed at all. It is not an easy task to sought out the one from the other.

Without amassing the vast array of statistical tools that can be brought to bear on this issue, we shall look at a couple of variables across time in the Los Medios y Mercados de Latinoamérica study.

First of all, we look at the percent of persons using radio by daypart across three years. These numbers are shown in Table 1. In most cases, the annual numbers are within 1 or 2 percentage points of each other. We can fairly conclude that these numbers are stable over time. If anything, we would wonder that such stable behavior needed to be tracked as frequently as we have.

Table 1. Percent of Persons Using Radio by Daypart (1994-1995-1996)

Daypart

1994 %PUR 1995 %PUR 1996 %PUR
Monday-Friday 6am-10am 22 24 24
Monday-Friday 10am-3pm 19 21 20
Monday-Friday 3pm-7pm 19 18 18
Monday-Friday 7pm-12m 11 9 9
Monday-Friday 12m-6am 4 3 2
Saturday 6am-10am 18 20 20
Saturday 10am-3pm 18 20 19
Saturday 3pm-7pm 17 16 15
Saturday 7pm-12m 9 7 8
Saturday 12m-6am 3 2 1
Sunday 6am-10am 15 16 15
Sunday 10am-3pm 16 15 14
Sunday 3pm-7pm 16 12 13
Sunday 7pm-12m 7 6 6
Sunday 12m-6am 2 1 1

(Source: Los Medios y Mercados de Latinoamérica 1994, 1995, 1996)

As another example, we consider the consumer confidence index. The definition is given elsewhere. The numbers are shown in Table 2. Here we see that there are some considerable fluctuations from year to year. From our first example, which is based upon the same sample of respondents, we know that the data can be relatively stable when the underlying phenomenon is unchanging. Therefore, we would now be inclined to believe that the consumer confidence indices may be reflecting real changes in perceptions and attitudes. In fact, we observe that the largest decreases in consumer confidence occurred in the two countries that were hurt most severely by the Mexican peso crisis --- Argentina and Mexico.

The magnitude of the changes from year to year would suggest that variables such as consumer confidence indices should be measured even more frequently (such as on a monthly basis) in order to react in a timely manner.

Table 2. Consumer Confidence Index Versus One Year Ago by Country, 1994-1995-1996

Country

1994 Index 1995 Index 1996 Index
Argentina 82 84 55
Brazil 133 101 129
Chile 109 86 113
Colombia 135 114 118
Mexico 107 80 62
Puerto Rico 125 87 138
Venezuela 57 89 74

(Source: Los Medios y Mercados de Latinoamérica 1994, 1995, 1996)

(posted by Roland Soong, June 20, 1997)


(Return to Zona Latina's Home Page)