When you are dealing with variables that are at least ordinal or higher in level of measurement, it is possible to talk about the relationship between those variables as having a direction.
A positive relationship is one in which
cases that score "low" on the independent variable tend to also score "low" on the dependent variable, and
cases that score "high" on the indepenent variable, tend to score "high" on the dependent variable.
A negative relationship is one in which
cases that score "low" on the independent variable tend to score "high" on the dependent variable, and
cases that score "high" on the indepenent variable, tend to score "low" on the dependent variable.
How can you tell?
In the absence of measures of association (which will tell you the direction of the relationship based on the sign of the statistic), you can typically determine the direction of the relationship by looking for specific patterns in the column percents found in the top and bottom rows of a crosstabulation containing ordinal level variables.
In a positive relationship the column percentages should:
Decrease as you look across to the right on the top row.
Increase as you look across to the right on the bottom row.
Examples
These tables represent examples of positive relationships
LOW
MED
HIGH
LOW
85%
70%
55%
MED
10%
15%
20%
HIGH
5%
15%
25%
LOW
MED
HIGH
LOW
60%
30%
10%
MED
30%
40%
30%
HIGH
10%
30%
60%
LOW
MED
HIGH
LOW
35%
20%
10%
MED
15%
15%
10%
HIGH
50%
65%
80%
In a negative relationship the column percentages should:
Increase as you look across to the right on the top row.
Decrease as you look across to the right on the bottom row.
Examples
These tables represent examples of negative relationships