MODULE 03

Lazarsfeld Paradigm

Start with a between X and Y. Introduce a Z. Then ask one question: what happened to the original relationship?

Research question
Variables
Variable roles (trained here)
Measurement levels
Method choice
Statistical logic (trained here)
SPSS workflow
Output interpretation (trained here)
Defensible conclusion (trained here)

Understand

One relationship, four possible stories

Graduates are 25 points more likely to be politically active. Pick a control variable, split the data into partial tables, and watch what happens to the 25-point gap.

Same zero-order relationship. Choose a control variable:

Zero-order model

% politically activeNo degreeDegree

Zero-order

+25 pp

Z: Women

+25 pp

Z: Men

+25 pp

pp = percentage-point difference between degree and no degree.

: graduates are 25 points more likely to be politically active.

The comparison that matters

Every Lazarsfeld pattern is defined by comparing the partial relationships with the zero-order relationship, and the partial relationships with each other.

Understand

The decision tree

Two questions are enough to classify any pattern. Explanation and interpretation produce identical tables; only the position of Z in the causal or temporal sequence tells them apart.

Question 1

What happens to the X–Y relationship after controlling for Z?

Summary

Replication

Same relationship after control.

partials ≈ zero-order

Specification

Relationship differs by subgroup.

partials differ from each other

Explanation

The original relationship is explained by an antecedent variable.

partials ≈ 0, Z antecedent

Interpretation

The original relationship operates through an intervening variable.

partials ≈ 0, Z intervening

Careful with causal language

Arrows show the proposed sequence, which comes from theory and timing, not from the tables. Survey crosstabs show associations; ‘explained by’ and ‘operates through’ describe how the pattern fits the model.

Try

Put the analysis in order

A Lazarsfeld analysis has five steps. Drag them into the right order (or use the arrow buttons).

Put the steps of a Lazarsfeld analysis in the correct order.

Drag the steps, or use the arrow buttons to move them.

1Prepare and analyse the two-dimensional crosstab.
2Summarise the results and classify the pattern within Lazarsfeld's paradigm.
3Set up the two-dimensional and three-dimensional model.
4Analyse the three-dimensional crosstab: inspect the partial tables and compare them with the zero-order relationship.
5Interpret the model, formulate hypotheses, and identify the type of control variable.

Why do we need the 2D table first?

Because the purpose of the three-dimensional analysis is to examine what changes relative to the original zero-order relationship.

Try

Practice scenarios

Ten short sociology scenarios. For each one: identify X, Y and Z, decide whether Z is antecedent or intervening, read the partial tables, and name the pattern.

Scenario 1 of 10

Education, participation and gender

A survey finds that graduates are more likely to have taken part in a political activity (petition, demonstration, contacting a politician) in the past year. The researcher checks whether this holds for both women and men.

% politically active by education, zero-order and within categories of gender
	Zero-order	Women	Men
No degree	22%	21%	23%
Degree	47%	45%	49%

Question 1 of 6

What is X (the independent variable)?

SPSS

Zero-order and partial tables in SPSS

The same Crosstabs dialog produces both tables: without a layer variable you get the zero-order table, with Z in the Layer box you get one partial table per category of Z.

Before you open SPSS

1What is the research question?Is education associated with political participation, and does this change when controlling for Z?
2What are X, Y and Z?Xeducation Yparticipation Zcontrol variable
3What are their measurement levels?Categorical (nominal or ordinal), so crosstabs fit.
4What are we comparing?The distribution of Y within categories of X, first overall, then within each category of Z.
5Which method?Two- and three-dimensional crosstabs with percentages within X, χ² and Cramer's V.

Analyze›Descriptive Statistics›Crosstabs

Crosstabs

gender

settlement

parstatus

polint

age

1Row(s):

educ (X)

2Column(s):

polpart (Y)

3Layer 1 of 1:

Z (e.g. gender)

Cells… › Row percentagesStatistics… › Chi-squareStatistics… › Phi and Cramer's V

PasteResetCancelOK

1
Row(s):
Put X in the rows so you can request row percentages (percentages within X).
2
Column(s):
The dependent variable goes in the columns.
3
Layer 1 of 1:
Leave empty for the zero-order table. Add Z here to get one partial table per category of Z.
Cells… › Row percentages
Percentages within each category of X (because X is in the rows).
Statistics… › Phi and Cramer's V
SPSS reports χ² and V separately for each layer, so you can compare partial relationships.
Run it twice: first without a layer (zero-order), then with Z as the layer (partial tables).

Test yourself

Mastery check

0/0

Identify Lazarsfeld pattern

X = education, Y = political participation, Z = gender. Which pattern does the table show?

% politically active, by education (X) and control variable (Z)
	Zero-order	Women	Men
Low education	22%	21%	23%
High education	47%	45%	49%
Difference	+25 pp	+24 pp	+26 pp

1 / 5