Starting point

The average effect is not the full psychological question

Population-average view

Test-related anxiety is usually treated as a negative predictor of achievement.

\[\text{Achievement}_{ij} = \beta_0 + \beta_1\text{Anxiety}_{ij} + e_{ij}\]

Question usually asked: Is anxiety negatively related to performance?

Contextualized view

The same individual difference may not have the same meaning everywhere.

\[\beta_{1j} = \text{local anxiety--achievement association}\]

Question here: Where is anxiety more prevalent, and where is it more consequential?

The gap

What the literature already suggests

  • Anxiety differs across students, schools, and systems.
  • Test and math anxiety show heterogeneous associations with achievement.
  • Classroom climate, support, comparison, and opportunity to learn matter.

What is still missing

  • Fine-grained geographical analysis within a single country.
  • Direct distinction between level variation and effect variation.
  • Socio-ecological correlates of both where anxiety is higher and where it matters more.

Study

Research questions

Does test-related anxiety show both a geography of prevalence and a geography of consequence?

  1. Do municipality-level anxiety estimates show stable geographical variation?
  2. Does the anxiety–achievement association vary across municipalities?
  3. Are anxiety levels and anxiety effects related to the same socio-ecological features?

Data

INVALSI population data

  • Italian Grade 5 students
  • Six cohorts: 2012/2013 to 2017/2018
  • Final analytic sample: 2,694,689 students
  • 3,000 municipalities
  • Median municipality sample size: 520 students

Measures and context

Student-level variables

  • Test-related anxiety: 4 items during standardized testing
  • Mathematics achievement
  • Reading achievement
  • Sex as covariate

Municipality indicators

  • Demographic: age, density, female share
  • Social: immigration, migration, marriage, children per family
  • Economic: income, education, low-productivity employment
  • Geographic: altitude, coast, area, north–south gradient

All variables were standardized.

Results

1. Anxiety levels were geographically structured

  • Split-half stability of municipality anxiety: r = .52
  • Spearman–Brown reliability: .69
  • Municipality average anxiety:
    • SD = 0.120
    • range: -0.36 to 0.63
  • Adjusted random intercepts:
    • SD = 0.088
    • range: -0.26 to 0.32

Even after accounting for the broad north–south gradient, municipalities differed meaningfully in average test-related anxiety.

2. The anxiety–achievement association varied by place

Average association

  • Mathematics: b = -0.189
  • Reading: b = -0.181

Slope variability

  • Mathematics slope SD: 0.027 [-0.26; -0.10]
  • Reading slope SD: 0.018
  • Stronger variability for mathematics

The negative association was not constant across municipalities.

3. Socio-ecological patterning was not random

Anxiety levels were higher in municipalities with

  • Lower income: r = -.42
  • Higher low education: r = .40
  • Higher low-productivity employment: r = .32
  • More southern location: r = .36

The anxiety–math association tended to be:

  • Weaker in more southern municipalities and in municipalities with more low-productivity employment.
  • Stronger in higher-income municipalities and municipalities with higher immigrant share.

Robustness

Checks performed

  • Added north–south gradient to interaction models.
  • Compared single-indicator models with conceptual blocks.
  • Adjusted models for province and region.

Main conclusion

Cross-level interaction estimates were more stable than main effects.

The core pattern remained: municipality characteristics were systematically related to local anxiety–achievement slopes, especially for mathematics.

Discussion

What this contributes to personality and socio-ecological psychology

Beyond educational description

  • The study extends geographical psychology to a contextualized emotional response in childhood.

  • It shows spatial patterning not only in where students are more anxious, but also in where anxiety is more tied to achievement.

Beyond average effects

  • Individual differences may have different correlates and consequences across local ecologies.

  • This is consistent with person–environment fit and broader socio-ecological models of psychological functioning.

But with caution

The municipality indicators should be read as ecological correlates, not as causal mechanisms.

  • The design is observational and cross-sectional at the student level.
  • Municipality indicators are strongly interrelated.
  • Municipality is only one contextual level; schools, classrooms, families, provinces, and regions also matter.
  • The anxiety measure is brief and task-embedded.

Take-home message

  1. Prevalence: municipalities differ in average anxiety.
  2. Consequence: municipalities differ in the anxiety–achievement association.
  3. Patterning: these two geographies overlap only partly and covary with socio-ecological indicators.

Psychological effects should be tested not only for whether they exist on average, but also for where they generalize.

Thank you

Tommaso Feraco
Department of General Psychology · University of Padova

GPS-A · The Geographical Psychology of School Anxiety

Contact: tommaso.feraco@unipd.it
Slides

SEM course website