When Urban Design Meets Evidence
As architects and urban designers, we shape physical space.
But the forces that shape where and how we build are often taken
for granted. Why does construction intensify rapidly in some districts while
remaining subdued in others? Why do certain places urbanise relentlessly, even
in the absence of strong income growth or population pressure?
This article approaches these questions empirically. Using
district-level data from Kerala’s Department of Economics & Statistics for
the period 2016–2022, it examines the drivers of urban construction intensity
across all fourteen districts of the state. A series of regression models is
used to test competing explanations commonly invoked in professional and policy
discourse; income growth, construction costs, migration, and temporal shocks against
spatial and institutional factors.
The results challenge several intuitive assumptions. Urban
construction in Kerala is not primarily driven by rising incomes or short-term
price movements. Instead, it is strongly shaped by where construction is
already concentrated: within urban jurisdictions, within particular
districts, and within institutional settings that reinforce cumulative
advantage. Migration-linked capital plays a role, but selectively amplifying
construction only where urban structures already enable it.
One result is particularly striking. Differences between
districts explain roughly three times more variation in urban
construction intensity than year-to-year change within districts. In practical
terms, whether construction accelerates depends far more on which district
one is in than on whether it is 2018, 2020, or 2022.
Many of these patterns will feel familiar to those who know
Kerala well. What has been missing is a systematic test of which explanations
actually hold once district structure, time effects, income, prices, and
migration are examined together. This study provides that test, using publicly
available data and transparent methods, with implications for how architects,
planners, and policymakers understand urbanisation in Kerala; not as a smooth
economic outcome, but as a spatially mediated, institutionally reinforced
process.
Data Assembly
This analysis integrates multiple public datasets from
Kerala’s Department of Economics & Statistics into a balanced district–year
panel. The final dataset comprises 98 observations, covering 14 districts over
seven years (2016–2022).
The study period spans pre-pandemic stability, the COVID-19
shock, and the immediate post-pandemic adjustment, making it particularly
informative for examining changes in construction dynamics under both normal
and disrupted conditions. All datasets were harmonised to a common
district–year structure using end-year conventions, allowing consistent
temporal alignment across construction activity, population, prices, wages,
income, and migration indicators.
This panel structure enables the analysis to distinguish
persistent district characteristics from year-specific shocks, an analytical
separation that is not possible in purely cross-sectional studies.
Figure 1: The urban-rural construction divide varies
dramatically across districts, from Thiruvananthapuram's urban dominance to
Idukki's rural character. Graph by Author.
Methodological Approach
The empirical strategy employs ordinary least squares (OLS)
regression at the district level, with district and year fixed effects.
Districts are used as the unit of analysis because this is the scale at which
planning jurisdictions, construction records, income accounts, and migration
statistics align consistently in Kerala’s administrative system. The approach
prioritises institutional coherence, comparability, and reproducibility.
The dependent variable; urban construction intensity is
a constructed measure derived from administrative building records and defined
as:
Urban construction intensity = (Number of urban buildings
constructed) / (District population)
This variable is log-transformed to stabilise variance and
to facilitate interpretation of proportional effects. Population is used as the
scaling denominator to enable comparison across districts of very different
sizes and densities, ensuring the measure reflects relative development
pressure rather than absolute scale.
Figure 2: Time-series trends reveal distinct district
trajectories, with Thiruvananthapuram and Ernakulam showing strong post-2020
surges.
Urban–Rural Classification
Urban and rural construction are classified using Kerala’s
official administrative definitions. Urban buildings are those constructed
within statutory urban local bodies (Municipal Corporations and
Municipalities), while rural buildings fall within Grama Panchayat
jurisdictions. The classification is therefore administrative rather than
morphological, reflecting governance structures rather than built form alone.
Figure 1 illustrates the sharp variation in the urban–rural
construction divide across districts, ranging from Thiruvananthapuram’s strong
urban dominance to Idukki’s predominantly rural construction profile.
Data Scope and Limitations
Construction activity is measured using counts of
residential buildings, rather than built-up area or floor space. As a result,
the analysis captures the direction and intensity of construction activity, not
total constructed volume. This distinction is particularly relevant in Kerala,
where construction ranges from single-storey individual houses to multi-storey
apartment complexes.
Construction data are drawn from statutory administrative
records compiled by the Government of Kerala and routinely used in official
housing and infrastructure assessments. Financial variables were standardised
to persons and rupees for consistency. No extreme outliers were removed;
skewness is addressed through log transformations where appropriate.
Figure 3: The
strong positive relationship between urban share of construction and per-capita
urban construction intensity (R² = 0.92).
Figure 3 demonstrates a striking empirical regularity:
districts with a higher share of construction occurring within urban
jurisdictions consistently exhibit much higher levels of per-capita urban
construction intensity.
This strong relationship raises a natural follow-up
question. If urban construction is accelerating so sharply, is it being driven
by conventional economic pressures such as rising costs and inflation, or does
the observed pattern reflect deeper structural dynamics tied to where
construction is institutionally concentrated? The next figure examines this
question directly by introducing inflation as a competing explanation.
Figure 4: Consumer Price Index (CPI), Kerala, 2012–2025. Graph by Author
Figure 4 shows that, despite visible movement in the CPI
series over the study period, the apparent discontinuity around the late-2010s
reflects a change in the CPI base year rather than a sudden inflationary shock;
underlying price movements are comparatively steady. The absence of a
significant inflation effect therefore suggests that short-term cost pressures
alone cannot explain the observed acceleration in urban construction. This
points to the need to examine other mechanisms beyond prices in explaining
district-level construction dynamics.
Descriptive Statistics
The empirical analysis draws on a balanced district–year
panel covering all 14 districts of Kerala over the period 2016–2022, yielding
98 observations. The panel integrates administrative records on residential
construction, population, district income, prices, wages, and migration,
allowing systematic comparison of urban construction intensity across space and
time.
|
Panel A.
Urbanisation & Construction |
|||||||
|
Variable |
N |
Mean |
Std. Dev. |
Min |
Max |
Between SD |
Within SD |
|
Households (total) |
98 |
27,813 |
12,830 |
7,674 |
58,039 |
12,758 |
3,452 |
|
Households (urban) |
98 |
6,482 |
4,638 |
683 |
25,194 |
4,327 |
1,987 |
|
Households (rural) |
98 |
21,331 |
9,566 |
6,889 |
47,879 |
9,431 |
2,840 |
|
Population |
98 |
24,84,100 |
10,01,648 |
8,35,842 |
43,25,608 |
10,33,853 |
23,719 |
|
Urban share |
98 |
0.217 |
0.087 |
0.078 |
0.507 |
0.075 |
0.046 |
|
Urban construction (per capita) |
98 |
0.0024 |
0.0011 |
0.0006 |
0.0074 |
0.0009 |
0.0007 |
|
Total construction (per capita) |
98 |
0.0112 |
0.0019 |
0.0067 |
0.0168 |
0.0012 |
0.0015 |
|
Panel B.
Income & Output |
|||||||
|
Variable |
N |
Mean |
Std. Dev. |
Min |
Max |
Between SD |
Within SD |
|
Per-capita GDDP (real) |
98 |
1,48,359 |
30,753 |
94,170 |
2,22,154 |
28,801 |
12,945 |
|
Log per-capita GDDP |
98 |
11.886 |
0.207 |
11.453 |
12.311 |
0.196 |
0.084 |
|
GDDP growth rate |
84 |
0.031 |
0.101 |
−0.271 |
0.254 |
0.014 |
0.1 |
|
Panel C.
Migration & External Income |
|||||||
|
Variable |
N |
Mean |
Std. Dev. |
Min |
Max |
Between SD |
Within SD |
|
Net emigration rate |
98 |
0.015 |
0.018 |
−0.023 |
0.049 |
0.019 |
0 |
|
Emigrant rate |
98 |
0.06 |
0.019 |
0.022 |
0.089 |
0.02 |
0 |
|
External income proxy |
98 |
0.173 |
0.215 |
−0.275 |
0.579 |
0.221 |
0.002 |
|
Panel D.
Wages (available only for 2022) |
|||||||
|
Variable |
N |
Mean |
Std. Dev. |
Min |
Max |
Between SD |
Within SD |
|
Real wage |
14 |
790.3 |
46.6 |
671.7 |
858.3 |
46.6 |
0 |
|
Log real wage |
14 |
6.671 |
0.061 |
6.51 |
6.755 |
0.061 |
0 |
Dataset compiled to order by Author. Source: Kerala Department of Economics & Statistics (ecostat.kerala.gov.in)
About the Data Panels:
Urbanisation & Construction | |||||||
Urban construction intensity (measured as urban residential
buildings per capita) exhibits substantial cross-district variation relative to
within-district change over time. The between-district standard deviation
exceeds the within-district component, indicating that persistent spatial
differences dominate year-to-year fluctuations. Total construction per capita
shows considerably less dispersion, suggesting that variation in construction
activity is driven more by where construction occurs than by overall
building volume.
Income and Output
Panel B reports district-level income indicators. Real
per-capita GDDP averages ₹1.48 lakh, with moderate variation across districts
compared to construction outcomes. While income differences are economically
meaningful, their dispersion is notably smaller than that observed for urban
construction intensity. Growth rates display substantial year-to-year
volatility, particularly around the pandemic period, but comparatively limited
between-district variation. This contrast motivates later analysis examining
whether income levels and growth can account for observed spatial patterns in
construction.
Migration and External Income
Panel C presents migration-related variables. Net emigration
rates vary both positively and negatively across districts, reflecting Kerala’s
long-standing role as a migration-linked economy with heterogeneous exposure to
external labour markets. The external income proxy exhibits high
between-district variation but minimal within-district change, consistent with
persistent structural differences in migration intensity rather than short-term
fluctuations. These patterns suggest a potential channel linking external
income flows to spatial development outcomes.
Wages and Prices
Panel D reports real wage data, available only for 2022.
Wage variation is entirely cross-sectional, with no within-district component,
reflecting the single-year nature of the data. Consumer price inflation (CPI),
based on the CPI (IW & AL) series with base year 2011–12, varies
meaningfully over time but shows limited cross-sectional dispersion. This
allows the analysis to distinguish nominal price movements from structural
drivers of construction activity.
Taken together, the descriptive statistics point to a
central empirical fact: urban construction intensity varies far more sharply
across districts than do income levels, growth rates, or prices. Persistent
spatial differences dominate temporal fluctuations, suggesting that standard
macroeconomic explanations alone are unlikely to account for Kerala’s observed
urbanisation patterns. These patterns raise a natural question—whether such
spatial variation reflects systematic structural drivers—which the subsequent
regression analysis is designed to test.
The Research Question: What Drives Urban Construction Intensity?
At its core, this study asks two related questions:
- Where
does urban construction concentrate across districts in Kerala?
- When
does construction intensity accelerate or stall within districts over
time?
The first question concerns persistent spatial differences
between districts, while the second addresses temporal dynamics common to all
districts. The econometric strategy is designed to separate these two
dimensions and to test whether standard economic explanations; prices, income,
growth, and migration can account for observed patterns in urban construction.
Model Framework and Estimation Strategy
To address these questions, the analysis estimates a
sequence of district–year fixed-effects regression models with log
per-capita urban construction intensity as the dependent variable. All
models include district and year fixed effects, ensuring that results are
identified from within-district variation over time, net of unobserved district
characteristics and common shocks.
Rather than presenting results selectively, Table 2
reports the full set of seven model specifications, which serve as the
reference point for the discussion that follows.
|
Table 2.
Determinants of Urban Construction Intensity in Kerala (2016–2022) |
|||||||
|
Dependent
variable: log of per-capita urban construction intensity |
|||||||
|
|
1 |
2 |
3 |
4 |
5 |
6 |
7 |
|
Variable |
Baseline
FE |
Urban
Share |
Urban
Share + CPI |
Urban
Share + GDDP |
Lagged
Growth |
Migration |
Migration
+ GDDP |
|
Urban
share |
5.324* |
5.324* |
5.234* |
5.324* |
5.234* |
||
|
|
|
(-0.355) |
(-0.355) |
(-0.347) |
|
(-0.355) |
(-0.347) |
|
CPI |
0 |
||||||
|
|
|
|
(-0.001) |
|
|
|
|
|
Log
per-capita GDDP |
−0.716* |
−0.716* |
|||||
|
|
|
|
|
(-0.299) |
|
|
(-0.299) |
|
Lagged log
urban construction |
0.074 |
||||||
|
|
|
|
|
|
(-0.065) |
|
|
|
Lagged urban
share |
−0.714 |
||||||
|
|
|
|
|
|
(-0.414) |
|
|
|
Net
emigration rate |
85.577** |
−31.746 |
|||||
|
|
|
|
|
|
|
(-25.423) |
(-54.827) |
|
Ernakulam |
0.602*** |
0.037 |
0.037 |
(0.103) |
(0.021) |
0.496* |
(-0.068) |
|
Idukki |
-0.533*** |
-0.159+ |
-0.159+ |
-0.288** |
(-0.038) |
(0.303) |
(-0.459) |
|
Kannur |
0.454** |
-0.131 |
-0.131 |
-0.307** |
(0.027) |
-2.247*** |
(0.478) |
|
Kasaragod |
-0.179 |
0.033 |
0.033 |
-0.241+ |
(-0.042) |
-0.853*** |
(0.087) |
|
Kollam |
-0.073 |
-0.079 |
-0.079 |
(-0.082) |
(-0.003) |
(0.095) |
(-0.146) |
|
Kottayam |
-0.343* |
-0.06 |
-0.06 |
(-0.117) |
(-0.019) |
-1.622*** |
(0.462) |
|
Kozhikode |
0.409** |
0.011 |
0.011 |
-0.203+ |
(-0.002) |
1.796** |
(-0.865) |
|
Malappuram |
0.158 |
0.187* |
0.187* |
(-0.199) |
(-0.059) |
0.998*** |
(-0.500) |
|
Palakkad |
-0.124 |
0.067 |
0.067 |
(-0.204) |
(-0.044) |
0.482** |
(-0.357) |
|
Pathanamthitta |
0.114 |
-0.07 |
-0.07 |
-0.346* |
(-0.012) |
-2.910*** |
(0.707) |
|
Thiruvananthapuram |
0.818*** |
-0.218* |
-0.218* |
-0.271** |
(0.042) |
3.067** |
(-1.489) |
|
Thrissur |
0.221 |
-0.198* |
-0.198* |
-0.263** |
(0.016) |
1.309** |
(-0.822) |
|
Wayanad |
0.242 |
0.256** |
0.256** |
(-0.095) |
(-0.062) |
|
|
|
District
fixed effects |
Yes |
Yes |
Yes |
Yes |
Yes |
Yes |
Yes |
|
Year fixed
effects |
Yes |
Yes |
Yes |
Yes |
Yes |
Yes |
Yes |
|
Model
diagnostics |
|||||||
|
Observations |
98 |
98 |
98 |
98 |
84 |
98 |
98 |
|
R² |
0.694 |
0.922 |
0.922 |
0.927 |
0.622 |
0.922 |
0.927 |
|
Adjusted R² |
0.62 |
0.902 |
0.902 |
0.907 |
0.502 |
0.902 |
0.907 |
|
Notes:
Robust standard errors in parentheses. p < 0.1, * p < 0.05, ** p <
0.01, *** p < 0.001. |
|||||||
Table 1: Regression results for all seven model
specifications. Dependent variable is log per-capita urban construction
intensity.
Model Progression (How to Read Table 2)
The seven specifications in Table 2 are structured to
incrementally test competing explanations for urban construction intensity:
Model 1 (Baseline FE): District and year fixed
effects only, capturing unexplained spatial and temporal variation.
Model 2 (Urban Share): Adds the urban share of
construction, testing whether where construction is institutionally
concentrated matters.
Model 3 (Urban Share + CPI): Adds consumer price
inflation to assess cost-pressure effects.
Model 4 (Urban Share + GDDP): Adds log per-capita
GDDP to test income-driven explanations.
Model 5 (Lagged Growth): Introduces lagged
construction and urban share to test dynamic persistence.
Model 6 (Migration): Adds net emigration rate to
capture migration-linked external income effects.
Model 7 (Migration + GDDP): Combines migration and
income to assess their joint explanatory power.
This stepwise structure allows each mechanism: spatial
concentration, prices, income, growth dynamics, and migration to be evaluated
both individually and conditionally.
Why District-Level Coefficients Are Reported
District coefficients are retained in all specifications to
make spatial heterogeneity explicit rather than absorbed silently into fixed
effects. Their inclusion allows direct comparison between districts after
controlling for structural drivers, and helps interpret whether observed
spatial differences persist once key explanatory variables are introduced.
Roadmap for the Results Discussion
The analysis below walks through the models in Table 2
sequentially, focusing on three core findings:
Urban concentration dominates:
where construction occurs matters more than prices or income.
Income and prices play limited
roles once spatial structure is accounted for.
Migration modifies—but does
not overturn—spatial patterns, reinforcing rather than replacing
district-specific dynamics.
Each subsection draws directly on Table 2, avoiding
selective reporting and maintaining transparency throughout.
Methodological Approach: A Progressive Inquiry
Rather than relying on a single specification, the analysis
estimates seven progressively enriched fixed-effects models (M1–M7), each
designed to test a specific hypothesis about the drivers of urban construction
intensity. This stepwise approach allows individual mechanisms to be isolated
and evaluated both independently and conditionally.
All models include district and year fixed effects, which
absorb time-invariant district characteristics such as geography, planning
regimes, and historical urban form, as well as common year-specific shocks.
Coefficients are therefore identified from within-district changes over time,
net of persistent spatial differences.
Model 1 (M1): Baseline Fixed Effects
Question: Are there systematic differences in urban
construction intensity across districts, and does intensity change over time?
The baseline model includes only district and year fixed
effects. It tests whether location and time period alone explain observed
variation in urban construction intensity.
Key finding: Yes, strongly so. District and year
effects alone explain a substantial share of the variation (R² = 0.694),
indicating that spatial and temporal structure matters even before introducing
economic or demographic variables.
Selected district coefficients illustrate this
heterogeneity:
Thiruvananthapuram (+0.818***): The highest baseline
intensity, consistent with its role as state capital and administrative centre.
Ernakulam (+0.602***): Elevated construction
intensity reflecting its position as Kerala’s primary commercial hub.
Idukki (−0.533***): Significantly lower intensity,
consistent with terrain constraints and limited urban jurisdiction.
These coefficients confirm that persistent district
characteristics dominate construction outcomes, motivating closer
examination of the mechanisms behind this spatial structure.
Models 2–3 (M2–M3): The Urban Share Hypothesis
Question: Does the location of
construction—urban versus rural—systematically shape construction intensity?
These models introduce urban share, defined as the
proportion of total construction occurring within statutory urban
jurisdictions.
Key finding: Urban share is the single most powerful
predictor in the analysis. The estimated coefficient (≈ 5.32, p < 0.001)
implies that a 10-percentage-point increase in urban share is associated
with a 0.53 increase in log per-capita urban construction intensity. Model
fit increases sharply (R² = 0.922).
Because the dependent variable is in logarithmic form, this
corresponds to an increase of roughly 65–70 percent in per-capita urban
construction intensity.
This result indicates a strongly nonlinear and
self-reinforcing relationship: districts that shift construction activity
toward urban jurisdictions experience disproportionate increases in urban
construction intensity. Once urban construction reaches a critical share,
subsequent construction increasingly concentrates in urban locations.
At this stage, the result establishes a structural empirical
fact: where construction is institutionally located matters more than how
much construction occurs overall.
Model 4 (M4): Prices and Cost Pressures
Question: Do inflationary cost pressures influence
urban construction intensity?
Model 4 adds the Consumer Price Index (CPI) to test
whether construction responds to short-term price movements.
Key finding: CPI has no statistically significant
effect (coefficient ≈ 0, p > 0.1). This remains true despite substantial
cumulative inflation over the study period.
This null result suggests that short-term cost pressures
do not systematically drive urban construction intensity once spatial structure
is accounted for. While prices undoubtedly affect project-level
feasibility, they do not explain the observed cross-district patterns in
construction intensity.
Model 6 (M6): The Income Hypothesis
Question: Does economic prosperity drive urban
construction intensity?
Model 6 introduces log per-capita Gross District Domestic
Product (GDDP) to test whether higher income levels translate into more
intensive urban construction, conditional on urban share.
Key finding: The coefficient on log per-capita GDDP
is negative and statistically significant (−0.716, p < 0.05). Holding
urban share constant, districts with higher income levels exhibit lower
per-capita urban construction intensity.
This result indicates that income alone does not explain
urban construction outcomes. Once the spatial concentration of construction
is accounted for, higher district income is not associated with intensified
urban building activity.
Interpretation: The negative coefficient is
consistent with a maturity effect. More prosperous districts may have
already completed major phases of residential expansion, face land constraints,
or experience slower marginal construction growth due to saturation,
regulation, or rising opportunity costs. In contrast, districts at intermediate
stages of urbanisation may exhibit higher construction intensity despite lower
income levels.
Crucially, this finding challenges the assumption that
economic growth mechanically translates into construction booms. Urban
construction intensity appears to depend less on aggregate prosperity than on
the spatial and institutional context in which construction occurs.
Model 7 (M7): The Growth Trajectory Hypothesis
Question: Do past urbanisation dynamics predict
future construction intensity?
Model 7 introduces lagged urban construction intensity
and lagged urban share to test whether construction follows smooth,
path-dependent trajectories.
Key finding: Evidence for dynamic persistence is
weak. The lagged urban share coefficient is marginally significant (−0.714, p
< 0.1), while lagged construction intensity is not statistically
significant. Overall model fit declines substantially (R² = 0.622).
These results indicate that past construction levels are
poor predictors of subsequent construction intensity once district and year
effects are controlled for.
Interpretation: Urban construction in Kerala does not
evolve along stable, deterministic growth paths. Instead, construction
intensity appears episodic and responsive to structural conditions rather than
governed by momentum alone. This weakens the case for extrapolative forecasting
models and underscores the importance of adaptive planning approaches that
respond to changing institutional and spatial conditions.
Models 6–7 (M6–M7): Migration and External Income Effects
Question: Does emigration influence urban
construction intensity?
Models M6 and M7 introduce net emigration rates as a
proxy for exposure to external income flows, recognising that direct remittance
data are unavailable at the district level.
Key finding: In Model 6, net emigration rate enters
with a positive and statistically significant coefficient (85.577, p <
0.01), suggesting that districts with higher out-migration tend to exhibit
higher urban construction intensity. However, this effect disappears once
district income (GDDP) is included in Model 7.
This pattern indicates that the migration coefficient is not
independently identified; instead, it captures broader economic and spatial
conditions correlated with income and urban structure.
Interpretation: Migration does not operate as an
autonomous driver of urban construction intensity. Rather, migration-linked
capital appears to reinforce existing urban structures, flowing preferentially
into districts where urban construction is already institutionally
concentrated. Emigration amplifies spatial differences instead of offsetting
them.
This finding helps reconcile apparently anomalous cases.
Districts with high migration exposure do not uniformly experience elevated
urban construction; outcomes depend on whether migration-linked income
interacts with established urban jurisdictions and development channels.
Migration thus conditions urban construction intensity, but does not
redistribute it spatially.
Synthesis of Models 6–7
Taken together, the income, growth, and migration models
reinforce a central conclusion: urban construction intensity in Kerala is
not driven by income growth, price dynamics, or migration in isolation.
These factors matter only insofar as they operate through and are constrained
by pre-existing spatial and institutional structures.
Urban construction accelerates not where districts are
wealthiest or fastest-growing, but where construction is already spatially
concentrated within urban jurisdictions.
Temporal Patterns: Episodic Shocks Rather Than Smooth
Trends
Year fixed effects indicate that urban construction
intensity does not evolve smoothly over time. Relative to the 2016 baseline,
construction intensity rises modestly in 2017–2018, increases sharply in 2019,
and peaks in 2020 before dropping in 2021 and partially recovering in 2022.
This pattern is consistent with commitment-driven
construction dynamics, where projects initiated under pre-pandemic
conditions continued through 2020, followed by a contraction as labour
constraints, material shortages, and uncertainty became binding. The subsequent
rebound in 2022 suggests recovery rather than a permanent structural break.
Importantly, these year effects are secondary to
district-level differences: temporal shocks matter, but they do not overturn
the dominant role of spatial structure identified in earlier models.
What the Evidence Shows (and What It Does Not)
1. Urban construction is spatially self-reinforcing
Across all specifications, the urban share of construction
emerges as the single strongest predictor of urban construction intensity. This
indicates that once construction becomes concentrated within urban
jurisdictions, subsequent construction increasingly follows the same spatial
channels.
This is not merely a scale effect. The result reflects
institutional concentration: approvals, infrastructure provision, developer
expectations, and financing mechanisms increasingly align with urban
jurisdictions once a district crosses a threshold of urban dominance.
2. Income levels do not mechanically translate into
construction intensity
After controlling for urban share, higher per-capita
district income is associated with lower, not higher, urban construction
intensity. This finding challenges the assumption that construction activity is
a simple function of economic prosperity.
Cases such as Malappuram illustrate this distinction
clearly: sustained construction activity can coexist with moderate income
levels when supported by dense settlement patterns and long-established
non-corporate building practices. Income matters—but only through the spatial
and institutional channels available for translating it into construction.
3. District identity remains consequential
District fixed effects remain significant across most
specifications, even after controlling for income, prices, migration, and urban
share. This indicates that geography, historical urban roles, and governance
regimes exert persistent influence on construction outcomes.
The implication is not that districts are idiosyncratic, but
that structural constraints differ systematically. Uniform policy prescriptions
are therefore unlikely to produce uniform outcomes.
4. Migration amplifies existing urban structures rather
than redistributing construction
Migration variables are significant only in models that omit
income controls and weaken once broader economic structure is accounted for.
This suggests that migration-linked capital does not independently drive urban
construction, but flows preferentially into districts where urban construction
is already institutionally feasible.
Migration thus acts as an amplifier, not an
equaliser: it reinforces existing spatial hierarchies rather than generating
new centres of urban growth.
5. Kerala’s urbanisation departs from standard
income-driven models
Taken together, the results indicate that Kerala’s
urbanisation cannot be understood as a linear outcome of income growth or price
dynamics. Instead, it reflects the interaction of administrative urban
boundaries, migration-linked capital, and episodic shocks operating within a
highly uneven spatial framework.
The models suggest that urban form in Kerala is shaped
less by how fast districts grow, and more by how and where construction is
institutionally organised.
Model Performance and Scope
The preferred specifications (Models 3, 4, and 7) achieve R²
values between 0.92 and 0.93, indicating that over 90 percent of the
variation in district-level urban construction intensity is explained once
urban share and fixed effects are accounted for. For a district-level panel in
the social sciences, this represents a strong model fit and suggests that the
primary structural drivers of urban construction intensity are captured by the
specification.
These high explanatory values should be interpreted
cautiously. Much of the explanatory power arises from persistent spatial
differences between districts, which is precisely the phenomenon the analysis
seeks to document and test rather than eliminate.
Limitations
This analysis is subject to several limitations.
First, the models omit potentially relevant variables such
as land prices, infrastructure investment, zoning or regulatory changes, and
project-level financing conditions, for which consistent district-level data
are not available. Second, the estimates identify associations rather than
causal effects; the results describe systematic relationships rather than
policy-invariant causal mechanisms. Third, aggregation at the district level
masks substantial intra-district variation in settlement form and construction
type. Finally, the study period (2016–2022) captures only a short time window;
longer panels would allow stronger inference on structural change and
persistence.
These limitations define the scope of the findings rather
than undermine them.
Synthesis
Taken together, the results point to a clear conclusion. Urban
construction intensity in Kerala is driven primarily by spatial concentration
rather than by income growth or price dynamics. Once construction activity
becomes institutionally concentrated within urban jurisdictions, subsequent
construction increasingly follows those channels, producing strongly
self-reinforcing patterns across districts.
Income levels and migration-linked capital matter only
conditionally. They amplify construction activity where urban structures
already exist, but do not independently generate new centres of urban
construction. As a result, a district’s position within Kerala’s urban system
matters more for construction outcomes than its aggregate economic prosperity.
This helps explain why districts with comparable incomes or
migration exposure exhibit sharply different construction trajectories: the
decisive factor is not wealth alone, but how and where construction is
organised.
Data and Methods (Brief)
- Data
source: Kerala Department of Economics & Statistics
(ecostat.kerala.gov.in)
- Unit
of analysis: District–year (14 districts, 2016–2022)
- Estimation:
OLS regression with district and year fixed effects
- Software:
R
- Models:
Seven progressive specifications testing urban concentration, prices,
income, growth dynamics, and migration
No comments:
Post a Comment