Beyond Climate Migration

Housing Precarity, Immobility, and Flood Vulnerability in Seoul

Sechang Kim

Question: How do physical flood risk and housing precarity jointly shape flood vulnerability in Seoul? How is housing precarity related to immobility?

Motivation: Personal Experience (During Master’s)

Why Seoul’s Flood Risk Matters

  • College of Education/Humanities vs. College of Law/Business?

  • Flood damage in Seoul is spatially uneven.

  • Some areas experience large and episodic damage.

  • Others experience repeated, smaller-scale or chronic damage.

  • This distinction matters because intensity & recurrence may reflect not only physical exposure, but also social and housing-market conditions.

Research Workflow

Observed and Expected Flood Exposure

Measuring Flood Damage

1) Magnitude

  • Number of damaged houses

  • Inundation area

  • Flood volume proxy: area × depth

2) Repeatability

  • Number of damage years

  • Evenness of annual damage distribution

  • Share of the most damaged year in total damage

Measuring Evenness (Shannon Entropy Index)

For each dong, annual flood damage is converted into a proportional distribution, where \(D_t\) is annual area-depth flood exposure.

\[ p_t = \frac{D_t}{\sum_t D_t},\quad H = - \sum_t p_t \log(p_t) \]

Normalized evenness (where n is the number of damaged years):

\[ E = \frac{H}{\log(n)} \]

Range: 0 - 1 (Completely Concentrated to Completely Distributed)

Measuring At-Risk Houses

  • Using 100m × 100m grid data and flood hazard map
  • Sum of houses that intersect with hazard polygon
  • Exposure to flood-risk
  • Reflects hydrological site conditions

Cluster Analysis (CA) Result: Box Plot

CA Result (Map)

Flood-damage Clusters, Seoul (k=5)

Multinomial Regression Result (1)

Stepwise multinomial logistic regression results: high-damage clusters
Outcome Predictor OR [95% CI] p
C1: High recurrent Non-housing share 1.30*** [1.16, 1.45] <.001
Single elderly proxy 1.00 [1.00, 1.00] 0.635
Foreign HH share 1.02 [0.85, 1.22] 0.868
At-risk houses 1.18** [1.05, 1.33] 0.005
C2: High episodic Non-housing share 1.28*** [1.15, 1.43] <.001
Single elderly proxy 1.00 [1.00, 1.00] 0.477
Foreign HH share 1.09 [0.91, 1.30] 0.354
At-risk houses 1.10† [0.98, 1.24] 0.099

Multinomial Regression Result (2)

Stepwise multinomial logistic regression results: lower-damage clusters
Outcome Predictor OR [95% CI] p
C3: Low recurrent Non-housing share 1.24*** [1.11, 1.39] <.001
Single elderly proxy 1.00 [1.00, 1.00] 0.931
Foreign HH share 1.02 [0.85, 1.22] 0.863
At-risk houses 1.06 [0.95, 1.18] 0.284
C4: Low episodic Non-housing share 1.22*** [1.09, 1.37] <.001
Single elderly proxy 1.00* [1.00, 1.00] 0.019
Foreign HH share 1.01 [0.84, 1.21] 0.957
At-risk houses 1.03 [0.91, 1.16] 0.657

Note: Reference category is the first level of cluster_label. Values are odds ratios with 95% confidence intervals. † p < .10, * p < .05, ** p < .01, *** p < .001. Variables were selected using stepwise AIC.

What Are “Non-Housing Residences”?

NonHous_R is central to the model results.

In the Korean housing context, “non-housing residences” often include highly precarious dwelling forms such as:

  • Jjokbang Example of jjokbang Link: goodnews1.com

  • Goshiwon

  • Shanty or informal settlements

  • Other substandard or quasi-housing arrangements

Housing Precarity and Residential Immobility

  • Jjokbang and goshiwon often require little or no deposit.
  • For extremely poor tenants, this makes them one of the few accessible housing options.
  • Monthly rate subsidy may become a de facto rent benchmark in this micro-housing market.
  • Rent can be very high relative to its unit size and facility quality.
  • Tenants have extremely limited ability to move to other housing because alternative housing requires deposits.
  • Landlords may have weak incentives to renovate because demand remains stable and returns can be high.

Key Point

Flood vulnerability may persist not simply because residents choose to stay, but because the housing market limits their ability to leave.