The repeat rent model in the presence of variable deposits

ABSTRACT

Accurate measurement of rent indices is essential in various contexts, including inflation assessment and housing market evaluation. We propose a novel repeat rent model specifically designed to accommodate variable deposits, a characteristic common in rental contracts in countries where substantial deposits are commonplace in rental contracts. To address deposit variability, we introduce time-varying deposit-rent conversion rates as additional parameters. Utilizing transaction-level data to apply this model to the rental market in Seoul, South Korea, our rent index reveals a noteworthy trend: rents began increasing in late 2019, predating the COVID-19 pandemic. This finding contrasts with other appraisal-based or survey-based rent indices, which posit that the rent surge began after the pandemic’s onset.

KEYWORDS:

• Repeat rent model
• repeat sales model
• CPI
• jeonse

JEL CLASSIFICATION:

• E31
• R21
• R31

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

¹ Note that this comparison with CPI should be interpreted with caution, given that our rent index exclusively relies on rental contracts for apartments and does not encompass other types of housing.

² Note that KREB employs a similar assumption when employing the repeat sales model (Korea Real Estate Board 2023b). Note that having access to information on the exact unit number could yield more precise estimates.

³ KREB also publishes other indices: transaction-based price indices for the multi-unit housing market, calculated using repeat sales methods (Korea Real Estate Board 2023b). This set of indices includes sales and jeonse price indices.

⁴ Case and Shiller (1989) introduce a unit-specific random walk term, which results in heteroskedasticity in the error term of the first-difference model. Specifically, the error term exhibits higher variance when observations of the same units are widely spaced apart. However, we do not explicitly incorporate this random walk term into our analysis as it introduces additional complexity to the estimation of a non-linear model. Note that the ordinary least squares estimator remains consistent regardless of the presence of the heteroskedasticity. Adams et al. (2022); Clark (2023) report that correcting for heteroskedasticity has a negligible effect on estimated rent indices.

⁵ We can substitute ${\beta }_t$ with $exp({\tilde{\beta }}_{\mathit{t}})$ to ensure the positivity condition ${\beta }_t>0$ for all $\mathit{t}$.

⁶ Outliers may arise when landlords offer rents that are substantially lower than the market rate, often due to personal relationships or other factors. In addition, from July 2020, tenants have the right to demand the renewal of their rental contracts for a two-year term, under the condition that any increase in rent and deposit is capped at 5%. This policy change led to a substantial divergence between the rents of new and renewal contracts immediately following its implementation (Hong 2022). However, such renewal contracts may not accurately reflect market rates. To detect these outliers, we employ the 3 times interquartile range (3IQR) rule, which is more lenient than the commonly-used 1.5IQR rule.