Are Move-In Ready Homes More Expensive?

Abstract

Most residential homes in developing countries such as India sell before construction is complete. Completed or move-in ready homes command a premium because of the compounded cost of capital and uncertainty costs incurred over time from holding under-construction homes. In this paper, we use listing data from India’s six largest urban agglomerations (UAs) between 2010–2012 and show that sellers expect 2–15% move-in ready premia in five UAs. Moreover, in four UAs, individuals reselling homes expect five to eight percentage points higher move-in ready premia than developers selling new homes because of additional costs incurred by individuals from holding under-construction homes. We do not find any evidence of substantial speculative gains among individual resellers. At mean listed prices, the expected move-in ready premium is 383% of an average household’s annual income in Mumbai, India’s most expensive city. Our findings indicate that within the context of a developing country, lengthy construction times and expensive capital exacerbate already poor affordability conditions.

Introduction

Figure 1 shows that 54-79% of listed homes for sale in the six largest metropolitan areas or urban agglomerations (UAs) in India during 2010-2012 were still under construction. In contrast, move-in ready homes, which include completed new homes and older resale homes, constituted 21-46% of listed homes for sale in these UAs during the same time.

Fig. 1

Move-in ready and under-construction homes by UA. Data source: Magicbricks (2014). Note: We use listing data for six UAs in India. All homes were listed during January 2010–October 2012. Darker bars indicate the percentage of under-construction homes, and lighter bars represent the percentage of move-in ready homes in a UA. Bars are labelled with the respective percentage values they represent

Homes often resell while still under construction because it takes a long time to build housing in India.¹ Gandhi et al. (2021) reports that the average time to complete a housing construction project in Mumbai is 8.5 years. One reason behind extended construction times is the difficulty in obtaining permits in Indian cities. According to a 2009 World Bank report on doing business in India, it takes on average 80 to 258 days to get construction permits across 17 Indian cities (World Bank, 2009).

Home sellers would expect buyers to pay higher prices for completed or move-in ready homes relative to ones under construction. The move-in ready premium exists because of additional hidden costs associated with under-construction home purchases.² These costs include wait-time opportunity costs, the cost of capital, and uncertainty costs, which get compounded over time from holding under-construction homes until construction completion.

There are opportunity costs to individuals holding under-construction homes until the end of construction. First-time homebuyers might be making a double payment that includes a mortgage on home loans and rent payments on their current homes. Besides, under federal law, individuals can claim tax exemption on mortgage payments only after the completion of construction of their purchased homes, thereby delaying the onset of tax benefits.³ Such costs are likely to increase with construction delays and contribute to individuals’ funding costs, and hence, the move-in ready premia expected by individuals reselling homes.⁴

On the other hand, developers incur a high cost of capital in India. This is because developers do not get formal institutional loans on land purchases, and interest rates on construction loans are between 16-18% compared to 8-10% on individual home loans.⁵ Hence, developers have the incentive to eliminate additional interest payments by financing construction using payments from early sales, at the time of initiation of construction projects, instead of costly construction loans. However, developers are usually unable to sell all homes before construction completion. Therefore, the market prices for move-in ready homes would reflect the higher cost of capital resulting from costlier construction loans.

Move-in ready home prices also include risk premia due to uncertainty costs from holding under-construction homes. Construction delays and the possibility of stalled construction are the primary sources of such risks. Construction delays occur due to cost overruns and litigation. Gandhi et al. (2021) show that litigation in Mumbai increases the average construction completion time by roughly 20%. Many developers’ lack of experience and poor management skills contribute to cost overruns (Shibani & Arumugam, 2015). Besides, developers using upfront payments from early sales to finance construction projects usually do not get the full amount of the sale price at the start of a construction project. Buyers typically pay for under-construction homes through payment plans involving installments—that are fractions of a base price—at various stages of construction (Sinha, 2013). Cost overruns and litigation risks are exacerbated by developers’ expensive capital.

Risk-loving individuals might also engage in speculative behavior, that is, purchasing new under-construction homes from developers at lower prices and reselling them when market prices increase.⁶ Figure 2a and b shows that a higher share of move-in ready homes are on resale, whereas under-construction homes are more likely to be sold new.

Fig. 2

Resale and new among move-in ready and under-construction homes by UA. Data source: Magicbricks (2014). Note: We use listing data for six UAs in India. All homes were listed during January 2010–October 2012. Darker bars in panel (a) indicate the percentage of resale among move-in ready homes, and lighter bars indicate the percentage of new for the given sample of move-in ready homes. Darker bars in panel (b) indicate the percentage of resale among under-construction homes, and lighter bars indicate the percentage of new for the given sample of under-construction homes. Bars are labelled with the respective percentage values they represent

In this paper, we use residential home listings data from a listing website covering the six largest UAs in India—Bangalore, Chennai, Delhi, Hyderabad, Kolkata, and Mumbai—between 2010-2012 and estimate sellers’ expected premia for move-in ready homes in each UA. Our key findings are three-fold. First, hedonic regressions of listed prices on home attributes reveal that the expected move-in ready premium is statistically significant in Bangalore, Chennai, Delhi, Kolkata, and Mumbai, and its magnitude varies between 2-15%. Second, individual resellers in Bangalore, Chennai, Delhi, and Hyderabad expect five to eight percentage points higher move-in ready premia than developers selling new homes. We argue that individual resellers expect higher move-in ready premia than developers selling new homes because individuals incur higher uncertainty and funding costs. Speculative behavior does not explain the residual resale premium after accounting for the move-in-ready premium. While resellers in Bangalore and Delhi expect 0.02% and 0.08% higher resale premia, respectively, with a 1% increase in home prices in the preceding year, resellers in Chennai and Mumbai expect lower real values as prices go up. Finally, for the average-priced home, the expected move-in ready premium is roughly between 24-383% of an average household’s annual income in five UAs.

Estimating the move-in ready premium is important for two reasons. First, the estimated move-in ready premium serves as a proxy for the institutional and regulatory costs borne by homebuyers. Previous literature on the costs of regulation and institutional frictions in the housing sector has mainly focused on the cost to builders and developers (Gandhi et al., 2021; Glaeser et al., 2005). However, a portion of the regulatory burden will be passed on to home buyers. An estimate of sellers’ expected premium on move-in ready homes serves as an upper bound for the direct cost of regulation borne by buyers.⁷ And second, the move-in ready premium provides us with an estimate of the decrease in housing affordability arising from prolonged construction times. Slower construction implies that the market supply of move-in ready homes is less responsive to price changes. The slower supply response effectively drives up the price of move-in ready homes.⁸

A previous body of literature has focused on the presale and forward contract markets—which are similar to under-construction home sales in this paper—in Southeast Asian countries (Chang & Ward, 1993). This literature has predominantly focused on two issues. First, there is a possibility that both buyers and sellers will renege on their contracts; buyers have the option to default, and developers can abandon construction (Chan et al., 2012). And second, under these contracts, moral hazard can lead to inferior construction quality (Chau et al., 2007). To the best of our knowledge, our paper is the first to provide a monetary value of the premium paid for move-in ready homes.

Prior academic literature has also examined home resale markets. Some papers have focused on the capitalization of additional improvements to existing buildings in resale values. Bruegge et al. (2016) show that energy efficiency affects resale home values, and He and Wu (2016) explores the capitalization of construction improvements in resale values. Another segment of this literature estimates the premium paid for the age of a home. Age is a good proxy for several home characteristics that change over time. These characteristics include depreciation, neighborhood effects, and improvements to the structure (Coulson & McMillen, 2008; Coulson et al., 2019). However, the listing data we use in this paper does not report the age of homes. This limits our ability to explain the resale premia expected by individual sellers fully.

We organize this paper as follows. In the next section, we discuss our use of listing data and what that means for the interpretability of our estimates. In section “Data”, we discuss the data used for analysis. In section “Context”, we discuss the composition and housing markets of the six Indian UAs in our sample. Section “The Expected Move-In Ready and Resale Premia” presents estimates of the expected move-in ready and resale premia. In section “Sources of the Move-In Ready and Resale Premia”, we test whether individuals or developers expect higher move-in ready premia and investigate the presence of speculative behavior among resellers. After discussing the affordability implications of the move-in ready premium in section “Implications for Affordable Housing”, we provide concluding remarks in section “Conclusion”.

A Few Notes on the Use of Listed Prices

Coefficient estimates obtained from hedonic regressions of listed prices on home attributes are not implicit prices paid for the attributes. The theoretical framework of Rosen (1974), interpreting hedonic coefficient estimates as implicit prices, applies when actual transaction data with equilibrium prices are observed. Using listed prices instead of transaction prices induces a measurement error in the dependent variable. Two major factors affecting this measurement error are the amount of time a home is listed on the market and the relative bargaining power of buyers and sellers.⁹

Since there is no prior literature on the relationship between listed and sales prices in India, we cannot ascertain the magnitude of the measurement error in listed prices and whether it correlates with the unobserved home attributes. Therefore, we interpret the hedonic coefficient estimates from regressions of listed prices on home attributes as sellers’ expected implicit prices for the attributes. Consequently, the resale and move-in ready premia estimated from our hedonic regressions can be interpreted as premia that sellers expect for these attributes. We use the terms “premia” and “expected premia” interchangeably.

There are, however, critical issues to be concerned about using the expected premia estimates obtained from listed prices as proxies for premia estimated using transaction data. First, hedonic estimates obtained using listing data can produce upward-biased estimates with large error variances (Kolbe et al., 2021). If move-in ready and resale homes are systematically listed at rates higher than transaction prices, our expected premia estimates would be upper bounds for true premia, with high standard errors. For instance, there could be hidden negative characteristics specific to move-in ready homes absent in under-construction homes, and these characteristics appear only when a buyer inspects a home in person. Then, the implicit prices for move-in ready homes relative to under-construction ones would be less than or equal to zero, ceteris paribus.

Under-construction homes might also have lower demand because of their endemic risk and remain listed longer than move-in ready homes, inducing an upward bias in the move-in ready premium. Furthermore, there could be differences in expected premia across the two seller segments in our analysis: developers and individual resellers. For instance, individuals may have less market information compared to developers. Individual resellers may also not be successful in selling listed properties, letting such listings “lapse” over time. Differences in market information and selling acumen could both explain part of the resale premia.¹⁰

To summarize our discussion, both our expected move-in ready and resale premia estimates obtained from hedonic regressions using listing data are upper bounds for true move-in ready and resale premia, respectively. Therefore, beyond interpreting our hedonic estimates as expected premia, we cannot assert that these expected premia fully translate into actual premia. Future studies could explore this issue with sales transaction data.

Having acknowledged the caveats in our data, we note a few additional points about using listed prices as proxies for transaction prices. First, prior literature has found that listed and transaction prices are highly correlated (Lyons, 2013, 2019). Hence, several transportation studies have used listed prices as proxies for transaction data (Du & Mulley, 2006; Efthymiou & Antoniou, 2013; Ghosh et al., 2023).¹¹ Second, past studies suggest that selling mispriced listed homes can take longer, increasing the costs incurred by sellers (Knight, 2002). This would be particularly true for sellers in Indian markets during 2010-2012 when internet-based property listing was still at its nascent stage. Hence, sellers would have the incentive to list homes at prices very close to market rates.

Data

We gather raw data on listed properties for sale in 11 Indian cities from Times Internet Limited, which owns the property listing website Magicbricks (Magicbricks, 2014). The data has been used by Ghosh et al. (2023) in their study of the impact of infrastructure improvements on real estate values in Bangalore. We map the 11 cities in the data to the Indian Census-designated urban agglomerations (UAs). Bangalore, Chennai, Hyderabad, Kolkata, and Mumbai are mapped to Bangalore UA, Chennai UA, Hyderabad UA, Kolkata UA, and Mumbai UA, respectively. We redefine the Census boundaries of Delhi UA to include, besides Delhi, the adjoining areas of Ghaziabad UA, Gurgaon UA, the municipal corporation of Faridabad, and the towns of Noida and Greater Noida (see section “Context” for details on UA boundaries).

The raw data contains details on 5.98 million residential and commercial properties listed between January 2007–October 2012. Among these were 1.37 million residential homes for sale.¹² Residential homes include multistory apartments, “builder floor apartments,” independent houses, and villas.¹³ We exclude data before 2010 from the regression analysis when information on move-in ready and under-construction homes was not collected. The data on listed residential homes contain details on the listed price, listing date, and a collection of hedonic attributes. These hedonic attributes include the number of bedrooms and bathrooms, built-up area or square footage, name of the home’s locality or neighborhood, whether the home was resale or new, and whether the home was move-in ready or under-construction. New listed homes are defined as those on sale for the first time and are sold only by developers. Resale homes have at least one prior transaction and are sold only by individuals. Based on these definitions, both under-construction and move-in ready homes can be listed as new (by developers) and resale (by individuals). Figure 2 provides a breakdown of the shares of new and resale homes that are move-in ready and under-construction.

The data contains information on 25 additional housing attributes, such as the presence of elevators, water storage, gymnasiums, air conditioning, etc. The reporting on these attributes is, unfortunately, incomplete. Therefore, we place these attributes into frequency-of-reporting bins—the five attributes reported most frequently are placed into one bin, the five reported second most frequently into another, and so forth. These bins translate into five dummy variables, which are then assigned to each home based on whether at least one attribute within a bin is reported. The composition of bins and details on the cutoff frequency for each bin are given in Table 1.

Table 1 Categorical attribute bins

We use the reported neighborhood names of each listed home to generate geographical coordinates. The neighborhood names are colloquial references to regions that do not necessarily have well-defined geographical or administrative boundaries. These neighborhoods are somewhat analogous to Census tracts in the United States (US) in terms of their covered geographical area and population, albeit with a lot of variation both within and across UAs. A detailed discussion of the land area, population, and composition of neighborhoods by each UA is given in section “Context”. We use the Google Maps Application Programming Interface (API) to retrieve the geographical coordinates for the centroid of each neighborhood. All homes in a neighborhood are assigned the geocodes of the centroid of the neighborhood.¹⁴ The generated geocodes for each home are then used to calculate the minimum geodesic distance of a home to a Central Business District (CBD) and a Secondary Business District (SBD).¹⁵ We use the distance of a listed home to the closest Business District as a covariate in the hedonic regressions.

The reported listing prices in the data are inflation-adjusted to 2001 values using the industrial worker Consumer Price Index (CPI) series for each UA obtained from the Labor Bureau of India (2012).¹⁶ The values of the real price (referred to interchangeably as both the price and the real price) contain many outliers. We identify outliers from each UA’s unconditional distribution of the absolute price and the price per square foot of built-up area, per bedroom, and per bathroom. We first remove the top and bottom five percentile values of per square feet, per bedroom, and per bathroom prices. In the remaining sample that satisfies this cutoff, we truncate the absolute price and built-up area values at the top and bottom five percentile values. We do not further truncate the values for bedrooms and bathrooms because these are already top-coded to ten in the raw data.¹⁷ The final data for analysis contain 221,783 listed residential homes on sale in the six UAs, combined, listed from January 2010 to October 2012. Detailed summary statistics for all listed home prices and attributes are given in Table 2.

Table 2 Summary statistics of listed home prices and attributes by UA

Context

In this section, we provide some details on the composition and nature of the six urban agglomerations (UAs) in our sample. We also discuss the residential housing markets in the UAs.

Urban Agglomerations and their Neighborhoods

The Census of India defines an urban agglomeration (UA) as “a continuous urban spread constituting a town and its adjoining outgrowths, or two or more physically contiguous towns together with or without outgrowths of such towns.” In other words, UAs are conurbations composed of one or more municipalities, towns, and adjoining areas. In 2011, there were 468 UAs with a population of 100,000 or more across India. Fifty-two had a population of at least one million people (Census of India, 2011).

The cities of Bangalore, Chennai, Hyderabad, Kolkata, and Mumbai are mapped to Bangalore UA, Chennai UA, Hyderabad UA, Kolkata UA, and Mumbai UA, respectively. In addition to the National Capital Territory (NCT) of Delhi, the surrounding areas of Ghaziabad UA, Gurgaon UA, the municipal corporation of Faridabad, and the Census towns of Noida and Greater Noida form part of one large urban cluster.¹⁸ Even though the Census of India does not recognize the five latter cities as part of the Delhi UA, these regions are very well-connected to the NCT of Delhi and with each other through extensive road networks and the Delhi metro. We will treat the redefined Delhi UA, including the above-mentioned areas, as one labor and housing market.¹⁹

The UAs comprising our listing data are the six most populous in India. Table 3 provides the population of each UA. Hyderabad is the least populous, with about 7.7 million people, and Delhi is the most populous, with nearly 22 million people. Except for Bangalore, the other five UAs in our sample draw urban areas from multiple districts—administrative divisions similar to counties in the US. Table 3 provides the number of districts comprising each UA.²⁰

Table 3 Characteristics of UAs

Sellers of each listed home report the colloquially used neighborhood name to reference the UA’s region in which a home is located. Table 3 indicates substantial heterogeneity in the number and density of neighborhoods across UAs. Assuming that neighborhoods are evenly spread over a UA, the average land area per neighborhood is roughly three square kilometers in Bangalore and Delhi, but more than seven square kilometers in Kolkata and Mumbai. The density of homes per neighborhood is also the highest in Mumbai. This is partly because Mumbai has very few reported neighborhoods. We use reported neighborhood names to generate geographical coordinates for the centroid of each neighborhood and assign the geocodes to homes listed in the neighborhood. The generated geocodes are then used to calculate the distance to the closest Business District.

We should note two important points here. First, since these neighborhoods are regions without any official spatial boundary, we do not know the exact spatial extent of a given neighborhood. This means that the consistency and spread of neighborhoods are unknown. Second, neighborhoods are mapped to the geographical boundaries of districts comprising a UA, not the UA itself.²¹ This implies that some homes may be listed in areas very close to a UA but not strictly within the confines of the Census-designated area of a UA. However, these problems are unlikely to cause major issues in our empirical strategy since the neighborhoods are only used to calculate the distance to CBD/SBD and as dummy variables to account for neighborhood-specific characteristics.

Housing Markets and Institutions

There is substantial variation in the distribution of prices across the six UAs. While Bangalore, Chennai, Hyderabad, and Kolkata are somewhat similar, with listed prices between 2.2-2.6 million rupees, Delhi and Mumbai are at least twice as expensive (see Table 2).²² With unconditional mean prices at seven million rupees in 2001 values, Mumbai is more than three times as expensive as Kolkata. Both Delhi and Mumbai also have a higher dispersion of prices (see Table 2). This suggests substantial variation even within the residential housing markets of Delhi and Mumbai. Finally, the number of observations and the share of resale homes is noticeably higher in Delhi and Mumbai compared to the other UAs (see Table 2 and Fig. 1). Assuming that the listing website was equally used in the UAs during 2010-2012, a higher share of resale homes would indicate that the markets in Delhi and Mumbai were relatively thicker.

Real estate and land governance are under the jurisdiction of the individual states of India. The UAs in our sample belong to different states.²³ Hence, the institutions and the regulatory environment are also different across the six UAs. Take the example of the most ubiquitous building regulation imposed in Indian cities—floor area ratio (FAR) limits.²⁴ While cities in Maharashtra, of which Mumbai is the capital, allowed on average an FAR of 1.3 in our study period, Delhi allowed for an FAR of 3.5 (Sridhar, 2010). Dutta et al. (2021) show that there is considerable inter-state variation in housing supply elasticities across India. The differences in regulatory and administrative setup across the UAs are our main motivation behind estimating each UA’s move-in ready and resale premia separately.

The Expected Move-In Ready and Resale Premia

In this section, we estimate sellers’ average expected move-in ready and resale premia. We hypothesize that sellers expect a premium from selling move-in ready homes because of the compounded cost of capital and risk premium from holding under-construction homes until the end of construction. Recall that, by our definition, resale homes are only listed by individuals as opposed to new homes that are listed only by developers. We, therefore, attribute the resale premia to individual resellers’ potentially higher costs from holding under-construction homes and possible speculative behavior, issues we explore further in section “Sources of the Move-In Ready and Resale Premia”.

We use the home listing data described in section “Data’ to run hedonic regressions of log listed prices on a collection of home attributes. The hedonic attributes include, among other things, a dummy variable representing move-in ready homes and another dummy for resale homes. The base categories for move-in ready and resale are under-construction and new homes, respectively. We run the following hedonic regression for each UA:

$$\text{log}\left({Price}_{j}\right)={\beta }_{0}+{\beta }_{1}{Ready}_{j}+{\beta }_{2}{Resale}_{j}+{\beta }_{3}{X}_{j}+{\lambda }_{n}{\phi }_{t}+{\varepsilon }_{j}$$

(1)

where ${Ready}_{j}$ and ${Resale}_{j}$ are dummy variables equal to one if home $j$ is resale and move-in ready, respectively; ${Price}_{j}$ is the listed price of home $j$; ${X}_{j}$ is a vector of additional hedonic attributes that include the number of bedrooms and bathrooms, log of built-up area or square footage, a dummy for independent house (similar to single-family homes), minimum geodesic distance to a CBD or an SBD, and dummy variables for five categorical attribute bins (defined in Table 1) representing the presence of 25 additional amenities. The error term is ${\varepsilon }_{j}$.

Both the move-in ready and the resale premia could be explained by unobserved neighborhood characteristics not captured by the observed hedonic attributes and the geodesic distance to CBDs/SBDs. If we imagine the geographical shape of a UA to be a circle, even with the same geodesic distance to the center (or CBD), two neighborhoods could be located on two different sides of the circle. Moreover, neighborhood characteristics could change over time. Hence, we include interactions of neighborhood dummies ${\lambda }_{n}$ and quarter-of-listing dummies ${\phi }_{t}$ in Eq. (1) to account for changes in neighborhoods over time.

The Least Square Dummy Variable (LSDV) estimates of Equation (1) are given in panel (a) of Table 4. The expected move-in ready premia are positive and significant at the 99% level in Bangalore, Delhi, Kolkata, and Mumbai; it is positive but marginally significant at 90% in Chennai and insignificant in Hyderabad. The magnitude of the premium varies considerably across UAs. Among the five cities with a significant premium, sellers in Chennai and Kolkata expect the lowest move-in ready premia with values of 1.8% and 3.7%, respectively. In Bangalore, Delhi, and Mumbai, sellers expect at least 10% premia from selling move-in ready homes. Sellers in Bangalore expect the highest premia at 15%. We also observe that individual resellers expect a resale premium everywhere but in Delhi. The magnitude of the resale premia varies between 3-13%. These results are somewhat consistent with discussions among academics, policymakers, and realtors in India, suggesting that individual investors might be engaging in speculative-type behavior in cities such as Bangalore and Delhi.²⁵

Table 4 Expected move-in ready and resale premia by UA

In equation (1), the interactions of neighborhood dummies and quarter-of-listing dummies capture changes in neighborhood-specific characteristics over time. Since changes in neighborhood-level prices can potentially account for changes in unobservables, we run regressions with quarterly mean neighborhood price lags and compare the estimates with those obtained from equation (1). We rewrite equation (1) as follows:

$$\text{log}({Price}_{j})= {\alpha }_{0}+ {\alpha }_{1}{Ready}_{j}+ {\alpha }_{2}{Resale}_{j}+ {\alpha }_{3}{X}_{j}+ \sum_{q=1}^{4}{\rho }_{q}{Z}_{q}+ {\nu }_{j}$$

(2)

where ${Price}_{j}$, ${Ready}_{j}$, ${Resale}_{j}$, and ${X}_{j}$ are as defined before; ${Z}_{q}$ represents a vector of four neighborhood-level mean price lag variables for the four quarters preceding the quarter of listing of a home; ${\rho }_{q}$ represents the corresponding coefficients for each quarterly price lag variable. We define ${Z}_{1}$ as the mean neighborhood price for the quarter immediately before the quarter of listing, ${Z}_{2}$ as the mean price for the quarter before that, and so on. The error term is ${\nu }_{j}$. Unlike in Eq. (1), here we cannot include the neighborhood dummies interacted with quarterly dummies because the interactions terms will be collinear with the quarterly price lag terms.

The Ordinary Least Squares (OLS) estimates of equation (2) are given in panel (b) of Table 4. We note some changes in these estimates by comparing them to panel (a) figures. First, the move-in ready premium estimate is 4.2% in Chennai—more than twice the magnitude of the estimate in panel (a)—and is significant at 99%. Second, the estimates for Delhi and Kolkata are around twice the magnitude of those obtained in panel (a). The move-in ready premia for Bangalore and Chennai are quite similar in the two specifications. The resale premium changed similarly. Chennai’s estimate more than doubles in panel (b), and the estimates for Hyderabad and Mumbai increase slightly. While the resale premium for Kolkata declines in the second specification, Bangalore’s estimate remains identical.

The generally higher magnitude of the premia estimates in panel (b) reflects the fact that while neighborhood- level price lags capture most neighborhood-specific changes that occurred in the past, it is likely that the price lag terms will not account for neighborhood characteristics in the quarter of listing. We use the estimates in panel (a) of Table 4 to calculate rupee values of the move-in ready premia in section “Implications for Affordable Housing”. In section “Sources of the Move-In Ready and Resale Premia”, we use variations of the specification given in equation (2) to investigate individual resellers’ additional expected move-in ready premia and speculative behavior-induced gains.

Sources of the Move-In Ready and Resale Premia

In this section, we examine the sources of the move-in ready and resale premia. First, we identify whether developers or individual resellers expect higher move-in ready premia and explain why differences in expected premia may exist. Next, we attempt to test for the presence of speculative behavior among individual resellers to explain the resale premia.

Developers Versus Resellers

To test whether individual resellers expect higher move-in ready premia than developers selling new homes, we include an interaction of the ready and resale dummy variables in equation (2) and rewrite it as follows:

$$\text{log}({Price}_{j})= {\gamma }_{0}+ {\gamma }_{1}{Ready}_{j}\times {Resale}_{j}+ {\gamma }_{2}{Ready}_{j}+ {\gamma }_{3}{Resale}_{j}+ \sum_{q=1}^{4}{\theta }_{q}{Z}_{q}+ {\gamma }_{4}{X}_{j}+ {\epsilon }_{j}$$

(3)

where ${Price}_{j}$, ${Ready}_{j}$, ${Resale}_{j}$, ${Z}_{q}$, and ${X}_{j}$ are as defined in Eq. (2). The error term is ${\epsilon }_{j}$. We cannot include the neighborhood dummies interacted with quarterly dummies, as in Eq. (1), because the quarterly price lag terms are collinear with the neighborhood dummies.

Since, in addition to their interaction, we include the ${Ready}_{j}$ and ${Resale}_{j}$ dummy variables individually in Eq. (3), we can break down the coefficient ${\gamma }_{1}$ as follows:

$$\begin{array}{c}\gamma_i=\left\{E\left(\log\left(Price_j\right)\vert Ready_j=1,\;Resale_j=1\right)-E\left(\log\left(Price_j\right)\vert Ready_j=0,\;Resale_j=1\right)\right\}-\\Move-in\;ready\;premia\;among\;resale\;\left(individual-sold\right)\;homes\\\left\{E\left(\log\left(Price_j\right)\vert Ready_j=1,\;Resale_j=1\right)-E\left(\log\left(Price_j\right)\vert Ready_j=0,\;Resale_j=1\right)\right\}\\Move-in\;ready\;premia\;among\;new\;\left(developer-sold\right)\;homes\end{array}$$

(4)

Equation (4) shows that the coefficient ${\gamma }_{1}$ is the additional move-in ready premia expected by individual resellers over and above what developers expect from selling new homes.

We run regressions using Eq. (3) with our listing data for each UA separately and present the key coefficient estimates in Table 5. We note a number of critical results here. First, the coefficient estimates of ${\gamma }_{1}$ on the interaction of ${Ready}_{j}$ and ${Resale}_{j}$ are significant and positive in Bangalore, Chennai, Delhi, and Hyderabad. The estimates imply that individual resellers expect five to eight percentage points higher move-in ready premia than developers selling new homes in these UAs. However, for Kolkata and Mumbai, the estimate of ${\gamma }_{1}$ is insignificant, indicating that resellers expect the same move-in ready premia as developers in these two cities. Note that the estimate of ${\gamma }_{1}$ is positive and marginally significant for Hyderabad, even though the estimated move-in ready premia for Hyderabad in panels (a) and (b) of Table 4 are insignificant. This may be explained by resellers’ positive expected move-in ready premia cancelling developers’ negative expected premia in Hyderabad, implying an overall average move-in ready premium close to zero.

Table 5 Expected move-in ready premium among resale homes by UA

Why do Individual Resellers Expect Higher Move-in Ready Premia?

The estimates of ${\gamma }_{1}$ are not negative in any of the UAs in our sample, implying that individual resellers expect at least as much move-in ready premia as developers. To examine why this might be the case, we employ accounting identities and explore the implications of risk premia and funding costs embedded in the move-in ready premia. For the sake of simplicity, we assume there are two periods: (i) the under-construction phase denoted with superscript $U$, which includes the entire period of time from the beginning of a project until completion, and (ii) the completed or move-in ready phase denoted with superscript $R$, when only completed homes are listed for sale.

Developers initiate construction projects, start selling at the beginning of construction, and continue selling until completion. Resellers purchase homes only at the beginning of construction and have the option to sell either during the construction phase or after completion. We denote developers and individuals with subscripts $d$ and $i$, respectively. Assuming markets are competitive, under-construction homes would sell for an equilibrium price of ${P}^{U}$, regardless of who sells it.²⁶

Developers list move-in ready properties for a price of ${P}_{d}^{R}= {P}^{U}(1+ {\kappa }_{d})(1+ {\eta }_{d})$. Here, ${\kappa }_{d}$ and ${\eta }_{d}$ are, respectively, developers’ risk premia from holding under-construction homes and their construction funding cost.²⁷ Developers’ expected move-in ready premia can then be written as follows:

$${\pi }_{d}={{P}_{d}^{R}- {P}^{U}= P}^{U}\left\{\left(1+ {\kappa }_{d}\right)\left(1+ {\eta }_{d}\right)-1\right\}$$

(5)

On the other hand, individual resellers incur funding cost ${\eta }_{i}$ and their risk premia from holding under-construction homes is ${\kappa }_{i}$.²⁸ Then, the move-in ready price listed by individual resellers would be ${P}_{i}^{R}= {P}^{U}(1+ {\kappa }_{i})(1+ {\eta }_{i})$.²⁹ We obtain individuals’ expected move-in ready premia as follows:

$${\pi }_{i}={{P}_{i}^{R}- {P}^{U}= P}^{U}\left\{\left(1+ {\kappa }_{i}\right)\left(1+ {\eta }_{i}\right)-1\right\}$$

(6)

If we assume that developers’ and individual resellers’ funding costs and risk premia are non-negative, their expected move-in ready premia will be unambiguously non-negative. If any of the costs ($\eta$ or $\kappa$) incurred by developers or individuals are positive, a move-in ready premium will exist on average. In other words, move-in ready homes are inevitably priced higher than under-construction homes since there are non-zero costs from holding under-construction homes that get compounded over time.

The difference between the move-in ready premia expected by individual resellers and developers can be obtained by subtracting equation (5) from equation (6), as follows:

$$\Delta \pi ={P}^{U}\left\{\Delta \kappa \left(1+{\eta }_{d}\right)+\Delta \eta \left(1+{\kappa }_{i}\right)\right\}$$

(7)

where the terms $\Delta \kappa ={\kappa }_{i}-{\kappa }_{d}$ and $\Delta \eta ={\eta }_{i}-{\eta }_{d}$ represent the risk premia and funding cost differences between individual resellers and developers. Note that $\Delta \pi$ is equivalent to the coefficient ${\gamma }_{1}$ on the interaction term ${Ready}_{j}\times {Resale}_{j}$ in Eq. (3).

Equation (7) shows that the difference between developers’ and individual resellers’ expected move-in ready premia is a function of the differences between their funding costs and risk premia. The expected move-in ready premia are higher for the segment of sellers that incur higher uncertainty and funding costs. In contrast, if only the funding or uncertainty costs are higher for one set of sellers, $\Delta \pi$ could go in either direction.³⁰ The fact that we obtain mostly positive (and some zero) estimates of $\Delta \pi$ (or ${\gamma }_{1}$) in our regression results shown in Table 5 indicates that individual resellers incur at least as much overall compounded uncertainty and funding costs as developers from holding under-construction homes. In the extreme case, individuals incur higher uncertainty and funding costs. In other cases, individuals incur either higher uncertainty or higher funding costs, and the total costs incurred by individuals outweigh developers’ costs.³¹

Note that there is likely to be considerable heterogeneity in move-in ready homes resold by individuals. Specifically, we do not observe whether resold move-in ready homes are “fresh” or “lived-in.” Ideally, we would have estimated individuals’ move-in ready premia from reselling only the former. “Lived-in” properties are depreciated and, therefore, command lower values than “fresh” properties. Coulson et al. (2019) show that prices of used properties are substantially lower than prices of unused properties. So, a mix of “fresh” and “lived-in” homes in our data implies that the estimated individual resellers’ expected move-in ready premia are biased downward. We, therefore, underestimate the additional move-in ready premia expected by individual resellers.³²

Resellers’ Speculative Behavior

To examine whether individuals reselling homes engage in speculative behavior, that is, they resell homes when market prices increase, we include interactions between the resale dummy and four quarterly mean neighborhood price lags in equation (3) and rewrite it as follows:

$$\text{log}({Price}_{j})= {\delta }_{0}+ {\delta }_{1}{Ready}_{j}\times {Resale}_{j}+ \sum_{q=1}^{4}{\mu }_{q}{Z}_{q} \times {Resale}_{j}++ {\delta }_{2}{Ready}_{j}+ \sum_{q=1}^{4}{\sigma }_{q}{Z}_{q}+ {\gamma }_{3}{X}_{j}+ {\nu }_{j}$$

(8)

where ${Price}_{j}$, ${Ready}_{j}$, ${Resale}_{j}$, ${Z}_{q}$, and ${X}_{j}$ are as defined in Eq. (2). The error term is ${\nu }_{j}$. We cannot include the resale dummy because of collinearity between the interaction terms, ${Z}_{q} \times {Resale}_{j}$, and ${Resale}_{j}$. This means there may be a residual resale premium in Eq. (8) that is not accounted for by speculation. We are also unable to include the neighborhood dummies interacted with quarterly dummies because the quarterly price lag terms are collinear with the neighborhood dummies.

The coefficients ${\mu }_{q}$ in Eq. (8) isolate the impact of changing market prices in the preceding four quarters on the listing price of a resale home. In other words, if ${\mu }_{q}$ does not equal 0, then individual resellers were possibly engaging in speculative behavior; a positive sign on ${\mu }_{q}$ would indicate possible gains from speculation, whereas a negative sign would mean individuals might have incurred losses from reselling homes.

Note that non-zero estimates of ${\mu }_{q}$ need not necessarily imply that individuals were actually gaining or losing from reselling homes. This is because the quarterly price lag terms capture changing market conditions. Since we do not have data on past transaction prices of resale homes, we cannot determine with certainty whether increasing market prices translate into equivalent property-level speculative gains or losses for individual resellers. We can only make a suggestive argument that reselling homes as markets turn bullish indicates the existence of speculative behavior.

We run regressions using equation (8) with our listing data and present the main results in Table 6. First, we find that none of the interaction terms of the quarterly price lags and the resale dummy are significant in Hyderabad and Kolkata, indicating that resellers in these two cities do not make speculative gains or losses. Second, we see that resellers in Bangalore, Chennai, Delhi, and Mumbai expect non-zero resale premia as prices in the preceding four quarters increase, hinting at the possible existence of speculative behavior in these four cities.

Table 6 Move-in ready premium among resale homes and speculative gains by UA

The interaction of the resale dummy with the fourth quarter price lag is significant and positive in Bangalore and significant and negative in Mumbai. The interaction of the resale dummy with the third-quarter price lag is significant and positive in Delhi and Mumbai. The interaction of the resale dummy with the second-quarter price lag is significant and positive in Bangalore. Finally, the interaction of the resale dummy with the first quarter price lag is significant and negative in Bangalore, Chennai, and Mumbai. Only two significant estimates on the price lag and resale interaction terms are significant at 99%, and another two at 95%.

When we add the coefficients for all quarterly price lag terms interacted with the resale dummy, we find that individual resellers in Bangalore and Delhi expect to gain 0.02% and 0.08%, respectively, in higher resale prices when market prices increase by 1% in the preceding four quarters. In contrast, resellers in Chennai and Mumbai expect 0.1% and 0.08%, respectively, in lower resale values when prices go up by 1% in the previous four quarters. The higher resale premia resulting from quarterly price lag growth represents a small fraction of the overall price growth in Bangalore and Delhi. In Chennai and Mumbai, resellers seem to be making losses in real terms. Hence, we conclude that even if there is speculative behavior, resellers do not gain substantially.

Because we do not include the resale dummy in the regressions for Table 6, the interactions between the four quarterly price lags and the resale dummy will not necessarily account for the entire amount of the resale premium. In fact, when we add the coefficients of the interactions of the resale dummy and the price lags in Table 6 and compare the resultants with the estimates of the resale dummy in Table 5, we observe a large difference in the two sets of estimates.³³ We attribute the residual resale premium to factors, such as the age of a home, unaccounted hidden attributes, unobserved changes in neighborhood characteristics, etc.

Note that the decision to include four quarterly price lag terms in our specifications in equations (2), (3) and (8) is based on the idea that individuals would list homes soon after market prices increase. Indian real estate institutions slow the transmission from bullish expectations to house prices. Selling homes is a lengthy process in India that involves the transfer of several titles, permits, and property rights.³⁴ This means that individuals engaging in speculative behavior will not wait long to list homes once market prices increase because finding buyers takes time. We also test whether price changes from more than a year back affect resale values by including six and eight quarterly lags and their interactions with the resale dummy in equation (8). The results of these regressions are given in Appendix Tables 12 and 13. We do not find any evidence of systematic speculative gains from these regressions.

Implications for Affordable Housing

Mumbai is the most expensive UA in our sample, with an average price of seven million rupees during 2010-2012 in 2001 values—equivalent to 700,000 US dollars (USD) at 2001 PPP-adjusted exchange rates.³⁵ At the peak of the real estate boom of 2006-07, the average house price in Los Angeles County was around half a million USD, not adjusting for inflation (Streitfeld, 2007). These figures, implying that Mumbai has one of the world’s most expensive real estate markets, are consistent with prior literature (Chakravorty, 2013; Nijman, 2000). Even though relatively cheaper, Delhi’s real estate market is close behind, with the average home priced at five million rupees or 500,000 USD in PPP-adjusted 2001 values.

At the same time, India has one of the largest slum populations in the world, with about 64 million people— larger than the combined populations of California and Texas or about the population of the United Kingdom in 2011—living in informal houses without adequate property rights (Census of India, 2011). One in six individuals in Indian cities lives in informal houses made of thatch, mud, bamboo, etc. (Dutta et al., 2021). The stark contrast in having some of the most expensive formal housing markets and a large share of the urban population living in informal houses underscores the need to study housing affordability in India.

We examine the implications of the move-in ready premium for housing affordability in India’s six largest UAs. We use the coefficient estimates presented in panel (a) of Table 4 to gain insight into how large the move-in ready premia are relative to household incomes in the six UAs. We use estimates of per capita annual income in our sample of UAs in 2014 from a Brookings Institution Report (Parilla et al., 2015). We also use household size figures from the Census of India (2011) to convert the per capita income into household-level income in each UA. We then obtain the rupee value of the move-in ready premium and calculate the share of an average household’s annual income it represents in each UA. The figures are given in Table 7.

Table 7 Move-in ready premium as percentage of household income by UA

The average move-in ready premia are particularly high in Mumbai, Delhi, and Bangalore, with values of 0.81 million, 0.49 million, and 0.36 million rupees, respectively. While in Bangalore and Delhi, the move-in ready premia are 268% and 103% of respective average household-level annual incomes, in Mumbai, it is 383%! In Chennai and Kolkata, the move-in ready premia are 24% and 67% of average household incomes, respectively. Since the move-in ready coefficient is insignificant for Hyderabad in all regression specifications, Hyderabad’s move-in ready premium is zero. These numbers provide an approximate estimate of the affordability constraint imposed by institutional frictions, such as lengthy construction times and expensive capital for developers.

Conclusion

In this paper, we gather data on 221,783 residential home listings in six Indian UAs and estimate the average move-in ready premium expected by sellers in each UA. We find that sellers in five UAs—Bangalore, Chennai, Delhi, Kolkata, and Mumbai—expect move-in ready premia of 2-15%. We also find that individual resellers expect five to eight percentage points higher move-in ready premia than developers selling new homes, implying that individuals incur higher costs from holding under-construction homes. The expected move-in ready premia are between 24-383% of average household-level annual incomes in five UAs.

Our research opens up avenues for future work in a couple of ways. First, we use listed prices instead of transaction prices in the hedonic regressions. To err on the side of caution, we interpret the hedonic coefficient estimates as sellers’ expected implicit prices for the attributes. However, the question remains as to whether the expected move-in ready and resale premia actually translate into higher transaction prices for buyers. Better substitutes for transaction prices can allow future researchers to probe deeper into this question. Of course, actual transaction prices would be ideal, but at this point, they are not recorded consistently by any administration in India. Finally, with data on attributes such as the age of a home, the exact time of transfer of property rights and ownership to buyers, exact geolocations, etc., future research can investigate the change in the premia paid for resale homes during various construction phases.

Appendix

Table 8 Expected move-in ready and resale premia by UA with neighborhood and quarterly dummies interacted and all covariates

Table 9 Expected move-in ready and resale premia by UA with quarterly price lags and other covariates

Table 10 Expected move-in ready premium among resale homes by UA with all covariates

Table 11 Speculative gains based on prices in the past four quarters by UA with all covariates

Table 12 Speculative gains based on prices in the past six quarters by UA with all covariates

Table 13 Speculative gains based on prices in the past eight quarters by UA with all covariates