Signalling Creditworthiness: Land Titles, Banking Practices, and Formal Credit In Indonesia

Abstract

Many land titling programs worldwide have produced lacklustre results in terms of achieving access to credit for the poor. This may reflect insufficient emphasis on local banking practices. Bankers commonly seek to ensure repayment by using methods other than securing collateral, such as targeting borrower characteristics that, on average, improve repayment rates. Formal land titles can signal these important characteristics to the bank. Using a household survey from Indonesia, we provide evidence that formal land titles have a positive and significant effect on access to credit and that at least part of this effect is best interpreted as an improvement in information flows. These results stand in contrast to the prevailing notion that land titles function only as collateral. Analysts who neglect local banking practices may misinterpret the observed effect of systematic land titling programs on credit access because these programs tend to reduce the signalling value of formal land titles.

Banyak program pemberian hak atas tanah di seluruh dunia berakhir buruk dalam kaitannya dengan penerimaan pinjaman kredit bagi masyarakat miskin. Hal ini bisa merefleksikan tidak memadainya praktik-praktik perbankan di daerah. Para bankir umumnya mencari kepastian atas pembayaran kembali melalui cara-cara selain penggunaan jaminan. Ini misalnya dilakukan dengan menargetkan peminjam berdasarkan karakteristik, yang secara rata-rata biasanya berhasil menaikkan tingkat pembayaran kembali. Adanya pemberian hak formal atas tanah dapat memberikan sinyal ini kepada bank. Dengan menggunakan survei rumah tangga dari Indonesia, para penulis membuktikan bahwa pemberian hak formal atas tanah memiliki efek yang positif dan signifikan kepada akses terhadap pinjaman. Sebagian dari efek ini dapat diinterpretasikan sebagai perbaikan dalam arus informasi. Hasil-hasil ini sangat kontras jika dibandingkan dengan studi-studi sebelumnya, yakni bahwa hak atas tanah berfungsi hanya sebagai jaminan. Para analis yang menafikan praktik perbankan lokal mungkin akan menyalahartikan efek yang diobservasi dari program sertifikasi tanah yang sistematis terhadap akses pinjaman, karena program-program demikian cenderung mengurangi nilai yang tersirat dari pemberian hak formal atas tanah.

Keywords:

• land title
• credit

JEL classification:

• O12
• O17
• Q15

We are grateful to a number of people for providing us with feedback about this project. We would like to thank Bill Easterly, Raquel Fernández, Daniel Gilligan, Sergei Guriev, Jonathan Morduch, Debraj Ray, Mario Rizzo, Sergei Stepanov, and Katia Zhuravskaya. We are also grateful to the Social Science Research Council for providing funding for the fieldwork discussed in this article and to Bank Rakyat Indonesia for facilitating our meetings with local bankers. Castañeda Dower acknowledges the support of the Ministry of Education and Science of the Russian Federation, grant No. 14.U04.31.0002, administered through the New Economic School's Center for the Study of Diversity and Social Interactions. Any errors that remain are our own.

Notes

¹ A title applicant usually requires a letter from the village head, verifying that the land is in the applicant's possession. The applicant then needs to have a survey conducted, which involves funding the boundary markers, the survey fee, and all transportation costs. The document must then be verified, mapped, and certified. In total, the process can easily take one year. Once the applicant has the certificate, they must pay a one-off tax on the right to have a title on a piece of land. (This is not a property tax or a tax on the sale.) Anecdotal evidence reveals that applicants can also accumulate significant informal costs while obtaining a title. For example, the stated fee of a land certificate is around Rp 300,000 (approximately $33 in 2002). Yet when we asked our survey respondents what the actual fee was, their answers ranged from Rp 1 million to Rp 2 million (approximately $111 to $222 in 2002).

² Even if the land title has been inherited, possession of it increases the incentive to learn how to interact in the formal sector.

³ The other 60% is made up mainly of salary guarantees, but it also includes other informal land documents, vehicle ownership certificates, or even no security at all. Of all first-time borrowers with a land title, 42% offered it as collateral. The relatively low usage of land title to secure a loan could reflect the poor state of land registration overall. Since the 2002 MASS, the Indonesian government has taken initiatives, such as Regulation 24/2007, to improve land titling (Bappenas 2012). With these initiatives, the use of land title for securing a loan could have changed. However, a recent World Bank report evaluating its latest land administration project, lasting from 2000 to 2009, finds mixed results (World Bank 2014), indicating that there is room for improvement and suggesting that our findings are still valid in 2014.

⁴ McLeod (1991) also argues that expanding credit access as a firm evolves is a socially optimal strategy for lenders, owing to the risk that first-time borrowers present.

⁵ Other factors can arise that undermine the possibility of a legal transfer if the borrower defaults: weak legal infrastructure, political pressure, and a thin land market. See Domeher and Abdulai's (2012) study for a discussion of why land registration may not have a positive impact on credit access.

⁶ On a more positive note, this argument could suggest that even non-transferable rights of exclusion to land might positively influence credit access if the process for receiving these documents is not automatic. These types of rights are often granted instead of fully transferable land rights when governments are worried that the formalisation of property rights will make it more likely for small-scale farmers to lose their land. This practice has been used in India.

⁷ Kochar (1997) argues that the existence of informal credit markets may cause the empirical data to misrepresent the extent of credit rationing: institutional credit may be accessed less, because individuals’ demand for credit may be satisfied by the informal sector. Moreover, McLeod (1991) points out that credit constraints are actually the reverse of what is often assumed—that is, it is institutional credit that lacks access to small-scale borrowers, because institutions often cannot compete with informal sources of credit.

⁸ On a relative scale, institutions that focus on microlending were affected less drastically by the 1997–98 Asian financial crisis than those that did not. In fact, BRI units were profitable throughout the crisis and it is reported that BRI units subsidised the branch banks during and after the crisis (Patten, Rosengard, and Johnston 2001; Rosengard et al. 2007).

⁹ BRI has been studied in the microfinance literature. For more information on the history and practices of BRI, see Maurer's (1999) study.

¹⁰ See Morduch's (1999) study for a thorough review of these methods, which are commonly used in the microfinance world.

¹¹ Consider, for example, the following method, which is often used: the loan applicant fills out a description of the condition of their assets, and then the BRI unit manager sends a bank employee to view the assets in order to make a comparison. What the bank cares about is not just the value of assets but also whether the description of the assets was honest.

¹² This number seems large, leading us to believe that managers have included threats to foreclose. Formal foreclosure is rare in Indonesia's small-scale loan market. However, just because foreclosure is not observed does not mean that collateral is not at work. A simple game-theoretic framework yields a Nash equilibrium where no one defaults yet the possibility of foreclosure is real. In the Indonesian context, this is probably not the equilibrium. Borrowers rather than lenders are generally favoured. Foreclosure is a socially sensitive issue, and the legal practice of foreclosure in Indonesia is unpredictable and lengthy. These high costs of foreclosure most likely are at the expense of the borrowers, and streamlining the process could benefit all parties involved.

¹³ Statistik Potensi Desa (Podes) stands for Village Potential Statistics, a village-level and urban-neighbourhood-level economic census. We use the 2003 census, which was collected in 2002.

¹⁴ In 2002, Papua included the yet-to-be-formed province of West Papua. In West Java, all sampled households were rural.

¹⁵ We could instead use the fraction of the value of the household's total land assets that are titled. In practice, this distinction is almost irrelevant in our dataset; even though 352 of our households report having more than one plot, all but 57 of these households have all their plots either titled or untitled.

¹⁶ In our data, we have 680 distinct loans, 489 of which are from formal banks. Of these formal loans, 475 report information on how the loan was secured. The difference between 475 and 405 can be explained by our excluding the landless from our analysis. Non-formal loans consist of micro-bank loans (14%) and informal sources (12%).

¹⁷ The Indonesian CPI is available from Badan Pusat Statistik, the central statistics agency: http://www.bps.go.id/eng/aboutus.php?inflasi=1. Over the range of years in which the formal loans were given, there were three changes to the weighting of the CPI and to the sample of cities included in the cost-of-living surveys. To account for these changes, we have run robustness checks for our main results by including dummies for urban and rural CPI-weighting groups. The results are similar.

¹⁸ We dropped one loan-size observation because of its unrealistic loan amount of Rp 1,000 (less than $1).

¹⁹ More than half of all formal loans did not use any type of land document. Instead, the most commonly cited security was an advance against future salaries, such as fixed income or the use of a guarantor.

²⁰ We tested all continuous variables by applying both the conventional t-test (allowing for the variances between the samples to differ) and the non-parametric Wilcoxon rank-sum test. We used a test of proportionality for all binary variables. There were no significant differences at the 10% level.

²¹ Land that has been designated as forest (roughly 60% of land area) is handled by the Ministry of Forestry.

²² A small number of households had more than one loan in the sample because they had loans from multiple formal banks. In all loan-size regressions, we include a dummy variable that indicates these households.

²³ Reverse causality is another possible explanation; borrowing households may use the loan to fund obtaining a land title. Since most loans are recent loans (those in the past four years) and the titling process is lengthy, having obtained a loan is not likely to mean that the household has obtained a land title. Unfortunately, we do not have information on when the land title was obtained.

²⁴ Using locality-level fixed effects is problematic from a statistical point of view, both because of the loss in degrees of freedom and because there is little within-locality variation in the main explanatory variable, the titled first-time borrower dummy—44 of 72 villages had no titled first-time borrowers, compared with 14 of 36 sub-districts.

²⁵ If banks use methods other than securing collateral, then the information effect observed in comparing first-time and experienced borrowers may reflect ex-post or ex-ante informational constraints. In giving an example of an ex-ante informational effect, Bester (1985) shows that banks can screen for borrower types with low repayment costs by offering loan contracts with higher collateral requirements at a lower interest rate. The ex-post information effect could work by using reputation-based contractual enforcement in which formal titles improve the observability of default.

²⁶ Two-thirds of sub-districts had zero titled first-time borrowers who did not offer this land title as security for their loan (compared with one-third of sub-districts with zero titled first-time borrowers).

²⁷ We employ a Heckman selection model for this procedure.

²⁸ The demand variable is a generated regressor and the standard errors are not correct, so the significance level should be interpreted with caution. We do not make the correction here, because we wanted to check only whether the inclusion of the demand variable affected the coefficients on having a land title and the interaction term with first-time borrowers.