Safety Net Perception and its Effects on Household Investment in Developing Countries: Chemical Fertilizer Input by Cambodian Farmers

Abstract

Farmers in developing countries are reluctant to make investments for fear of failure and the economic distress resulting from this, but the perception of protection by a safety net may induce farmers to invest by reducing that fear. Using Cambodian farm household data, this paper examines factors affecting the perception of protection by a safety net and then assesses the effect of this perception on farmers' investment. For empirical analyses, perceived credit availability from relatives represents this perceived safety net; and chemical fertilizer input signifies the investment size. The results of the econometric analysis demonstrate that farm households with higher economic status, who are able to repay a loan or favour, are more likely to perceive such a safety net. It is also shown that the safety net perception positively affects chemical fertilizer input, implying that farmers accept risk when they perceive a safety net. These findings suggest that it is not only a lack of capital that deters poor households from investment, but also the perceived lack of a safety net.

Notes

¹ This issue is investigated theoretically by Feder (1980) and empirically by Shakya & Flinn (1985), Green & Ng'ong'ola (1993), Jha & Hojjatti (1993), Bisanda et al. (1998) and Ouma et al. (2001), among others.

² Jha & Hojatti (1993) used cooperative membership as an indicator of credit access, but this also seems to represent credit for investment because cooperatives usually provide loan funds for production.

³ For example, a dummy variable in Isham (2002) takes a value of one if households report that credit is available to them.

⁴ As of 2003, 100000 riels was equal to about US$25 ($1 = 4000 riels).

⁵ To see the degree of the risk of fertilizer application we need data of more than 1 year; but CSES contains the data of only 1 year of farming activities.

⁶ According to FAO data, a mere 4.9% of agricultural land was irrigated as of 2001 (FAO, 2008).

⁷ Return to fertilizer application could not be assessed using CSES data because information on some other input (such as labour input) is not available.

⁸ “Farm household” here is defined as a household whose head's primary occupation was listed as farmer.

⁹ Although neither the amount of chemical fertilizer used nor its price was recorded in the survey, 90 000 riels of chemical fertilizer is estimated as equivalent to two sacks (50 kg per sack) in amount, because the price of urea or DAP fertilizer was 40 000–50 000 riels per sack as of 2004.

¹⁰ We cannot present the figure of fertilizer input per fertilized area because the CSES data do not contain fertilized area.

¹¹ The local administration units in Cambodia are province (or municipality), district, commune, and village. According to the government definition as of 2004, urban areas correspond to a district with a provincial government office (provincial centre), four districts among seven districts in Phnom Penh municipality, and the entire area of the other three municipalities. All other areas are defined as rural.

¹² We used statistical software (STATA ver. 10) for estimation. The probit sample selection model of the debt function and perceived safety net function were estimated using the HECKPROB command, whereas the censored regression model of the chemical fertilizer input function and the planted area function were estimated using an author-written program that applies the ML command.

¹³ Actually, if we estimate a bivariate probit model of two dummy variables that represent whether a household has positive planted area of a certain category of planted area, the estimated correlation coefficient of the error terms of the two probit equations is non-zero and significant. This means that the set of the planted area functions must be estimated simultaneously.

¹⁴ In fact, we tried to estimate bivariate tobit models, in which the error terms are assumed to follow bivariate normal distribution, of two categories of planted area (such as rice versus non-rice crop or irrigated versus non-irrigated area) but failed to maximize the likelihood functions.

¹⁵ Estimating the fertilizer input function separately for each land type is also impossible because CSES does not include data on fertilizer input by land type.

¹⁶ Data on the percentage of cultivated area fertilized and the value of fertilizer applied per fertilized area are not collected in CSES.

¹⁷ As the perceived safety net is represented by whether one can borrow money from relatives, the relatives' economic status should be included as an explanatory variable. We were unable to do so because such data are not available.

¹⁸ We use the time span of “5 years ago” merely because of the availability of data.

¹⁹ In the rural Cambodian villages studied in Yagura (2008), for most married couples, both the husband and wife are natives of the village in which they live, implying that relatives generally live near each other.

²⁰ Regarding the ethnicity of household heads in this sample, Cham is the second largest ethnic group after Khmer. Although the survey data identify other ethnic groups such as Chinese and Vietnamese, we do not include dummy variables for those groups because of the small number of such households.

²¹ The provincial dummy variable is defined for each province and municipality, with Phnom Penh and Pailing municipalities set to zero. Exceptions are Mondul Kiri and Ratana Kiri provinces, which are combined because the number of sample households is too small.

²² Excluding SINGLE from the debt function is necessary to meet the exclusion restriction. In fact, SINGLE is not significant even when included in the debt function.

²³ In fact, land titling has yet to be widely implemented in Cambodia. Consequently, a large fraction of the total farmland area lacks an official title deed. To introduce a land titling system, the government once urged de facto landowners to apply for registration and issued a receipt for the application form. At least in some regions, this receipt is regarded as the title deed by local people (Yagura, 2008). In the present study, those who have such a receipt are also regarded as title holders.

²⁴ This means that the other 30% of land is irrigated, but the CSES questionnaire designates having any water source other than rain as “irrigated” even if the water supply is very small or irregular.

²⁵ Dercon (1996) shows this kind of effect of draught animals using Tanzanian data.

²⁶ In Cambodia, those who do not own draught animals must hire other farmers to plough land. In such cases, the hired farmer tends to put a higher priority on his own land.