ABSTRACT
This study analyses asymmetry in price transmission between wholesale and retail rice markets in Sri Lanka, using the threshold autoregressive model. We found that the wholesale and the retail rice markets in Sri Lanka are integrated, with price changes moving from the wholesale to the retail market. However, the price transmission process is asymmetric. In particular, price increases at the wholesale market transmit immediately to the retail market while price decreases transmit more slowly. Parameter stability test and follow-up analysis indicated that the price transmission process is asymmetric only during periods of price surges, suggesting that the rice market is not efficient during these periods.
Disclosure statement
No potential conflict of interest was reported by the authors.
Notes
1 Market integration generally refers to the fact that traders in one market are able to adjust their prices and volumes in response to the changes originated in the other market. Initially, researchers believed that market integration was the same as co-movement of prices in two markets. However, with time, this belief was questioned by Ravallion (Citation1986), Baulch (Citation1997), Goodwin and Piggott (Citation2001), Richardson (Citation1978), Moser, Barrett, and Minten (Citation2005), Silvapulle and Jayasuriya (Citation1994) and Alexander and Wyeth (Citation1994).
2 Nadu is one of the main rice varieties consumed by Sri Lankans. Per capita monthly consumption is around 3.3 kg in 2012/13.
3 Pettah market is the main wholesale and retail rice market in Sri Lanka.
4 See Appendix for a list of empirical studies that have analysed the presence of asymmetry in agricultural markets.
5 Usually, the Granger causality test is performed on stationary time-series variables. Thus, in performing the Granger causality test for this data set, we first convert the data into first deference and then perform the test. Alternatively, we could have estimated a Vector Autoregressive model and then executed the Granger causality test. Both methods yield the same results.
6 Bai–Perron test is another test for detecting structural breaks. This test provides the exact point of time where the structural breaks exist. However, in this study, this test is not applicable for the M-TAR model due to singularity in the model. Nevertheless, we performed the Bai–Perron test for the linear model (Equation 1) and confirmed that there are two structural breaks in the data set: one at the end of 2007 and the second at the beginning of 2010.