28.01.2015 Views

The Future of Smallholder Farming in Eastern Africa - Uganda ...

The Future of Smallholder Farming in Eastern Africa - Uganda ...

The Future of Smallholder Farming in Eastern Africa - Uganda ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Table 7b--Unit root tests for first differences <strong>of</strong> prices <strong>in</strong> the postliberalization era<br />

Market p Criterion ADF test<br />

statistic<br />

Null hypothesis (unit<br />

root present)<br />

Kiritiri (no trend) 1 AIC, HQC -8.373 Reject<br />

Kiritiri (with trend) 1 AIC, HQC -8.365 Reject<br />

Siakago (no trend) 0 AIC, SBC, HQC -9.779 Reject<br />

Siakago (with trend) 0 AIC, SBC, HQC -9.812 Reject<br />

Kitale (no trend) 0 SBC, HQC -7.134 Reject<br />

Kitale (with trend) 0 SBC, HQC -7.083 Reject<br />

Nairobi (no trend) 0 AIC, SBC, HQC -6.116 Reject<br />

Nairobi (with trend) 0 AIC, SBC, HQC -6.101 Reject<br />

Migori (no trend) 0 AIC, SBC, HQC -9.143 Reject<br />

Migori (with trend) 0 AIC, SBC, HQC -9.082 Reject<br />

Awendo (no trend) 0 AIC, SBC, HQC -8.420 Reject<br />

Awendo (with trend) 0 AIC, SBC, HQC -8.360 Reject<br />

95% critical value for the ADF statistic (no trend) -2.902<br />

95% critical value for the ADF statistic (with -3.472<br />

trend)<br />

Source: Authors’ survey, 2001.<br />

To test for co<strong>in</strong>tegration, residuals were obta<strong>in</strong>ed from regressions <strong>of</strong> the price <strong>in</strong> market i,<br />

P i , t on the price <strong>in</strong> market j, P j , t as given already <strong>in</strong> equation 2. <strong>The</strong> residuals were then<br />

tested for the presence <strong>of</strong> unit roots. If the t-value for δ 1 is greater than the critical 95 percent<br />

level ADF statistic, the residuals are <strong>in</strong>tegrated <strong>of</strong> order zero and the prices are co<strong>in</strong>tegrated.<br />

Aga<strong>in</strong> as noted earlier, <strong>in</strong> theory, if market i is co<strong>in</strong>tegrated with market j, then market j<br />

should be co<strong>in</strong>tegrated with market i. In practice, however, test results may differ. For this<br />

reason the co<strong>in</strong>tegration test was repeated for all markets by <strong>in</strong>terchang<strong>in</strong>g the left-hand and<br />

right-hand price variables.<br />

Results are presented <strong>in</strong> Tables 8a and 8b with the first column show<strong>in</strong>g the dependent<br />

variable <strong>in</strong> the first regression and the first row represent<strong>in</strong>g the dependent variable <strong>of</strong> the<br />

second regression. Accord<strong>in</strong>g to Engle and Granger (1991) a l<strong>in</strong>k between market i and<br />

market j is said to be segmented if there is no co<strong>in</strong>tegration <strong>in</strong> either direction, or rather if<br />

regress<strong>in</strong>g series i on series j and regress<strong>in</strong>g j on i both yield nonstationary residuals. This<br />

implies that if there is co<strong>in</strong>tegration <strong>in</strong> at least one direction, then the l<strong>in</strong>k is considered to be<br />

<strong>in</strong>tegrated. Follow<strong>in</strong>g this argument, the co<strong>in</strong>tegration tests show that for the preliberalization<br />

sample the hypothesis <strong>of</strong> co<strong>in</strong>tegration <strong>of</strong> OLS residuals is rejected only for the Kiritiri-<br />

Siakago, Kitale-Nairobi, Kitale-Migori, and Nairobi-Migori l<strong>in</strong>ks. This implies that all other<br />

markets were not <strong>in</strong>tegrated <strong>in</strong> the preliberalization period. <strong>The</strong> result highlights the<br />

importance <strong>of</strong> the restriction that limited the amount <strong>of</strong> gra<strong>in</strong> that could be moved out <strong>of</strong> a<br />

district <strong>in</strong> the era <strong>of</strong> market controls. In the postliberalization period the hypothesis <strong>of</strong> no<br />

long-term relations among pairs <strong>of</strong> markets is rejected <strong>in</strong> all cases except for the Nairobi-<br />

Awendo, Siakago-Migori, and Siakago-Awendo. We concluded that only the three market<br />

l<strong>in</strong>ks listed were not co<strong>in</strong>tegrated <strong>in</strong> the postliberalization period considered <strong>in</strong> this study.<br />

Divid<strong>in</strong>g the sample <strong>in</strong>to preliberalization and postliberalization subsamples and carry<strong>in</strong>g out<br />

the appropriate co<strong>in</strong>tegration tests led us to observe that most markets became co<strong>in</strong>tegrated <strong>in</strong><br />

the postliberalization era. We conclude that liberalization improved the transmission <strong>of</strong> price<br />

signals among various markets <strong>in</strong> the country, thereby strengthen<strong>in</strong>g the l<strong>in</strong>ks between<br />

markets.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!