Menu
Home
People
Places
Arts
History
Plants & Animals
Science
Life & Culture
Technology
Reference.org
Chow test
open-in-new
<h2 id="illustrations">Illustrations</h2> Applications of the Chow test<table><tbody><tr><th>Structural break (slopes differ)</th><th>Program evaluation (intercepts differ)</th></tr><tr><td></td><td></td></tr><tr><td>At x = 1.7 {\displaystyle x=1.7} there is a structural break; separate regressions on the subintervals [ 0 , 1.7 ] {\displaystyle [0,1.7]} and [ 1.7 , 4 ] {\displaystyle [1.7,4]} delivers a better model than the combined regression (dashed) over the whole interval.</td><td>Comparison of two different programs (red, green) in a common data set: separate regressions for both programs deliver a better model than a combined regression (black).</td></tr></tbody></table> <h2 id="first-chow-test">First Chow Test</h2> <p>Suppose that we model our data as </p> y t = a + b x 1 t + c x 2 t + ε . {\displaystyle y_{t}=a+bx_{1t}+cx_{2t}+\varepsilon .\,} <p>If we split our data into two groups, then we have </p> y t = a 1 + b 1 x 1 t + c 1 x 2 t + ε {\displaystyle y_{t}=a_{1}+b_{1}x_{1t}+c_{1}x_{2t}+\varepsilon \,} <p>and </p> y t = a 2 + b 2 x 1 t + c 2 x 2 t + ε . {\displaystyle y_{t}=a_{2}+b_{2}x_{1t}+c_{2}x_{2t}+\varepsilon .\,} <p>The <a href="/facts/Null_hypothesis/C8NwSE9J">null hypothesis</a> of the Chow test asserts that a 1 = a 2 {\displaystyle a_{1}=a_{2}} , b 1 = b 2 {\displaystyle b_{1}=b_{2}} , and c 1 = c 2 {\displaystyle c_{1}=c_{2}} , and there is the assumption that the <a href="/facts/Errors_and_residuals_in_statistics/Ef0LgAS4">model errors</a> ε {\displaystyle \varepsilon } are <a href="/facts/Independent_and_identically_distributed/othIRaWt">independent and identically distributed</a> from a <a href="/facts/Normal_distribution/UapjjPyQ">normal distribution</a> with unknown <a href="/facts/Variance/ULBJKXD1">variance</a>. </p><p>Let S C {\displaystyle S_{C}} be the sum of squared <a href="/facts/Errors_and_residuals_in_statistics/Ef0LgAS4">residuals</a> from the combined data, S 1 {\displaystyle S_{1}} be the sum of squared residuals from the first group, and S 2 {\displaystyle S_{2}} be the sum of squared residuals from the second group. N 1 {\displaystyle N_{1}} and N 2 {\displaystyle N_{2}} are the number of observations in each group and k {\displaystyle k} is the total number of parameters (in this case 3, i.e. 2 independent variables coefficients + intercept). Then the Chow test statistic is </p> ( S C − ( S 1 + S 2 ) ) / k ( S 1 + S 2 ) / ( N 1 + N 2 − 2 k ) . {\displaystyle {\frac {(S_{C}-(S_{1}+S_{2}))/k}{(S_{1}+S_{2})/(N_{1}+N_{2}-2k)}}.} <p>The test statistic follows the <a href="/facts/F-distribution/XNuuJtWY"><i>F</i>-distribution</a> with k {\displaystyle k} and N 1 + N 2 − 2 k {\displaystyle N_{1}+N_{2}-2k} <a href="/facts/Degrees_of_freedom_(statistics)/fvwKCU86">degrees of freedom</a>. </p><p>The same result can be achieved via dummy variables. </p><p>Consider the two data sets which are being compared. Firstly there is the 'primary' data set i={1,..., n 1 {\displaystyle n_{1}} } and the 'secondary' data set i={ n 1 {\displaystyle n_{1}} +1,...,n}. Then there is the union of these two sets: i={1,...,n}. If there is no structural change between the primary and secondary data sets a regression can be run over the union without the issue of biased estimators arising. </p><p>Consider the regression: </p><p> y t = β 0 + β 1 x 1 t + β 2 x 2 t + . . . + β k x k t + γ 0 D t + ∑ i = 1 k γ i x i t D t + ε t . {\displaystyle y_{t}=\beta _{0}+\beta _{1}x_{1t}+\beta _{2}x_{2t}+...+\beta _{k}x_{kt}+\gamma _{0}D_{t}+\sum _{i=1}^{k}\gamma _{i}x_{it}D_{t}+\varepsilon _{t}.\,} </p><p>Which is run over i={1,...,n}. </p><p>D is a dummy variable taking a value of 1 for i={ n 1 {\displaystyle n_{1}} +1,...,n} and 0 otherwise. </p><p>If both data sets can be explained fully by ( β 0 , β 1 , . . . , β k ) {\displaystyle (\beta _{0},\beta _{1},...,\beta _{k})} then there is no use in the dummy variable as the data set is explained fully by the restricted equation. That is, under the assumption of no structural change we have a null and alternative hypothesis of: </p><p> H 0 : γ 0 = 0 , γ 1 = 0 , . . . , γ k = 0 {\displaystyle H_{0}:\gamma _{0}=0,\gamma _{1}=0,...,\gamma _{k}=0} </p><p> H 1 : otherwise {\displaystyle H_{1}:{\text{otherwise}}} </p><p>The null hypothesis of joint insignificance of D can be run as an F-test with n − 2 ( k + 1 ) {\displaystyle n-2(k+1)} degrees of freedom (DoF). That is: F = ( R S S R − R S S U ) / ( k + 1 ) R S S U / D o F {\displaystyle F={\frac {(RSS^{R}-RSS^{U})/(k+1)}{RSS^{U}/DoF}}} . </p><p>Remarks </p> <ul><li>The global sum of squares (SSE) is often called the Restricted Sum of Squares (RSSM) as we basically test a constrained model where we have 2 k {\displaystyle 2k} assumptions (with k {\displaystyle k} the number of regressors).</li> <li>Some software like SAS will use a predictive Chow test when the size of a subsample is less than the number of regressors.</li></ul> <ul><li>Chow, Gregory C. (1960). <a href="https://web.archive.org/web/20191228155733/http://pdfs.semanticscholar.org/0f70/219160c8ad2f9db02e226d3f7d7320e729b8.pdf">"Tests of Equality Between Sets of Coefficients in Two Linear Regressions"</a> (PDF). <i>Econometrica</i>. 28 (3): 591–605. <a href="/facts/Doi_(identifier)/muM9Etpq">doi</a>:<a href="https://doi.org/10.2307%2F1910133">10.2307/1910133</a>. <a href="/facts/JSTOR_(identifier)/YTeVmaJ7">JSTOR</a> <a href="https://www.jstor.org/stable/1910133">1910133</a>. <a href="/facts/S2CID_(identifier)/ldJsHa2Y">S2CID</a> <a href="https://api.semanticscholar.org/CorpusID:116311724">116311724</a>. Archived from <a href="http://pdfs.semanticscholar.org/0f70/219160c8ad2f9db02e226d3f7d7320e729b8.pdf">the original</a> (PDF) on 2019-12-28.</li> <li>Doran, Howard E. (1989). <i>Applied Regression Analysis in Econometrics</i>. CRC Press. p. 146. <a href="/facts/ISBN_(identifier)/15AdSPa9">ISBN</a> 978-0-8247-8049-4.</li> <li>Dougherty, Christopher (2007). <i>Introduction to Econometrics</i>. Oxford University Press. p. 194. <a href="/facts/ISBN_(identifier)/15AdSPa9">ISBN</a> 978-0-19-928096-4.</li> <li><a href="/facts/Jan_Kmenta/RKLurjPH">Kmenta, Jan</a> (1986). <a href="https://archive.org/details/elementsofeconom0003kmen"><i>Elements of Econometrics</i></a> (Second ed.). New York: Macmillan. pp. <a href="https://archive.org/details/elementsofeconom0003kmen/page/412">412–423</a>. <a href="/facts/ISBN_(identifier)/15AdSPa9">ISBN</a> 978-0-472-10886-2.</li> <li><a href="/facts/Jeffrey_Wooldridge/XdwAoA1m">Wooldridge, Jeffrey M.</a> (2009). <i>Introductory Econometrics: A Modern Approach</i> (Fourth ed.). Mason: South-Western. pp. 243–246. <a href="/facts/ISBN_(identifier)/15AdSPa9">ISBN</a> 978-0-324-66054-8.</li></ul> <h2 id="external-links">External links</h2> Wikimedia Commons has media related to Chow test. <ul><li><a href="https://www.stata.com/support/faqs/stat/chow.html">Computing the Chow statistic</a>, <a href="https://www.stata.com/support/faqs/stat/chow2.html">Chow and Wald tests</a>, <a href="https://www.stata.com/support/faqs/stat/chow3.html">Chow tests</a>: Series of FAQ explanations from the <a href="/facts/Stata/7Sx8QoJH">Stata</a> Corporation at <a href="https://www.stata.com/support/faqs/">https://www.stata.com/support/faqs/</a></li> <li><a href="http://support.sas.com/documentation/cdl/en/etsug/60372/HTML/default/viewer.htm#etsug_model_sect051.htm">[1]</a>: Series of FAQ explanations from the <a href="/facts/SAS_Institute/8AgnAKbf">SAS</a> Corporation</li></ul>