New cointegration tests using signs of the variables as instrumental variables have been developed. These tests have a standard normal distribution and are not affected by the number of variables when there is no cointegration. Through simulations, it was found that these tests have consistent accuracy and are more powerful, especially with small amounts of data.